result: per-paper CSVs

result/per_paper/2105.06643/accuracy_efficiency.csv

@@ -0,0 +1 @@
+metric_kind,entity,model_variant,metric,value,uncertainty

result/per_paper/2105.06643/accuracy_efficiency_traced.csv

@@ -0,0 +1 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col

result/per_paper/2105.06643/components_architecture.csv

@@ -0,0 +1,6 @@
+component,what_it_is,provenance,citation,evidence
+Monash Time Series Forecasting Archive,"A comprehensive time series forecasting archive containing 20 publicly available datasets with equal and variable lengths, including real-world and competition datasets across varied domains.",proposed_here,,"We introduce the first comprehensive time series forecasting archive containing datasets of related time series, available at https://forecastingdata.org/. This archive contains 20 publicly available time series datasets..."
+.tsf File Format,"A new time series data storage format based on the Weka ARFF file format, designed to store meta-information and series-specific details in a non-redundant, flexible manner.",proposed_here,"Paynter et al., 2008; Löning et al., 2019","We introduce a new format to store time series data, based on the Weka ARFF file format (Paynter et al., 2008) and overcoming some of the shortcomings we observe in the .ts format used in the sktime time series repository (Löning et al., 2019). We use a .tsf extension for this new format."
+tsfeatures,"A feature extraction method for time series data, used to analyze characteristics of datasets and identify similarities/differences among series.",reused_cited,"Hyndman et al., 2020","We analyse the characteristics of different series to identify the similarities and differences among them. For that, we conduct a feature analysis using tsfeatures (Hyndman et al., 2020)..."
+catch22 Features,"A set of 22 time series features designed for classification tasks, used here for feature analysis of the datasets.",reused_cited,"Lubba et al., 2019","We analyse the characteristics of different series to identify the similarities and differences among them. For that, we conduct a feature analysis using ... catch22 features (Lubba et al., 2019) extracted from all series of all datasets."
+Baseline Forecasting Evaluation Framework,"A framework for evaluating baseline forecasting models (traditional univariate and global models) across eight error metrics, with publicly available forecasts and results.",proposed_here,,We evaluate the performance of a set of baseline forecasting models including both traditional univariate forecasting models and global forecasting models over all datasets across eight error metrics. The forecasts and evaluation results of the baseline methods are publicly available...

result/per_paper/2105.06643/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,not_reported
+num_devices,,,not_reported,not_reported
+training_cost,,,not_reported,not_reported
+training_batch_size,,,not_reported,not_reported
+training_steps_or_epochs,,,not_reported,not_reported
+precision,,,not_reported,not_reported
+inference_latency,,,not_reported,not_reported
+inference_throughput,,,not_reported,not_reported
+peak_memory,,,not_reported,not_reported
+flops_or_macs,,,not_reported,not_reported
+num_inference_samples,,,not_reported,not_reported
+params,,,not_reported,not_reported
+context_lengths_evaluated,,,not_reported,not_reported
+horizon_lengths_evaluated,,,not_reported,not_reported
+inference_batch_size,,,not_reported,not_reported

result/per_paper/2302.11939/accuracy_efficiency.csv

@@ -0,0 +1,98 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,Weather,GPT4TS / One-Fits-All (Large),MSE,0.221,
+accuracy,Weather,GPT4TS / One-Fits-All (Small),MSE,0.297,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.342,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.318,
+accuracy,ETTh1,GPT4TS / One-Fits-All (Large),MSE,0.613,
+accuracy,ETTh1,GPT4TS / One-Fits-All (Small),MSE,0.552,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.596,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.915,
+accuracy,ETTh2,GPT4TS / One-Fits-All (Large),MSE,0.413,
+accuracy,ETTh2,GPT4TS / One-Fits-All (Small),MSE,0.451,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.499,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.462,
+accuracy,ETTm1,GPT4TS / One-Fits-All (Large),MSE,0.774,
+accuracy,ETTm1,GPT4TS / One-Fits-All (Small),MSE,0.614,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.628,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.797,
+accuracy,ETTm2,GPT4TS / One-Fits-All (Large),MSE,0.352,
+accuracy,ETTm2,GPT4TS / One-Fits-All (Small),MSE,0.454,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.930,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.332,
+accuracy,ECL,GPT4TS / One-Fits-All (Large),MSE,0.261,
+accuracy,ECL,GPT4TS / One-Fits-All (Small),MSE,0.348,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.478,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.444,
+accuracy,Traffic,GPT4TS / One-Fits-All (Large),MSE,0.672,
+accuracy,Traffic,GPT4TS / One-Fits-All (Small),MSE,0.405,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.446,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,1.453,
+accuracy,Average,GPT4TS / One-Fits-All (Large),MSE,0.559,
+accuracy,Average,GPT4TS / One-Fits-All (Small),MSE,0.674,
+accuracy,Weather,GPT4TS / One-Fits-All (Large),MSE,0.227,
+accuracy,Weather,GPT4TS / One-Fits-All (Small),MSE,0.299,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.353,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.327,
+accuracy,ETTh1,GPT4TS / One-Fits-All (Large),MSE,0.681,
+accuracy,ETTh1,GPT4TS / One-Fits-All (Small),MSE,0.570,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.598,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.943,
+accuracy,ETTh2,GPT4TS / One-Fits-All (Large),MSE,0.428,
+accuracy,ETTh2,GPT4TS / One-Fits-All (Small),MSE,0.468,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.489,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.470,
+accuracy,ETTm1,GPT4TS / One-Fits-All (Large),MSE,0.726,
+accuracy,ETTm1,GPT4TS / One-Fits-All (Small),MSE,0.578,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.620,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.857,
+accuracy,ETTm2,GPT4TS / One-Fits-All (Large),MSE,0.232,
+accuracy,ETTm2,GPT4TS / One-Fits-All (Small),MSE,0.322,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.433,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.341,
+accuracy,ECL,GPT4TS / One-Fits-All (Large),MSE,0.297,
+accuracy,ECL,GPT4TS / One-Fits-All (Small),MSE,0.367,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.404,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.627,
+accuracy,Traffic,GPT4TS / One-Fits-All (Large),MSE,0.795,
+accuracy,Traffic,GPT4TS / One-Fits-All (Small),MSE,0.481,
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.502,
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,1.526,
+accuracy,ETTm1,GPT2(3),MSE,0.017,
+accuracy,Avg,GPT2(3),MSE,0.105,
+accuracy,ETTm2,GPT2(3),MSE,0.017,
+accuracy,Avg,GPT2(3),MSE,0.084,
+accuracy,ETTh1,GPT2(3),MSE,0.043,
+accuracy,Avg,GPT2(3),MSE,0.173,
+accuracy,ETTh2,GPT2(3),MSE,0.039,
+accuracy,Avg,GPT2(3),MSE,0.141,
+accuracy,ECL,GPT2(3),MSE,0.080,
+accuracy,Avg,GPT2(3),MSE,0.207,
+accuracy,Weather,GPT2(3),MSE,0.026,
+accuracy,Avg,GPT2(3),MSE,0.056,
+accuracy,GPT(6),GPT4TS / One-Fits-All (6),F1,6,
+accuracy,EthanolConcentration,GPT4TS / One-Fits-All (6),Imputation Error,34.2,
+accuracy,FaceDetection,GPT4TS / One-Fits-All (6),Imputation Error,69.2,
+accuracy,Handwriting,GPT4TS / One-Fits-All (6),Imputation Error,32.7,
+accuracy,Heartbeat,GPT4TS / One-Fits-All (6),Imputation Error,77.2,
+accuracy,JapaneseVowels,GPT4TS / One-Fits-All (6),Imputation Error,98.6,
+accuracy,PEMS-SF,GPT4TS / One-Fits-All (6),Imputation Error,87.9,
+accuracy,SelfRegulationSCP1,GPT4TS / One-Fits-All (6),Imputation Error,93.2,
+accuracy,SelfRegulationSCP2,GPT4TS / One-Fits-All (6),Imputation Error,59.4,
+accuracy,SpokenArabicDigits,GPT4TS / One-Fits-All (6),Imputation Error,99.2,
+accuracy,UWaveGestureLibrary,GPT4TS / One-Fits-All (6),Imputation Error,88.1,
+accuracy,Average,GPT4TS / One-Fits-All (6),Imputation Error,74.0,
+accuracy,Yearly Quarterly,GPT4TS / One-Fits-All (6),SMAPE,13.531,
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,2.996,
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.786,
+accuracy,SMAPE,GPT4TS / One-Fits-All (6),SMAPE,10.100,
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,1.182,
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.890,
+accuracy,Monthly Others,GPT4TS / One-Fits-All (6),SMAPE,12.894,
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,0.933,
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.878,
+accuracy,SMAPE,GPT4TS / One-Fits-All (6),SMAPE,4.891,
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,3.302,
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,1.035,
+accuracy,Average,GPT4TS / One-Fits-All (6),SMAPE,11.991,
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,1.585,
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.851,

result/per_paper/2302.11939/accuracy_efficiency_traced.csv

@@ -0,0 +1,98 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,Weather,GPT4TS / One-Fits-All (Large),MSE,0.221,,1,2,14
+accuracy,Weather,GPT4TS / One-Fits-All (Small),MSE,0.297,,1,2,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.342,,1,6,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.318,,1,6,15
+accuracy,ETTh1,GPT4TS / One-Fits-All (Large),MSE,0.613,,1,7,14
+accuracy,ETTh1,GPT4TS / One-Fits-All (Small),MSE,0.552,,1,7,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.596,,1,11,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.915,,1,11,15
+accuracy,ETTh2,GPT4TS / One-Fits-All (Large),MSE,0.413,,1,12,14
+accuracy,ETTh2,GPT4TS / One-Fits-All (Small),MSE,0.451,,1,12,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.499,,1,16,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.462,,1,16,15
+accuracy,ETTm1,GPT4TS / One-Fits-All (Large),MSE,0.774,,1,17,14
+accuracy,ETTm1,GPT4TS / One-Fits-All (Small),MSE,0.614,,1,17,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.628,,1,21,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.797,,1,21,15
+accuracy,ETTm2,GPT4TS / One-Fits-All (Large),MSE,0.352,,1,22,14
+accuracy,ETTm2,GPT4TS / One-Fits-All (Small),MSE,0.454,,1,22,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.930,,1,26,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.332,,1,26,15
+accuracy,ECL,GPT4TS / One-Fits-All (Large),MSE,0.261,,1,27,14
+accuracy,ECL,GPT4TS / One-Fits-All (Small),MSE,0.348,,1,27,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.478,,1,31,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.444,,1,31,15
+accuracy,Traffic,GPT4TS / One-Fits-All (Large),MSE,0.672,,1,32,14
+accuracy,Traffic,GPT4TS / One-Fits-All (Small),MSE,0.405,,1,32,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.446,,1,36,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,1.453,,1,36,15
+accuracy,Average,GPT4TS / One-Fits-All (Large),MSE,0.559,,1,37,14
+accuracy,Average,GPT4TS / One-Fits-All (Small),MSE,0.674,,1,37,15
+accuracy,Weather,GPT4TS / One-Fits-All (Large),MSE,0.227,,2,2,14
+accuracy,Weather,GPT4TS / One-Fits-All (Small),MSE,0.299,,2,2,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.353,,2,6,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.327,,2,6,15
+accuracy,ETTh1,GPT4TS / One-Fits-All (Large),MSE,0.681,,2,7,14
+accuracy,ETTh1,GPT4TS / One-Fits-All (Small),MSE,0.570,,2,7,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.598,,2,11,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.943,,2,11,15
+accuracy,ETTh2,GPT4TS / One-Fits-All (Large),MSE,0.428,,2,12,14
+accuracy,ETTh2,GPT4TS / One-Fits-All (Small),MSE,0.468,,2,12,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.489,,2,16,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.470,,2,16,15
+accuracy,ETTm1,GPT4TS / One-Fits-All (Large),MSE,0.726,,2,17,14
+accuracy,ETTm1,GPT4TS / One-Fits-All (Small),MSE,0.578,,2,17,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.620,,2,21,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.857,,2,21,15
+accuracy,ETTm2,GPT4TS / One-Fits-All (Large),MSE,0.232,,2,22,14
+accuracy,ETTm2,GPT4TS / One-Fits-All (Small),MSE,0.322,,2,22,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.433,,2,26,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.341,,2,26,15
+accuracy,ECL,GPT4TS / One-Fits-All (Large),MSE,0.297,,2,27,14
+accuracy,ECL,GPT4TS / One-Fits-All (Small),MSE,0.367,,2,27,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.404,,2,31,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,0.627,,2,31,15
+accuracy,Traffic,GPT4TS / One-Fits-All (Large),MSE,0.795,,2,32,14
+accuracy,Traffic,GPT4TS / One-Fits-All (Small),MSE,0.481,,2,32,15
+accuracy,Avg.,GPT4TS / One-Fits-All (Large),MSE,0.502,,2,36,14
+accuracy,Avg.,GPT4TS / One-Fits-All (Small),MSE,1.526,,2,36,15
+accuracy,ETTm1,GPT2(3),MSE,0.017,,3,2,2
+accuracy,Avg,GPT2(3),MSE,0.105,,3,6,2
+accuracy,ETTm2,GPT2(3),MSE,0.017,,3,7,2
+accuracy,Avg,GPT2(3),MSE,0.084,,3,11,2
+accuracy,ETTh1,GPT2(3),MSE,0.043,,3,12,2
+accuracy,Avg,GPT2(3),MSE,0.173,,3,16,2
+accuracy,ETTh2,GPT2(3),MSE,0.039,,3,17,2
+accuracy,Avg,GPT2(3),MSE,0.141,,3,21,2
+accuracy,ECL,GPT2(3),MSE,0.080,,3,22,2
+accuracy,Avg,GPT2(3),MSE,0.207,,3,26,2
+accuracy,Weather,GPT2(3),MSE,0.026,,3,27,2
+accuracy,Avg,GPT2(3),MSE,0.056,,3,31,2
+accuracy,GPT(6),GPT4TS / One-Fits-All (6),F1,6,,4,2,0
+accuracy,EthanolConcentration,GPT4TS / One-Fits-All (6),Imputation Error,34.2,,6,2,18
+accuracy,FaceDetection,GPT4TS / One-Fits-All (6),Imputation Error,69.2,,6,3,18
+accuracy,Handwriting,GPT4TS / One-Fits-All (6),Imputation Error,32.7,,6,4,18
+accuracy,Heartbeat,GPT4TS / One-Fits-All (6),Imputation Error,77.2,,6,5,18
+accuracy,JapaneseVowels,GPT4TS / One-Fits-All (6),Imputation Error,98.6,,6,6,18
+accuracy,PEMS-SF,GPT4TS / One-Fits-All (6),Imputation Error,87.9,,6,7,18
+accuracy,SelfRegulationSCP1,GPT4TS / One-Fits-All (6),Imputation Error,93.2,,6,8,18
+accuracy,SelfRegulationSCP2,GPT4TS / One-Fits-All (6),Imputation Error,59.4,,6,9,18
+accuracy,SpokenArabicDigits,GPT4TS / One-Fits-All (6),Imputation Error,99.2,,6,10,18
+accuracy,UWaveGestureLibrary,GPT4TS / One-Fits-All (6),Imputation Error,88.1,,6,11,18
+accuracy,Average,GPT4TS / One-Fits-All (6),Imputation Error,74.0,,6,12,18
+accuracy,Yearly Quarterly,GPT4TS / One-Fits-All (6),SMAPE,13.531,,7,1,2
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,2.996,,7,2,2
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.786,,7,3,2
+accuracy,SMAPE,GPT4TS / One-Fits-All (6),SMAPE,10.100,,7,4,2
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,1.182,,7,5,2
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.890,,7,6,2
+accuracy,Monthly Others,GPT4TS / One-Fits-All (6),SMAPE,12.894,,7,7,2
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,0.933,,7,8,2
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.878,,7,9,2
+accuracy,SMAPE,GPT4TS / One-Fits-All (6),SMAPE,4.891,,7,10,2
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,3.302,,7,11,2
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,1.035,,7,12,2
+accuracy,Average,GPT4TS / One-Fits-All (6),SMAPE,11.991,,7,13,2
+accuracy,MASE,GPT4TS / One-Fits-All (6),SMAPE,1.585,,7,14,2
+accuracy,OWA,GPT4TS / One-Fits-All (6),SMAPE,0.851,,7,15,2

result/per_paper/2302.11939/components_architecture.csv

@@ -0,0 +1,7 @@
+component,what_it_is,provenance,citation,evidence
+Frozen Pretrained Transformer (FPT),"A pre-trained transformer model (e.g., GPT2) with frozen self-attention and feedforward layers, adapted for time series analysis through fine-tuning.",proposed_here,"Zhou et al., 2023","This model, known as the Frozen Pretrained Transformer (FPT), is evaluated through fine-tuning on all major types of tasks involving time series."
+Self-Attention Mechanism,A module in the transformer architecture that captures long-range dependencies in time series data by computing attention weights across all positions.,reused_cited,"Vaswani et al., 2017","The self-attention module behaviors similarly to principle component analysis (PCA), an observation that helps explains how transformer bridges the domain gap."
+Feedforward Layers,Non-linear transformations applied to the output of self-attention mechanisms in each residual block of the transformer.,reused_cited,"Vaswani et al., 2017",We refrain from altering the self-attention and feedforward layers of the residual blocks in the pre-trained language or image model.
+Residual Blocks,Structural units in the transformer that combine self-attention and feedforward layers with skip connections to enable deeper network training.,reused_cited,"Vaswani et al., 2017",We refrain from altering the self-attention and feedforward layers of the residual blocks in the pre-trained language or image model.
+Fine-Tuning Mechanism,"A process of adapting the pre-trained model's parameters (excluding frozen layers) to specific time series tasks (e.g., forecasting, classification).",reused_cited,"Devlin et al., 2019",This model [...] is evaluated through fine-tuning on all major types of tasks involving time series.
+PCA Analogy for Self-Attention,A theoretical insight comparing the self-attention module's behavior to principal component analysis (PCA) for dimensionality reduction in time series.,proposed_here,"Zhou et al., 2023",We also found both theoretically and empirically that the self-attention module behaviors similarly to principle component analysis (PCA).

result/per_paper/2302.11939/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,
+num_devices,,,not_reported,
+training_cost,,,not_reported,
+training_batch_size,,,not_reported,
+training_steps_or_epochs,,,not_reported,
+precision,,,not_reported,
+inference_latency,,,not_reported,
+inference_throughput,,,not_reported,
+peak_memory,,,not_reported,
+flops_or_macs,,,not_reported,
+num_inference_samples,,,not_reported,
+params,,,not_reported,
+context_lengths_evaluated,,,not_reported,
+horizon_lengths_evaluated,,,not_reported,
+inference_batch_size,,,not_reported,

result/per_paper/2310.01728/accuracy_efficiency.csv

@@ -0,0 +1,90 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,ETTh1,Time-LLM,MSE,0.613,
+accuracy,Avg,Time-LLM,MSE,0.596,
+accuracy,ETTh2,Time-LLM,MSE,0.413,
+accuracy,Avg,Time-LLM,MSE,0.499,
+accuracy,ETTm1,Time-LLM,MSE,0.774,
+accuracy,Avg,Time-LLM,MSE,0.628,
+accuracy,ETTm2,Time-LLM,MSE,0.352,
+accuracy,Avg,Time-LLM,MSE,0.930,
+accuracy,Weather,Time-LLM,MSE,0.221,
+accuracy,Avg,Time-LLM,MSE,0.342,
+accuracy,Electricity,Time-LLM,MSE,0.261,
+accuracy,Avg,Time-LLM,MSE,0.478,
+accuracy,Traffic,Time-LLM,MSE,0.672,
+accuracy,Avg,Time-LLM,MSE,0.446,
+accuracy,ETTh1,Time-LLM,MSE,0.681,
+accuracy,Avg,Time-LLM,MSE,0.598,
+accuracy,ETTh2,Time-LLM,MSE,0.428,
+accuracy,Avg,Time-LLM,MSE,0.457,
+accuracy,ETTm1,Time-LLM,MSE,0.726,
+accuracy,Avg,Time-LLM,MSE,0.620,
+accuracy,ETTm2,Time-LLM,MSE,0.232,
+accuracy,Avg,Time-LLM,MSE,0.433,
+accuracy,Weather,Time-LLM,MSE,0.227,
+accuracy,Avg,Time-LLM,MSE,0.353,
+accuracy,Electricity,Time-LLM,MSE,0.297,
+accuracy,Avg,Time-LLM,MSE,0.404,
+accuracy,Traffic,Time-LLM,MSE,0.795,
+accuracy,Avg,Time-LLM,MSE,0.502,
+accuracy,$ETTh1 \rightarrow ETTh2$,Time-LLM,MSE,0.400,
+accuracy,Avg,Time-LLM,MSE,0.380,
+accuracy,$ETTh1 \rightarrow ETTm2$,Time-LLM,MSE,0.357,
+accuracy,Avg,Time-LLM,MSE,0.314,
+accuracy,$ETTh2 \rightarrow ETTh1$,Time-LLM,MSE,0.555,
+accuracy,Avg,Time-LLM,MSE,0.565,
+accuracy,$ETTh2 \rightarrow ETTm2$,Time-LLM,MSE,0.336,
+accuracy,Avg,Time-LLM,MSE,0.325,
+accuracy,$ETTm1 \rightarrow ETTh2$,Time-LLM,MSE,0.415,
+accuracy,Avg,Time-LLM,MSE,0.439,
+accuracy,$ETTm1 \rightarrow ETTm2$,Time-LLM,MSE,0.314,
+accuracy,Avg,Time-LLM,MSE,0.296,
+accuracy,$ETTm2 \rightarrow ETTh2$,Time-LLM,MSE,0.391,
+accuracy,Avg,Time-LLM,MSE,0.409,
+accuracy,$ETTm2 \rightarrow ETTm1$,Time-LLM,MSE,0.490,
+accuracy,Avg,Time-LLM,MSE,0.568,
+accuracy,ETTh1,Time-LLM,MSE,0.933,
+accuracy,Avg,Time-LLM,MSE,0.656,
+accuracy,ETTh2,Time-LLM,MSE,0.390,
+accuracy,Avg,Time-LLM,MSE,0.532,
+accuracy,ETTm1,Time-LLM,MSE,1.091,
+accuracy,Avg,Time-LLM,MSE,0.697,
+accuracy,ETTm2,Time-LLM,MSE,0.435,
+accuracy,Avg,Time-LLM,MSE,0.634,
+accuracy,Weather,Time-LLM,MSE,0.255,
+accuracy,Avg,Time-LLM,MSE,0.448,
+accuracy,Electricity,Time-LLM,MSE,0.520,
+accuracy,Avg,Time-LLM,MSE,0.510,
+accuracy,Traffic,Time-LLM,MSE,1.068,
+accuracy,Avg,Time-LLM,MSE,0.765,
+accuracy,ILI,Time-LLM,MSE,4.909,
+accuracy,Avg,Time-LLM,MSE,1.346,
+accuracy,Yearly,Time-LLM,SMAPE,13.419,
+accuracy,MASE,Time-LLM,SMAPE,3.565,
+accuracy,OWA,Time-LLM,SMAPE,0.911,
+accuracy,Quarterly,Time-LLM,SMAPE,10.110,
+accuracy,MASE,Time-LLM,SMAPE,1.253,
+accuracy,OWA,Time-LLM,SMAPE,0.938,
+accuracy,Monthly,Time-LLM,SMAPE,12.980,
+accuracy,MASE,Time-LLM,SMAPE,1.003,
+accuracy,OWA,Time-LLM,SMAPE,0.931,
+accuracy,Others,Time-LLM,SMAPE,4.795,
+accuracy,MASE,Time-LLM,SMAPE,4.116,
+accuracy,OWA,Time-LLM,SMAPE,1.259,
+accuracy,Average,Time-LLM,SMAPE,11.983,
+accuracy,MASE,Time-LLM,SMAPE,1.808,
+accuracy,OWA,Time-LLM,SMAPE,0.94,
+accuracy,ETTh1,Time-LLM,MSE,0.556,
+accuracy,ETTh1,Time-LLM,MSE,0.522,
+accuracy,ETTh2,Time-LLM,MSE,0.370,
+accuracy,ETTh2,Time-LLM,MSE,0.394,
+accuracy,ETTm1,Time-LLM,MSE,0.404,
+accuracy,ETTm1,Time-LLM,MSE,0.427,
+accuracy,ETTm2,Time-LLM,MSE,0.277,
+accuracy,ETTm2,Time-LLM,MSE,0.323,
+accuracy,Weather,Time-LLM,MSE,0.234,
+accuracy,Weather,Time-LLM,MSE,0.273,
+accuracy,ECL,Time-LLM,MSE,0.175,
+accuracy,ECL,Time-LLM,MSE,0.270,
+accuracy,Traffic,Time-LLM,MSE,0.429,
+accuracy,Traffic,Time-LLM,MSE,0.306,

result/per_paper/2310.01728/accuracy_efficiency_traced.csv

@@ -0,0 +1,90 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,ETTh1,Time-LLM,MSE,0.613,,1,2,14
+accuracy,Avg,Time-LLM,MSE,0.596,,1,6,14
+accuracy,ETTh2,Time-LLM,MSE,0.413,,1,7,14
+accuracy,Avg,Time-LLM,MSE,0.499,,1,11,14
+accuracy,ETTm1,Time-LLM,MSE,0.774,,1,12,14
+accuracy,Avg,Time-LLM,MSE,0.628,,1,16,14
+accuracy,ETTm2,Time-LLM,MSE,0.352,,1,17,14
+accuracy,Avg,Time-LLM,MSE,0.930,,1,21,14
+accuracy,Weather,Time-LLM,MSE,0.221,,1,22,14
+accuracy,Avg,Time-LLM,MSE,0.342,,1,26,14
+accuracy,Electricity,Time-LLM,MSE,0.261,,1,27,14
+accuracy,Avg,Time-LLM,MSE,0.478,,1,31,14
+accuracy,Traffic,Time-LLM,MSE,0.672,,1,32,14
+accuracy,Avg,Time-LLM,MSE,0.446,,1,36,14
+accuracy,ETTh1,Time-LLM,MSE,0.681,,2,2,14
+accuracy,Avg,Time-LLM,MSE,0.598,,2,6,14
+accuracy,ETTh2,Time-LLM,MSE,0.428,,2,7,14
+accuracy,Avg,Time-LLM,MSE,0.457,,2,11,14
+accuracy,ETTm1,Time-LLM,MSE,0.726,,2,12,14
+accuracy,Avg,Time-LLM,MSE,0.620,,2,16,14
+accuracy,ETTm2,Time-LLM,MSE,0.232,,2,17,14
+accuracy,Avg,Time-LLM,MSE,0.433,,2,21,14
+accuracy,Weather,Time-LLM,MSE,0.227,,2,22,14
+accuracy,Avg,Time-LLM,MSE,0.353,,2,26,14
+accuracy,Electricity,Time-LLM,MSE,0.297,,2,27,14
+accuracy,Avg,Time-LLM,MSE,0.404,,2,31,14
+accuracy,Traffic,Time-LLM,MSE,0.795,,2,32,14
+accuracy,Avg,Time-LLM,MSE,0.502,,2,36,14
+accuracy,$ETTh1 \rightarrow ETTh2$,Time-LLM,MSE,0.400,,3,2,9
+accuracy,Avg,Time-LLM,MSE,0.380,,3,6,9
+accuracy,$ETTh1 \rightarrow ETTm2$,Time-LLM,MSE,0.357,,3,7,9
+accuracy,Avg,Time-LLM,MSE,0.314,,3,11,9
+accuracy,$ETTh2 \rightarrow ETTh1$,Time-LLM,MSE,0.555,,3,12,9
+accuracy,Avg,Time-LLM,MSE,0.565,,3,16,9
+accuracy,$ETTh2 \rightarrow ETTm2$,Time-LLM,MSE,0.336,,3,17,9
+accuracy,Avg,Time-LLM,MSE,0.325,,3,21,9
+accuracy,$ETTm1 \rightarrow ETTh2$,Time-LLM,MSE,0.415,,3,22,9
+accuracy,Avg,Time-LLM,MSE,0.439,,3,26,9
+accuracy,$ETTm1 \rightarrow ETTm2$,Time-LLM,MSE,0.314,,3,27,9
+accuracy,Avg,Time-LLM,MSE,0.296,,3,31,9
+accuracy,$ETTm2 \rightarrow ETTh2$,Time-LLM,MSE,0.391,,3,32,9
+accuracy,Avg,Time-LLM,MSE,0.409,,3,36,9
+accuracy,$ETTm2 \rightarrow ETTm1$,Time-LLM,MSE,0.490,,3,37,9
+accuracy,Avg,Time-LLM,MSE,0.568,,3,41,9
+accuracy,ETTh1,Time-LLM,MSE,0.933,,4,2,8
+accuracy,Avg,Time-LLM,MSE,0.656,,4,6,8
+accuracy,ETTh2,Time-LLM,MSE,0.390,,4,7,8
+accuracy,Avg,Time-LLM,MSE,0.532,,4,11,8
+accuracy,ETTm1,Time-LLM,MSE,1.091,,4,12,8
+accuracy,Avg,Time-LLM,MSE,0.697,,4,16,8
+accuracy,ETTm2,Time-LLM,MSE,0.435,,4,17,8
+accuracy,Avg,Time-LLM,MSE,0.634,,4,21,8
+accuracy,Weather,Time-LLM,MSE,0.255,,4,22,8
+accuracy,Avg,Time-LLM,MSE,0.448,,4,26,8
+accuracy,Electricity,Time-LLM,MSE,0.520,,4,27,8
+accuracy,Avg,Time-LLM,MSE,0.510,,4,31,8
+accuracy,Traffic,Time-LLM,MSE,1.068,,4,32,8
+accuracy,Avg,Time-LLM,MSE,0.765,,4,36,8
+accuracy,ILI,Time-LLM,MSE,4.909,,4,37,8
+accuracy,Avg,Time-LLM,MSE,1.346,,4,41,8
+accuracy,Yearly,Time-LLM,SMAPE,13.419,,5,1,2
+accuracy,MASE,Time-LLM,SMAPE,3.565,,5,2,2
+accuracy,OWA,Time-LLM,SMAPE,0.911,,5,3,2
+accuracy,Quarterly,Time-LLM,SMAPE,10.110,,5,4,2
+accuracy,MASE,Time-LLM,SMAPE,1.253,,5,5,2
+accuracy,OWA,Time-LLM,SMAPE,0.938,,5,6,2
+accuracy,Monthly,Time-LLM,SMAPE,12.980,,5,7,2
+accuracy,MASE,Time-LLM,SMAPE,1.003,,5,8,2
+accuracy,OWA,Time-LLM,SMAPE,0.931,,5,9,2
+accuracy,Others,Time-LLM,SMAPE,4.795,,5,10,2
+accuracy,MASE,Time-LLM,SMAPE,4.116,,5,11,2
+accuracy,OWA,Time-LLM,SMAPE,1.259,,5,12,2
+accuracy,Average,Time-LLM,SMAPE,11.983,,5,13,2
+accuracy,MASE,Time-LLM,SMAPE,1.808,,5,14,2
+accuracy,OWA,Time-LLM,SMAPE,0.94,,5,15,2
+accuracy,ETTh1,Time-LLM,MSE,0.556,,7,2,1
+accuracy,ETTh1,Time-LLM,MSE,0.522,,7,2,2
+accuracy,ETTh2,Time-LLM,MSE,0.370,,7,3,1
+accuracy,ETTh2,Time-LLM,MSE,0.394,,7,3,2
+accuracy,ETTm1,Time-LLM,MSE,0.404,,7,4,1
+accuracy,ETTm1,Time-LLM,MSE,0.427,,7,4,2
+accuracy,ETTm2,Time-LLM,MSE,0.277,,7,5,1
+accuracy,ETTm2,Time-LLM,MSE,0.323,,7,5,2
+accuracy,Weather,Time-LLM,MSE,0.234,,7,6,1
+accuracy,Weather,Time-LLM,MSE,0.273,,7,6,2
+accuracy,ECL,Time-LLM,MSE,0.175,,7,7,1
+accuracy,ECL,Time-LLM,MSE,0.270,,7,7,2
+accuracy,Traffic,Time-LLM,MSE,0.429,,7,8,1
+accuracy,Traffic,Time-LLM,MSE,0.306,,7,8,2

result/per_paper/2310.01728/components_architecture.csv

@@ -0,0 +1,5 @@
+component,what_it_is,provenance,citation,evidence
+Text Prototype Representation,A method to reprogram input time series into text-like token sequences that align with the modality of large language models (LLMs).,proposed_here,,We begin by reprogramming the input time series with text prototypes before feeding it into the frozen LLM to align the two modalities.
+Prompt-as-Prefix (PaP),"A technique to enrich input context with declarative prompts (e.g., domain knowledge, task instructions) to guide the LLM's reasoning about time series data.",proposed_here,,"We propose Prompt-as-Prefix (PaP), which enriches the input context and directs the transformation of reprogrammed input patches."
+Forecast Projection Module,A component that projects the output of the LLM (after processing reprogrammed time series) into time series forecasts.,proposed_here,,The transformed time series patches from the LLM are finally projected to obtain the forecasts.
+Frozen LLM Backbone,"The pre-trained large language model (e.g., GPT variants) used as a fixed backbone for processing reprogrammed time series inputs.",reused_cited,"Chen, 2022; Zhou et al., 2023a","Our approach aligns with model reprogramming (Chen, 2022) and leverages pre-trained language models without altering their self-attention and feedforward layers (Zhou et al., 2023a)."

result/per_paper/2310.01728/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,NVIDIA A100-80G,GPU,stated,Our experiments are conducted on NVIDIA A100-80G GPUs.
+num_devices,,,not_reported,"Number of devices (e.g., GPUs) is not explicitly stated."
+training_cost,,,not_reported,"Training cost (e.g., monetary or computational cost) is not reported."
+training_batch_size,,,not_reported,"Batch sizes vary per task (e.g., 16 for ETTh1, 8 for Weather), but no single value is stated for Time-LLM."
+training_steps_or_epochs,,,not_reported,"Epochs vary per task (e.g., 50 for ETTh1, 100 for ETTm1), but no single value is stated for Time-LLM."
+precision,,,not_reported,"Precision (e.g., 16-bit/32-bit) is not explicitly stated."
+inference_latency,,,not_reported,Inference latency is not mentioned.
+inference_throughput,,,not_reported,Inference throughput is not mentioned.
+peak_memory,,,not_reported,"Peak memory is not explicitly stated, though GPU memory overhead is mentioned as a metric."
+flops_or_macs,,,not_reported,FLOPs or MACs are not mentioned.
+num_inference_samples,,,not_reported,Number of inference samples is not mentioned.
+params,,,not_reported,"Trainable parameters are mentioned as a metric, but no absolute value is provided for Time-LLM."
+context_lengths_evaluated,,,not_reported,"Input length $T$ is mentioned (e.g., 512 for LTF tasks), but not explicitly stated as 'context lengths evaluated.'"
+horizon_lengths_evaluated,"96, 336",steps,stated,Forecasting tasks on ETTh1 involve horizons of 96 and 336 steps.
+inference_batch_size,,,not_reported,Inference batch size is not mentioned.

result/per_paper/2310.03589/accuracy_efficiency.csv

@@ -0,0 +1 @@
+metric_kind,entity,model_variant,metric,value,uncertainty

result/per_paper/2310.03589/accuracy_efficiency_traced.csv

@@ -0,0 +1 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col

result/per_paper/2310.03589/components_architecture.csv

@@ -0,0 +1,10 @@
+component,what_it_is,provenance,citation,evidence
+Transformer-based Architecture,"The foundational structure of TimeGPT, utilizing self-attention mechanisms to process temporal dependencies.",reused_cited,"Vaswani et al., 2017","TimeGPT is a Transformer-based time series model with self-attention mechanisms based on [Vaswani et al., 2017]."
+Positional Encoding,A method to inject information about the position of elements in the sequence into the model's input.,reused_cited,"Vaswani et al., 2017","The architecture consists of an encoder-decoder structure with multiple layers, each with residual connections and layer normalization. Finally, a linear layer maps the decoder's output to the forecasting window dimension."
+Encoder-Decoder Structure,A two-part architecture where the encoder processes input sequences and the decoder generates output sequences.,reused_cited,"Vaswani et al., 2017","The architecture consists of an encoder-decoder structure with multiple layers, each with residual connections and layer normalization."
+Residual Connections,A technique to facilitate the training of deeper networks by allowing gradients to flow through the network more effectively.,reused_cited,"Vaswani et al., 2017","The architecture consists of an encoder-decoder structure with multiple layers, each with residual connections and layer normalization."
+Layer Normalization,A normalization technique applied to the outputs of each layer to stabilize the training process.,reused_cited,"Vaswani et al., 2017","The architecture consists of an encoder-decoder structure with multiple layers, each with residual connections and layer normalization."
+Multi-head Attention,A mechanism that allows the model to focus on different parts of the input sequence simultaneously.,reused_cited,"Vaswani et al., 2017",The general intuition is that attention-based mechanisms are able to capture the diversity of past events and correctly extrapolate potential future distributions.
+CNN (Convolutional Neural Network),A component integrated into the model to process local patterns in the time series data.,proposed_here,,"The architecture includes a CNN as part of the processing pipeline, as indicated in the mermaid diagram."
+Linear Layer,A final layer that maps the decoder's output to the forecasting window dimension.,proposed_here,,"Finally, a linear layer maps the decoder's output to the forecasting window dimension."
+Output Embedding,A representation of the output sequence used during both training and inference phases.,proposed_here,,The mermaid diagram shows 'Output Embedding' as a component in both training and inference subgraphs.

result/per_paper/2310.03589/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,NVIDIA A10G GPUs,,stated,TimeGPT underwent a multi-day training period on a cluster of NVIDIA A10G GPUs.
+num_devices,,,not_reported,The paper does not specify the number of devices in the cluster.
+training_cost,,,not_reported,"The paper does not report training cost (e.g., monetary or energy cost)."
+training_batch_size,,,not_reported,The paper mentions hyperparameter exploration but does not specify the training batch size.
+training_steps_or_epochs,,,not_reported,The paper does not report training steps or epochs.
+precision,,,not_reported,"The paper does not specify training or inference precision (e.g., FP16, FP32)."
+inference_latency,0.6,milliseconds per series,stated,"For zero-shot inference, our internal tests recorded an average GPU inference speed of 0.6 milliseconds per series for TimeGPT."
+inference_throughput,,,not_reported,"The paper does not report inference throughput (e.g., samples per second)."
+peak_memory,,,not_reported,The paper does not report peak memory usage during training or inference.
+flops_or_macs,,,not_reported,The paper does not report FLOPs or MACs for the model.
+num_inference_samples,,,not_reported,The paper does not specify the number of inference samples tested.
+params,,,not_reported,The paper does not report the number of parameters in TimeGPT-1.
+context_lengths_evaluated,,,not_reported,The paper does not specify context lengths evaluated during training or testing.
+horizon_lengths_evaluated,,,not_reported,The paper does not specify forecast horizon lengths evaluated.
+inference_batch_size,,,not_reported,The paper does not report inference batch size.

result/per_paper/2310.04948/accuracy_efficiency.csv

@@ -0,0 +1,63 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,GPT2,TEMPO,MSE/MAE,0.193,
+accuracy,T5,TEMPO,MSE/MAE,0.185,
+accuracy,PatchTST,TEMPO,MSE/MAE,0.489,
+accuracy,Timesnet,TEMPO,MSE/MAE,0.293,
+accuracy,FEDformer,TEMPO,MSE/MAE,0.300,
+accuracy,ETSformer,TEMPO,MSE/MAE,0.707,
+accuracy,Informer,TEMPO,MSE/MAE,0.512,
+accuracy,DLinear,TEMPO,MSE/MAE,0.195,
+accuracy,GPT2,TEMPO,MSE/MAE,0.207,
+accuracy,T5,TEMPO,MSE/MAE,0.205,
+accuracy,PatchTST,TEMPO,MSE/MAE,0.465,
+accuracy,Timesnet,TEMPO,MSE/MAE,0.283,
+accuracy,FEDformer,TEMPO,MSE/MAE,0.390,
+accuracy,ETSformer,TEMPO,MSE/MAE,0.721,
+accuracy,Informer,TEMPO,MSE/MAE,0.625,
+accuracy,DLinear,TEMPO,MSE/MAE,0.204,
+accuracy,GPT2,TEMPO,MSE/MAE,0.231,
+accuracy,T5,TEMPO,MSE/MAE,0.229,
+accuracy,PatchTST,TEMPO,MSE/MAE,0.531,
+accuracy,Timesnet,TEMPO,MSE/MAE,0.733,
+accuracy,FEDformer,TEMPO,MSE/MAE,0.317,
+accuracy,ETSformer,TEMPO,MSE/MAE,0.862,
+accuracy,Informer,TEMPO,MSE/MAE,1.222,
+accuracy,DLinear,TEMPO,MSE/MAE,0.231,
+accuracy,GPT2,TEMPO,MSE/MAE,0.262,
+accuracy,T5,TEMPO,MSE/MAE,0.266,
+accuracy,PatchTST,TEMPO,MSE/MAE,0.475,
+accuracy,Timesnet,TEMPO,MSE/MAE,1.166,
+accuracy,FEDformer,TEMPO,MSE/MAE,0.423,
+accuracy,ETSformer,TEMPO,MSE/MAE,0.666,
+accuracy,Informer,TEMPO,MSE/MAE,0.881,
+accuracy,DLinear,TEMPO,MSE/MAE,0.259,
+accuracy,GPT2,TEMPO,MSE/MAE,0.223,
+accuracy,T5,TEMPO,MSE/MAE,0.221,
+accuracy,PatchTST,TEMPO,MSE/MAE,0.49,
+accuracy,Timesnet,TEMPO,MSE/MAE,0.619,
+accuracy,FEDformer,TEMPO,MSE/MAE,0.358,
+accuracy,ETSformer,TEMPO,MSE/MAE,0.750,
+accuracy,Informer,TEMPO,MSE/MAE,0.810,
+accuracy,DLinear,TEMPO,MSE/MAE,0.222,
+accuracy,ETTh1,TEMPO,MSE/MAE,0.400,
+accuracy,Avg.,TEMPO,MSE/MAE,0.454,
+accuracy,ETTh2,TEMPO,MSE/MAE,0.301,
+accuracy,Avg.,TEMPO,MSE/MAE,0.383,
+accuracy,ECL,TEMPO,MSE/MAE,0.19,
+accuracy,Avg,TEMPO,MSE/MAE,0.277,
+accuracy,Traffic,TEMPO,MSE/MAE,0.56,
+accuracy,Avg,TEMPO,MSE/MAE,0.621,
+accuracy,Weather,TEMPO,MSE/MAE,0.217,
+accuracy,Avg,TEMPO,MSE/MAE,0.328,
+accuracy,CC,TEMPO,SMAPE,32.27,
+accuracy,CD,TEMPO,SMAPE,25.9,
+accuracy,Ind,TEMPO,SMAPE,26.7,
+accuracy,RE,TEMPO,SMAPE,29.46,
+accuracy,ECL,TEMPO,MSE/MAE,0.178,
+accuracy,Avg,TEMPO,MSE/MAE,0.228,
+accuracy,Ettm1,TEMPO,MSE/MAE,0.438,
+accuracy,Avg,TEMPO,MSE/MAE,0.575,
+accuracy,Avg.,TEMPO,MSE/MAE,0.216,
+accuracy,Avg.,TEMPO,MSE/MAE,0.503,
+accuracy,Avg.,TEMPO,MSE/MAE,0.287,
+accuracy,Avg.,TEMPO,MSE/MAE,0.280,

result/per_paper/2310.04948/accuracy_efficiency_traced.csv

@@ -0,0 +1,63 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,GPT2,TEMPO,MSE/MAE,0.193,,0,3,1
+accuracy,T5,TEMPO,MSE/MAE,0.185,,0,4,1
+accuracy,PatchTST,TEMPO,MSE/MAE,0.489,,0,5,1
+accuracy,Timesnet,TEMPO,MSE/MAE,0.293,,0,6,1
+accuracy,FEDformer,TEMPO,MSE/MAE,0.300,,0,7,1
+accuracy,ETSformer,TEMPO,MSE/MAE,0.707,,0,8,1
+accuracy,Informer,TEMPO,MSE/MAE,0.512,,0,9,1
+accuracy,DLinear,TEMPO,MSE/MAE,0.195,,0,10,1
+accuracy,GPT2,TEMPO,MSE/MAE,0.207,,0,12,1
+accuracy,T5,TEMPO,MSE/MAE,0.205,,0,13,1
+accuracy,PatchTST,TEMPO,MSE/MAE,0.465,,0,14,1
+accuracy,Timesnet,TEMPO,MSE/MAE,0.283,,0,15,1
+accuracy,FEDformer,TEMPO,MSE/MAE,0.390,,0,16,1
+accuracy,ETSformer,TEMPO,MSE/MAE,0.721,,0,17,1
+accuracy,Informer,TEMPO,MSE/MAE,0.625,,0,18,1
+accuracy,DLinear,TEMPO,MSE/MAE,0.204,,0,19,1
+accuracy,GPT2,TEMPO,MSE/MAE,0.231,,0,21,1
+accuracy,T5,TEMPO,MSE/MAE,0.229,,0,22,1
+accuracy,PatchTST,TEMPO,MSE/MAE,0.531,,0,23,1
+accuracy,Timesnet,TEMPO,MSE/MAE,0.733,,0,24,1
+accuracy,FEDformer,TEMPO,MSE/MAE,0.317,,0,25,1
+accuracy,ETSformer,TEMPO,MSE/MAE,0.862,,0,26,1
+accuracy,Informer,TEMPO,MSE/MAE,1.222,,0,27,1
+accuracy,DLinear,TEMPO,MSE/MAE,0.231,,0,28,1
+accuracy,GPT2,TEMPO,MSE/MAE,0.262,,0,30,1
+accuracy,T5,TEMPO,MSE/MAE,0.266,,0,31,1
+accuracy,PatchTST,TEMPO,MSE/MAE,0.475,,0,32,1
+accuracy,Timesnet,TEMPO,MSE/MAE,1.166,,0,33,1
+accuracy,FEDformer,TEMPO,MSE/MAE,0.423,,0,34,1
+accuracy,ETSformer,TEMPO,MSE/MAE,0.666,,0,35,1
+accuracy,Informer,TEMPO,MSE/MAE,0.881,,0,36,1
+accuracy,DLinear,TEMPO,MSE/MAE,0.259,,0,37,1
+accuracy,GPT2,TEMPO,MSE/MAE,0.223,,0,39,1
+accuracy,T5,TEMPO,MSE/MAE,0.221,,0,40,1
+accuracy,PatchTST,TEMPO,MSE/MAE,0.49,,0,41,1
+accuracy,Timesnet,TEMPO,MSE/MAE,0.619,,0,42,1
+accuracy,FEDformer,TEMPO,MSE/MAE,0.358,,0,43,1
+accuracy,ETSformer,TEMPO,MSE/MAE,0.750,,0,44,1
+accuracy,Informer,TEMPO,MSE/MAE,0.810,,0,45,1
+accuracy,DLinear,TEMPO,MSE/MAE,0.222,,0,46,1
+accuracy,ETTh1,TEMPO,MSE/MAE,0.400,,1,2,2
+accuracy,Avg.,TEMPO,MSE/MAE,0.454,,1,6,2
+accuracy,ETTh2,TEMPO,MSE/MAE,0.301,,1,7,2
+accuracy,Avg.,TEMPO,MSE/MAE,0.383,,1,11,2
+accuracy,ECL,TEMPO,MSE/MAE,0.19,,3,2,2
+accuracy,Avg,TEMPO,MSE/MAE,0.277,,3,6,2
+accuracy,Traffic,TEMPO,MSE/MAE,0.56,,3,7,2
+accuracy,Avg,TEMPO,MSE/MAE,0.621,,3,11,2
+accuracy,Weather,TEMPO,MSE/MAE,0.217,,3,12,2
+accuracy,Avg,TEMPO,MSE/MAE,0.328,,3,16,2
+accuracy,CC,TEMPO,SMAPE,32.27,,4,2,1
+accuracy,CD,TEMPO,SMAPE,25.9,,4,3,1
+accuracy,Ind,TEMPO,SMAPE,26.7,,4,4,1
+accuracy,RE,TEMPO,SMAPE,29.46,,4,5,1
+accuracy,ECL,TEMPO,MSE/MAE,0.178,,5,2,2
+accuracy,Avg,TEMPO,MSE/MAE,0.228,,5,6,2
+accuracy,Ettm1,TEMPO,MSE/MAE,0.438,,5,7,2
+accuracy,Avg,TEMPO,MSE/MAE,0.575,,5,11,2
+accuracy,Avg.,TEMPO,MSE/MAE,0.216,,6,7,1
+accuracy,Avg.,TEMPO,MSE/MAE,0.503,,6,7,3
+accuracy,Avg.,TEMPO,MSE/MAE,0.287,,6,7,5
+accuracy,Avg.,TEMPO,MSE/MAE,0.280,,6,7,7

result/per_paper/2310.04948/components_architecture.csv

@@ -0,0 +1,6 @@
+component,what_it_is,provenance,citation,evidence
+Time Series Decomposition Module,"Decomposes input time series into three additive components: trend, seasonality, and residuals using locally weighted scatterplot smoothing (LOESS).",proposed_here,,"TEMPO first decomposes time series input into three additive components, i.e., trend, seasonality, and residuals via locally weighted scatterplot smoothing (Cleveland et al., 1990)."
+Hidden Space Mapping,"Maps each decomposed time series component (trend, seasonality, residuals) to a corresponding hidden space to construct the input embedding for the generative pre-trained transformer (GPT).",proposed_here,,Each of these temporal inputs is subsequently mapped to its corresponding hidden space to construct the time series input embedding of the generative pre-trained transformer (GPT).
+Soft Prompt Mechanism,"A prompt-based tuning strategy that uses learnable continuous vector representations to encode temporal knowledge of trend and seasonality, guiding the GPT for forecasting tasks.",proposed_here,,TEMPO utilizes a soft prompt to efficiently tune the GPT [...] by guiding the reuse of a collection of learnable continuous vector representations that encode temporal knowledge of trend and seasonality.
+Generative Pre-trained Transformer (GPT) Core,"The foundational transformer architecture adapted for time series forecasting, leveraging pre-trained weights and attention mechanisms.",reused_cited,"Radford, A., et al. (2019)",TEMPO [...] expands the capability for dynamically modeling real-world temporal phenomena from data within diverse domains.
+Interpretable Framework for Component Interaction,"A framework that leverages the three additive components (trend, seasonality, residuals) to provide interpretable insights into interactions among input components.",proposed_here,"Hastie, T. (2017)","We leverage the three key additive components of time series data [...] to provide an interpretable framework for comprehending the interactions among input components (Hastie, 2017)."

result/per_paper/2310.04948/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,
+num_devices,,,not_reported,
+training_cost,,,not_reported,
+training_batch_size,,,not_reported,
+training_steps_or_epochs,,,not_reported,
+precision,,,not_reported,
+inference_latency,,,not_reported,
+inference_throughput,,,not_reported,
+peak_memory,,,not_reported,
+flops_or_macs,,,not_reported,
+num_inference_samples,,,not_reported,
+params,,,not_reported,
+context_lengths_evaluated,,,not_reported,
+horizon_lengths_evaluated,,,not_reported,
+inference_batch_size,,,not_reported,

result/per_paper/2310.07446/accuracy_efficiency.csv

@@ -0,0 +1 @@
+metric_kind,entity,model_variant,metric,value,uncertainty

result/per_paper/2310.07446/accuracy_efficiency_traced.csv

@@ -0,0 +1 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col

result/per_paper/2310.07446/components_architecture.csv

@@ -0,0 +1,6 @@
+component,what_it_is,provenance,citation,evidence
+Distributional Forecasting Approach,"Method for generating forecasts, ranging from point forecasts to predefined distribution heads based on data assumptions.",reused_cited,"56, 70","The first aspect involves the approach to distributional forecasting, ranging from models focused on point forecasts [49, 40] to those using pre-defined distribution heads based on specific data assumptions [56, 70]."
+Decoding Scheme,"Strategy for generating multi-step forecasts, either autoregressive (AR) or non-autoregressive (NAR).",reused_cited,"58, 57, 32","The second aspect is the decoding scheme used to generate multi-step forecasts, which can be either autoregressive (AR) or non-autoregressive (NAR)."
+Normalization Strategy,"Choice of normalization technique, such as reversible instance normalization (RevIN) for long-term point forecasting or mean scaling for short-term probabilistic forecasting.",reused_cited,"32, 58, 57","The third aspect pertains to the normalization choice, where the long-term point forecasting models typically employ reversible instance normalization (RevIN) [32] while short-term probabilistic ones often use mean scaling strategies [58, 57]."
+Trend and Seasonality Strength Measurement,Metric to quantify the strength of trends and seasonality in time-series data for benchmarking.,proposed_here,,"We measure three essential data characteristics in ProbTS: the strength of trends and seasonality, and the complexity of the data distribution."
+Data Distribution Complexity Measurement,Metric to assess the complexity of data distributions in time-series for benchmarking.,proposed_here,,"We measure three essential data characteristics in ProbTS: the strength of trends and seasonality, and the complexity of the data distribution."

result/per_paper/2310.07446/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,NVIDIA Tesla V100,GPU,stated,run on single NVIDIA Tesla V100 GPUs with CUDA 11.3
+num_devices,1,device,stated,run on single NVIDIA Tesla V100 GPUs
+training_cost,,,not_reported,
+training_batch_size,,,not_reported,
+training_steps_or_epochs,50,epochs,stated,limited training to 50 epochs
+precision,,,not_reported,
+inference_latency,,,not_reported,
+inference_throughput,,,not_reported,
+peak_memory,,,not_reported,
+flops_or_macs,,,not_reported,
+num_inference_samples,100,samples,stated,100 samples are employed to estimate the empirical CDF
+params,,,not_reported,
+context_lengths_evaluated,,,not_reported,
+horizon_lengths_evaluated,"[24, 48, 96, 192, 336, 720]",time steps,stated,"selected prediction horizons of {24, 48, 96, 192, 336, 720}"
+inference_batch_size,,,not_reported,

result/per_paper/2310.08278/accuracy_efficiency.csv

@@ -0,0 +1,13 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,PATCHTST,Lag-Llama,CRPS,0.169,
+accuracy,TFT,Lag-Llama,CRPS,0.154,
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.136,
+accuracy,PATCHTST,Lag-Llama,CRPS,0.171,
+accuracy,TFT,Lag-Llama,CRPS,0.156,
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.135,
+accuracy,PATCHTST,Lag-Llama,CRPS,0.174,
+accuracy,TFT,Lag-Llama,CRPS,0.152,
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.133,
+accuracy,PATCHTST,Lag-Llama,CRPS,0.174,
+accuracy,TFT,Lag-Llama,CRPS,0.148,
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.132,

result/per_paper/2310.08278/accuracy_efficiency_traced.csv

@@ -0,0 +1,13 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,PATCHTST,Lag-Llama,CRPS,0.169,,0,3,1
+accuracy,TFT,Lag-Llama,CRPS,0.154,,0,4,1
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.136,,0,5,1
+accuracy,PATCHTST,Lag-Llama,CRPS,0.171,,0,7,1
+accuracy,TFT,Lag-Llama,CRPS,0.156,,0,8,1
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.135,,0,9,1
+accuracy,PATCHTST,Lag-Llama,CRPS,0.174,,0,11,1
+accuracy,TFT,Lag-Llama,CRPS,0.152,,0,12,1
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.133,,0,13,1
+accuracy,PATCHTST,Lag-Llama,CRPS,0.174,,0,15,1
+accuracy,TFT,Lag-Llama,CRPS,0.148,,0,16,1
+accuracy,LAG-LLAMA,Lag-Llama,CRPS,0.132,,0,17,1

result/per_paper/2310.08278/components_architecture.csv

@@ -0,0 +1,6 @@
+component,what_it_is,provenance,citation,evidence
+Decoder-only Transformer Architecture,"A transformer-based model structure that uses only the decoder component, designed for sequence generation tasks.",proposed_here,"Rasul et al., 2023",Lag-Llama is based on a decoder-only transformer architecture that uses lags as covariates.
+Lags as Covariates,A method where historical time series values (lags) are used as input features to predict future values.,proposed_here,"Rasul et al., 2023",uses lags as covariates.
+Pretraining on Diverse Time Series Data,"A pretraining strategy involving a large, diverse corpus of time series data from multiple domains to enable zero-shot generalization.",proposed_here,"Rasul et al., 2023",pretrained on a large corpus of diverse time series data from several domains.
+Probabilistic Forecasting Mechanism,"A framework for generating probabilistic forecasts, providing uncertainty estimates alongside point predictions.",proposed_here,"Rasul et al., 2023",univariate probabilistic time series forecasting.
+Univariate Time Series Processing,A focus on modeling single-variable (univariate) time series data rather than multivariate sequences.,proposed_here,"Rasul et al., 2023",univariate probabilistic time series forecasting.

result/per_paper/2310.08278/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,Nvidia Tesla-P100 GPU,,stated,"We use a single Nvidia Tesla-P100 GPU with 12 GB of memory, 4 CPU cores, and 24 GB of RAM."
+num_devices,1,,stated,We use a single Nvidia Tesla-P100 GPU...
+training_cost,,,not_reported,
+training_batch_size,256,,stated,"During pretraining, we use the batch size of 256..."
+training_steps_or_epochs,50,epochs,stated,We use an early stopping criterion of 50 epochs based on the average validation loss...
+precision,,,not_reported,
+inference_latency,,,not_reported,
+inference_throughput,,,not_reported,
+peak_memory,,,not_reported,
+flops_or_macs,,,not_reported,
+num_inference_samples,,,not_reported,
+params,,,not_reported,
+context_lengths_evaluated,,,not_reported,
+horizon_lengths_evaluated,,,not_reported,
+inference_batch_size,,,not_reported,

result/per_paper/2310.10688/accuracy_efficiency.csv

@@ -0,0 +1,48 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,australian electricity demand,TimesFM(ZS),MAE,448.81,
+accuracy,bitcoin,TimesFM(ZS),MAE,1.3,
+accuracy,pedestrian counts,TimesFM(ZS),MAE,40.71,
+accuracy,weather,TimesFM(ZS),MAE,2.07,
+accuracy,nn5 daily,TimesFM(ZS),MAE,3.54,
+accuracy,nn5 weekly,TimesFM(ZS),MAE,14.67,
+accuracy,tourism yearly,TimesFM(ZS),MAE,109977.29,
+accuracy,tourism quarterly,TimesFM(ZS),MAE,12102.04,
+accuracy,tourism monthly,TimesFM(ZS),MAE,3183.77,
+accuracy,cif 2016,TimesFM(ZS),MAE,773980.44,
+accuracy,covid deaths,TimesFM(ZS),MAE,209.80,
+accuracy,fred md,TimesFM(ZS),MAE,947.12,
+accuracy,traffic hourly,TimesFM(ZS),MAE,0.01,
+accuracy,traffic weekly,TimesFM(ZS),MAE,1.12,
+accuracy,saugeenday,TimesFM(ZS),MAE,24.63,
+accuracy,us births,TimesFM(ZS),MAE,437.27,
+accuracy,hospital,TimesFM(ZS),MAE,19.41,
+accuracy,solar weekly,TimesFM(ZS),MAE,1258.27,
+accuracy,Scaled MAE (Arithmetic Mean),TimesFM(ZS),MAE,0.8005,
+accuracy,Scaled MAE (Geometric Mean),TimesFM(ZS),MAE,0.6846,
+accuracy,ETTh1,TimesFM,MAE,96,
+accuracy,Avg,TimesFM,MAE,0.426,
+accuracy,ETTh2,TimesFM,MAE,96,
+accuracy,Avg,TimesFM,MAE,0.410,
+accuracy,ETTm1,TimesFM,MAE,96,
+accuracy,Avg,TimesFM,MAE,0.388,
+accuracy,ETTm2,TimesFM,MAE,96,
+accuracy,Avg,TimesFM,MAE,0.334,
+accuracy,AirPassengersDataset,TimesFM(ZS),MAE,62.51,
+accuracy,AusBeerDataset,TimesFM(ZS),MAE,11.94,
+accuracy,GasRateCO2Dataset,TimesFM(ZS),MAE,2.50,
+accuracy,MonthlyMilkDataset,TimesFM(ZS),MAE,28.09,
+accuracy,SunspotsDataset,TimesFM(ZS),MAE,41.40,
+accuracy,WineDataset,TimesFM(ZS),MAE,2871.33,
+accuracy,WoolyDataset,TimesFM(ZS),MAE,728.92,
+accuracy,HeartRateDataset,TimesFM(ZS),MAE,5.85,
+accuracy,Scaled MAE (Arithmetic Mean),TimesFM(ZS),MAE,0.6829,
+accuracy,Scaled MAE (Geometric Mean),TimesFM(ZS),MAE,0.5767,
+accuracy,ETTh1 (horizon=96),TimesFM(ZS),MAE,0.45,
+accuracy,ETTh1 (horizon=192),TimesFM(ZS),MAE,0.53,
+accuracy,ETTh2 (horizon=96),TimesFM(ZS),MAE,0.35,
+accuracy,ETTh2 (horizon=192),TimesFM(ZS),MAE,0.62,
+accuracy,ETTm1 (horizon=96),TimesFM(ZS),MAE,0.19,
+accuracy,ETTm1 (horizon=192),TimesFM(ZS),MAE,0.26,
+accuracy,ETTm2 (horizon=96),TimesFM(ZS),MAE,0.24,
+accuracy,ETTm2 (horizon=192),TimesFM(ZS),MAE,0.27,
+accuracy,Avg,TimesFM(ZS),MAE,0.36,

result/per_paper/2310.10688/accuracy_efficiency_traced.csv

@@ -0,0 +1,48 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,australian electricity demand,TimesFM(ZS),MAE,448.81,,0,1,15
+accuracy,bitcoin,TimesFM(ZS),MAE,1.3,,0,2,15
+accuracy,pedestrian counts,TimesFM(ZS),MAE,40.71,,0,3,15
+accuracy,weather,TimesFM(ZS),MAE,2.07,,0,4,15
+accuracy,nn5 daily,TimesFM(ZS),MAE,3.54,,0,5,15
+accuracy,nn5 weekly,TimesFM(ZS),MAE,14.67,,0,6,15
+accuracy,tourism yearly,TimesFM(ZS),MAE,109977.29,,0,7,15
+accuracy,tourism quarterly,TimesFM(ZS),MAE,12102.04,,0,8,15
+accuracy,tourism monthly,TimesFM(ZS),MAE,3183.77,,0,9,15
+accuracy,cif 2016,TimesFM(ZS),MAE,773980.44,,0,10,15
+accuracy,covid deaths,TimesFM(ZS),MAE,209.80,,0,11,15
+accuracy,fred md,TimesFM(ZS),MAE,947.12,,0,12,15
+accuracy,traffic hourly,TimesFM(ZS),MAE,0.01,,0,13,15
+accuracy,traffic weekly,TimesFM(ZS),MAE,1.12,,0,14,15
+accuracy,saugeenday,TimesFM(ZS),MAE,24.63,,0,15,15
+accuracy,us births,TimesFM(ZS),MAE,437.27,,0,16,15
+accuracy,hospital,TimesFM(ZS),MAE,19.41,,0,17,15
+accuracy,solar weekly,TimesFM(ZS),MAE,1258.27,,0,18,15
+accuracy,Scaled MAE (Arithmetic Mean),TimesFM(ZS),MAE,0.8005,,0,19,15
+accuracy,Scaled MAE (Geometric Mean),TimesFM(ZS),MAE,0.6846,,0,20,15
+accuracy,ETTh1,TimesFM,MAE,96,,1,1,1
+accuracy,Avg,TimesFM,MAE,0.426,,1,5,1
+accuracy,ETTh2,TimesFM,MAE,96,,1,6,1
+accuracy,Avg,TimesFM,MAE,0.410,,1,10,1
+accuracy,ETTm1,TimesFM,MAE,96,,1,11,1
+accuracy,Avg,TimesFM,MAE,0.388,,1,15,1
+accuracy,ETTm2,TimesFM,MAE,96,,1,16,1
+accuracy,Avg,TimesFM,MAE,0.334,,1,20,1
+accuracy,AirPassengersDataset,TimesFM(ZS),MAE,62.51,,2,1,7
+accuracy,AusBeerDataset,TimesFM(ZS),MAE,11.94,,2,2,7
+accuracy,GasRateCO2Dataset,TimesFM(ZS),MAE,2.50,,2,3,7
+accuracy,MonthlyMilkDataset,TimesFM(ZS),MAE,28.09,,2,4,7
+accuracy,SunspotsDataset,TimesFM(ZS),MAE,41.40,,2,5,7
+accuracy,WineDataset,TimesFM(ZS),MAE,2871.33,,2,6,7
+accuracy,WoolyDataset,TimesFM(ZS),MAE,728.92,,2,7,7
+accuracy,HeartRateDataset,TimesFM(ZS),MAE,5.85,,2,8,7
+accuracy,Scaled MAE (Arithmetic Mean),TimesFM(ZS),MAE,0.6829,,2,9,7
+accuracy,Scaled MAE (Geometric Mean),TimesFM(ZS),MAE,0.5767,,2,10,7
+accuracy,ETTh1 (horizon=96),TimesFM(ZS),MAE,0.45,,3,1,7
+accuracy,ETTh1 (horizon=192),TimesFM(ZS),MAE,0.53,,3,2,7
+accuracy,ETTh2 (horizon=96),TimesFM(ZS),MAE,0.35,,3,3,7
+accuracy,ETTh2 (horizon=192),TimesFM(ZS),MAE,0.62,,3,4,7
+accuracy,ETTm1 (horizon=96),TimesFM(ZS),MAE,0.19,,3,5,7
+accuracy,ETTm1 (horizon=192),TimesFM(ZS),MAE,0.26,,3,6,7
+accuracy,ETTm2 (horizon=96),TimesFM(ZS),MAE,0.24,,3,7,7
+accuracy,ETTm2 (horizon=192),TimesFM(ZS),MAE,0.27,,3,8,7
+accuracy,Avg,TimesFM(ZS),MAE,0.36,,3,9,7

result/per_paper/2310.10688/components_architecture.csv

@@ -0,0 +1,6 @@
+component,what_it_is,provenance,citation,evidence
+Decoder-only Attention Architecture,"A transformer-based model trained exclusively in decoder mode, where the model predicts the next patch based on all past patches in parallel across the context window.",proposed_here,,"Our model is trained in decoder-only mode [LSP+18]. In other words, given a sequence of input patches, the model is optimized to predict the next patch as a function of all past patches."
+Input Patching,"A method to divide time-series data into fixed-length segments (patches) during training, analogous to tokenization in language models.",reused_cited,NNSK22,Inspired by the success of patch based modeling in the recent long horizon forecasting work [NNSK22] we also choose to break down the time-series into patches during training.
+Longer Output Patches,"An architectural choice where output patches for prediction are longer than input patches, enabling simultaneous forecasting of multiple future time-steps.",proposed_here,,"We propose a middle ground by allowing our output patches for prediction to be longer than the input patches. As an example, suppose the input patch length is 32 and output patch length is 128."
+Large-Scale Time-Series Corpus,"A pretraining dataset combining real-world (e.g., web search queries, Wikipedia page visits) and synthetic time-series data to ensure diversity and volume for training.",proposed_here,,"Our model is based on pretraining a decoder style attention model with input patching, using a large time-series corpus comprising both real-world and synthetic datasets."
+Zero-shot Forecasting Capability,"The ability to generate accurate forecasts on previously unseen datasets without task-specific fine-tuning, leveraging the pretraining on diverse data.",proposed_here,,"Our model can yield accurate zero-shot forecasts across different domains, forecasting horizons and temporal granularities."

result/per_paper/2310.10688/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,
+num_devices,,,not_reported,
+training_cost,,,not_reported,
+training_batch_size,,,not_reported,
+training_steps_or_epochs,,,not_reported,
+precision,,,not_reported,
+inference_latency,,,not_reported,
+inference_throughput,,,not_reported,
+peak_memory,,,not_reported,
+flops_or_macs,,,not_reported,
+num_inference_samples,,,not_reported,
+params,,,not_reported,
+context_lengths_evaluated,,,not_reported,
+horizon_lengths_evaluated,,,not_reported,
+inference_batch_size,,,not_reported,

result/per_paper/2311.01933/accuracy_efficiency.csv

@@ -0,0 +1,5 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,Data Budget = 50,ForecastPFN,MAE,50,
+accuracy,Data Budget = 500,ForecastPFN,MAE,500,
+accuracy,"ForecastPFN, low noise","ForecastPFN, lowest noise",MSE,0.359,
+accuracy,ForecastPFN,"ForecastPFN, lowest noise",MSE,0.207,

result/per_paper/2311.01933/accuracy_efficiency_traced.csv

@@ -0,0 +1,5 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,Data Budget = 50,ForecastPFN,MAE,50,,0,1,0
+accuracy,Data Budget = 500,ForecastPFN,MAE,500,,0,12,0
+accuracy,"ForecastPFN, low noise","ForecastPFN, lowest noise",MSE,0.359,,2,2,1
+accuracy,ForecastPFN,"ForecastPFN, lowest noise",MSE,0.207,,2,3,1

result/per_paper/2311.01933/components_architecture.csv

@@ -0,0 +1,6 @@
+component,what_it_is,provenance,citation,evidence
+Transformer Encoder,"The base architecture of ForecastPFN, consisting of one multi-head attention layer and two feedforward layers.",reused_cited,"2,25,39","As with the original PFNs [2,25,39], we use a transformer [56] as the base architecture. We use an encoder-based transformer, consisting of one multi-head attention layer and two feedforward layer"
+Multi-Head Attention Layer,A component of the transformer encoder that enables the model to attend to different parts of the input sequence simultaneously.,reused_cited,56,"As with the original PFNs [2,25,39], we use a transformer [56] as the base architecture. We use an encoder-based transformer, consisting of one multi-head attention layer and two feedforward layer"
+Feedforward Layers,Two sequential feedforward networks within the transformer encoder that process information after attention mechanisms.,reused_cited,56,"As with the original PFNs [2,25,39], we use a transformer [56] as the base architecture. We use an encoder-based transformer, consisting of one multi-head attention layer and two feedforward layer"
+Multiplicative Noise,A synthetic data generation technique used to balance signal-to-noise ratios across time series by scaling noise relative to the base series.,proposed_here,,We choose multiplicative noise to better balance the amount of signal to noise across all series.
+Weibull Distribution,"A synthetic data distribution parameterized to interpolate between Gaussian and exponential distributions, used to model noise in the synthetic training data.",proposed_here,,"The Weibull distribution is a simple and natural method to parameterize between Gaussian and exponential distributions, two types of distributions that frequently come up in real-world time series."

result/per_paper/2311.01933/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,"No hardware type (e.g., GPU/TPU) is explicitly mentioned in the text."
+num_devices,,,not_reported,"The number of devices (e.g., GPUs) used for training or inference is not reported."
+training_cost,,,not_reported,"Training cost (e.g., monetary or computational resources) is not quantified."
+training_batch_size,,,not_reported,The training batch size is not specified.
+training_steps_or_epochs,,,not_reported,The number of training steps or epochs is not reported.
+precision,,,not_reported,"The precision (e.g., 32-bit vs. 16-bit) used during training is not mentioned."
+inference_latency,0.2,seconds,stated,"The text explicitly states: 'producing predictions on a brand new dataset in just a single forward pass, taking 0.2 seconds.'"
+inference_throughput,,,not_reported,"Inference throughput (e.g., samples per second) is not quantified."
+peak_memory,,,not_reported,Peak memory usage during training or inference is not reported.
+flops_or_macs,,,not_reported,FLOPs or MACs (computational complexity) are not mentioned.
+num_inference_samples,,,not_reported,The number of inference samples evaluated is not specified.
+params,,,not_reported,The total number of model parameters is not reported.
+context_lengths_evaluated,,,not_reported,"The context lengths (e.g., input sequence lengths) evaluated during testing are not explicitly stated."
+horizon_lengths_evaluated,,,not_reported,"The prediction horizons (e.g., output sequence lengths) evaluated during testing are not explicitly stated."
+inference_batch_size,,,not_reported,The inference batch size is not specified.

result/per_paper/2311.11413/accuracy_efficiency.csv

@@ -0,0 +1,39 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,BasicMotions,LPTM,accuracy,1.00,
+accuracy,FaceDetection,LPTM,accuracy,0.79,
+accuracy,FingerMovements,LPTM,accuracy,0.78,
+accuracy,PEMS-SF,LPTM,accuracy,0.93,
+accuracy,RacketSports,LPTM,accuracy,0.93,
+accuracy,EigenWorms,LPTM,accuracy,0.94,
+accuracy,ArticularyWordRecognition,LPTM,accuracy,0.98,
+accuracy,AtrialFibrillation,LPTM,accuracy,0.93,
+accuracy,CharacterTrajectories,LPTM,accuracy,0.98,
+accuracy,Cricket,LPTM,accuracy,0.99,
+accuracy,DuckGeese,LPTM,accuracy,0.79,
+accuracy,Epilepsy,LPTM,accuracy,0.97,
+accuracy,ERing,LPTM,accuracy,0.97,
+accuracy,EthanolConcentration,LPTM,accuracy,0.53,
+accuracy,HandMovementDirection,LPTM,accuracy,0.53,
+accuracy,Handwriting,LPTM,accuracy,0.51,
+accuracy,Heartbeat,LPTM,accuracy,0.74,
+accuracy,InsectWingbeat,LPTM,accuracy,0.72,
+accuracy,JapaneseVowels,LPTM,accuracy,0.98,
+accuracy,Libras,LPTM,accuracy,0.95,
+accuracy,LSST,LPTM,accuracy,0.98,
+accuracy,MotorImagery,LPTM,accuracy,0.57,
+accuracy,NATOPS,LPTM,accuracy,0.94,
+accuracy,PenDigits,LPTM,accuracy,0.92,
+accuracy,Phoneme,LPTM,accuracy,0.32,
+accuracy,SelfRegulation,LPTM,accuracy,0.92,
+accuracy,SpokenArabicDigits,LPTM,accuracy,1.00,
+accuracy,StandWalkJump,LPTM,accuracy,0.58,
+accuracy,UWaveGesture,LPTM,accuracy,0.94,
+accuracy,PAMAP2,LPTM,accuracy,0.97,
+accuracy,OpportunityGestures,LPTM,accuracy,0.92,
+accuracy,OpportunityLocomotion,LPTM,accuracy,0.89,
+accuracy,SelfRegulationSCP2,LPTM,accuracy,0.691,
+accuracy,Occupancy,LPTM,accuracy,0.836,
+accuracy,MosquitoSound,LPTM,accuracy,0.715,
+accuracy,TS2Vec,LPTM,RMSE,2,
+accuracy,TS2Vec,LPTM,RMSE,2,
+accuracy,TS2Vec,LPTM,RMSE,2,

result/per_paper/2311.11413/accuracy_efficiency_traced.csv

@@ -0,0 +1,39 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,BasicMotions,LPTM,accuracy,1.00,,0,1,10
+accuracy,FaceDetection,LPTM,accuracy,0.79,,0,2,10
+accuracy,FingerMovements,LPTM,accuracy,0.78,,0,3,10
+accuracy,PEMS-SF,LPTM,accuracy,0.93,,0,4,10
+accuracy,RacketSports,LPTM,accuracy,0.93,,0,5,10
+accuracy,EigenWorms,LPTM,accuracy,0.94,,0,6,10
+accuracy,ArticularyWordRecognition,LPTM,accuracy,0.98,,0,7,10
+accuracy,AtrialFibrillation,LPTM,accuracy,0.93,,0,8,10
+accuracy,CharacterTrajectories,LPTM,accuracy,0.98,,0,9,10
+accuracy,Cricket,LPTM,accuracy,0.99,,0,10,10
+accuracy,DuckGeese,LPTM,accuracy,0.79,,0,11,10
+accuracy,Epilepsy,LPTM,accuracy,0.97,,0,12,10
+accuracy,ERing,LPTM,accuracy,0.97,,0,13,10
+accuracy,EthanolConcentration,LPTM,accuracy,0.53,,0,14,10
+accuracy,HandMovementDirection,LPTM,accuracy,0.53,,0,15,10
+accuracy,Handwriting,LPTM,accuracy,0.51,,0,16,10
+accuracy,Heartbeat,LPTM,accuracy,0.74,,0,17,10
+accuracy,InsectWingbeat,LPTM,accuracy,0.72,,0,18,10
+accuracy,JapaneseVowels,LPTM,accuracy,0.98,,0,19,10
+accuracy,Libras,LPTM,accuracy,0.95,,0,20,10
+accuracy,LSST,LPTM,accuracy,0.98,,0,21,10
+accuracy,MotorImagery,LPTM,accuracy,0.57,,0,22,10
+accuracy,NATOPS,LPTM,accuracy,0.94,,0,23,10
+accuracy,PenDigits,LPTM,accuracy,0.92,,0,24,10
+accuracy,Phoneme,LPTM,accuracy,0.32,,0,25,10
+accuracy,SelfRegulation,LPTM,accuracy,0.92,,0,26,10
+accuracy,SpokenArabicDigits,LPTM,accuracy,1.00,,0,27,10
+accuracy,StandWalkJump,LPTM,accuracy,0.58,,0,28,10
+accuracy,UWaveGesture,LPTM,accuracy,0.94,,0,29,10
+accuracy,PAMAP2,LPTM,accuracy,0.97,,0,30,10
+accuracy,OpportunityGestures,LPTM,accuracy,0.92,,0,31,10
+accuracy,OpportunityLocomotion,LPTM,accuracy,0.89,,0,32,10
+accuracy,SelfRegulationSCP2,LPTM,accuracy,0.691,,0,33,10
+accuracy,Occupancy,LPTM,accuracy,0.836,,0,34,10
+accuracy,MosquitoSound,LPTM,accuracy,0.715,,0,35,10
+accuracy,TS2Vec,LPTM,RMSE,2,,1,19,0
+accuracy,TS2Vec,LPTM,RMSE,2,,2,15,0
+accuracy,TS2Vec,LPTM,RMSE,2,,4,6,0

result/per_paper/2311.11413/components_architecture.csv

@@ -0,0 +1,7 @@
+component,what_it_is,provenance,citation,evidence
+Transformer-based Architecture,"The core model structure based on the transformer architecture, adapted for time-series data.",reused_cited,"Vaswani et al., 2017",LPTM uses a simple transformer-based architecture.
+Adaptive Segmentation Module,"A novel module that dynamically identifies optimal segmentation strategies for time-series data during pre-training, tailored to domain-specific characteristics.",proposed_here,,We propose a novel adaptive segmentation module that segments the time-series of each domain based on how well it performs on self-supervised pre-training.
+Segmentation Scoring Mechanism,A method within the adaptive segmentation module that evaluates segmentation effectiveness using losses from self-supervised learning tasks.,proposed_here,,The segmentation module uses a novel scoring mechanism during pre-training to identify an effective segmentation strategy for a domain.
+Masking-based Self-Supervised Pre-training,"A pre-training strategy where patches of input data are masked, and the model is trained to reconstruct them, enabling learning of temporal patterns without labeled data.",reused_cited,"Devlin et al., 2019",These tasks mask patches of the input data and train the model to...
+Time-Series Tokenization via Segmentation,"A method of converting raw time-series data into tokens by segmenting the series into meaningful chunks, replacing per-time-step inputs.",proposed_here,,We propose a novel method of adaptive segmentation that automatically identifies optimal dataset-specific segmentation strategy during pre-training.
+Multi-Domain Pre-Training Dataset,"A collection of heterogeneous time-series datasets from diverse domains (e.g., epidemiology, energy, economics) used to train the model for cross-domain generalization.",proposed_here,,"The entire set of heterogeneous multi-domain pre-train dataset is denoted as $D_{pre} = \{D'_{1}, D'_{2}, \ldots, D'_{K}\}$."

result/per_paper/2311.11413/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,"No hardware specifications (e.g., GPU/TPU types) are mentioned for LPTM."
+num_devices,,,not_reported,"The number of devices (e.g., GPUs) used for training or inference is not reported."
+training_cost,,,not_reported,"Training cost (e.g., computational resources, energy, or monetary cost) is not mentioned."
+training_batch_size,,,not_reported,Batch size during training is not specified.
+training_steps_or_epochs,,,not_reported,Number of training steps or epochs is not reported.
+precision,,,not_reported,"Training or inference precision (e.g., FP16, FP32) is not mentioned."
+inference_latency,,,not_reported,Latency per inference is not provided.
+inference_throughput,,,not_reported,"Throughput (e.g., samples/second) is not reported."
+peak_memory,,,not_reported,Peak memory usage during training or inference is not mentioned.
+flops_or_macs,,,not_reported,FLOPs or MACs (computational complexity) is not provided.
+num_inference_samples,,,not_reported,Number of samples used for inference evaluation is not specified.
+params,,,not_reported,Model parameter count is not reported.
+context_lengths_evaluated,,,not_reported,"Context lengths (e.g., sequence lengths) evaluated during testing are not mentioned."
+horizon_lengths_evaluated,,,not_reported,Forecasting horizons evaluated are not specified.
+inference_batch_size,,,not_reported,Batch size during inference is not reported.

result/per_paper/2401.03955/accuracy_efficiency.csv

@@ -0,0 +1,220 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,ETTH1,TTM_B,MSE,0.36,
+accuracy,ETTH1,TTM_E,MSE,0.362,
+accuracy,ETTH1,TTM_A,MSE,0.363,
+accuracy,ETTH2,TTM_B,MSE,0.269,
+accuracy,ETTH2,TTM_E,MSE,0.273,
+accuracy,ETTH2,TTM_A,MSE,0.262,
+accuracy,ETTM1,TTM_B,MSE,0.291,
+accuracy,ETTM1,TTM_E,MSE,0.293,
+accuracy,ETTM1,TTM_A,MSE,0.283,
+accuracy,ETTM2,TTM_B,MSE,0.164,
+accuracy,ETTM2,TTM_E,MSE,0.158,
+accuracy,ETTM2,TTM_A,MSE,0.158,
+accuracy,Weather,TTM_B,MSE,0.146,
+accuracy,Weather,TTM_E,MSE,0.154,
+accuracy,Weather,TTM_A,MSE,0.149,
+accuracy,Electricity,TTM_B,MSE,0.129,
+accuracy,Electricity,TTM_E,MSE,0.129,
+accuracy,Electricity,TTM_A,MSE,0.128,
+accuracy,Traffic,TTM_B,MSE,0.368,
+accuracy,Traffic,TTM_E,MSE,0.372,
+accuracy,Traffic,TTM_A,MSE,0.352,
+accuracy,ETTH1,TTM_B,MSE,0.373,
+accuracy,ETTH1,TTM_E,MSE,0.36,
+accuracy,ETTH1,TTM_A,MSE,0.362,
+accuracy,ETTH2,TTM_B,MSE,0.283,
+accuracy,ETTH2,TTM_E,MSE,0.269,
+accuracy,ETTH2,TTM_A,MSE,0.273,
+accuracy,ETTM1,TTM_B,MSE,0.286,
+accuracy,ETTM1,TTM_E,MSE,0.291,
+accuracy,ETTM1,TTM_A,MSE,0.293,
+accuracy,ETTM2,TTM_B,MSE,0.165,
+accuracy,ETTM2,TTM_E,MSE,0.164,
+accuracy,ETTM2,TTM_A,MSE,0.158,
+accuracy,Weather,TTM_B,MSE,0.144,
+accuracy,Weather,TTM_E,MSE,0.146,
+accuracy,Weather,TTM_A,MSE,0.154,
+accuracy,Electricity,TTM_B,MSE,0.13,
+accuracy,Electricity,TTM_E,MSE,0.129,
+accuracy,Electricity,TTM_A,MSE,0.129,
+accuracy,Traffic,TTM_B,MSE,0.367,
+accuracy,Traffic,TTM_E,MSE,0.368,
+accuracy,Traffic,TTM_A,MSE,0.372,
+accuracy,ETTH1,TTM_B,MAE,0.195,
+accuracy,ETTH1,TTM_E,MAE,0.243,
+accuracy,ETTH1,TTM_A,MAE,0.217,
+accuracy,ETTH2,TTM_B,MAE,0.12,
+accuracy,ETTH2,TTM_E,MAE,0.149,
+accuracy,ETTH2,TTM_A,MAE,0.158,
+accuracy,ETTM1,TTM_B,MAE,0.141,
+accuracy,ETTM1,TTM_E,MAE,0.247,
+accuracy,ETTM1,TTM_A,MAE,0.187,
+accuracy,ETTM2,TTM_B,MAE,0.04,
+accuracy,ETTM2,TTM_E,MAE,0.063,
+accuracy,ETTM2,TTM_A,MAE,0.068,
+accuracy,Weather,TTM_B,MAE,0.015,
+accuracy,Weather,TTM_E,MAE,0.015,
+accuracy,Weather,TTM_A,MAE,0.016,
+accuracy,Electricity,TTM_B,MAE,0.409,
+accuracy,Electricity,TTM_E,MAE,0.421,
+accuracy,Electricity,TTM_A,MAE,0.433,
+accuracy,Traffic,TTM_B,MAE,0.231,
+accuracy,Traffic,TTM_E,MAE,0.232,
+accuracy,Traffic,TTM_A,MAE,0.239,
+accuracy,ETTH1,TTM-B,MAE,0.364,
+accuracy,ETTH1,TTM-E,MAE,0.363,
+accuracy,ETTH1,TTM-A,MAE,0.359,
+accuracy,ETTH2,TTM-B,MAE,0.277,
+accuracy,ETTH2,TTM-E,MAE,0.271,
+accuracy,ETTH2,TTM-A,MAE,0.267,
+accuracy,ETTM1,TTM-B,MAE,0.313,
+accuracy,ETTM1,TTM-E,MAE,0.326,
+accuracy,ETTM1,TTM-A,MAE,0.317,
+accuracy,ETTM2,TTM-B,MAE,0.171,
+accuracy,ETTM2,TTM-E,MAE,0.178,
+accuracy,ETTM2,TTM-A,MAE,0.17,
+accuracy,Weather,TTM-B,MAE,0.154,
+accuracy,Weather,TTM-E,MAE,0.162,
+accuracy,Weather,TTM-A,MAE,0.155,
+accuracy,Electricity,TTM-B,MAE,0.146,
+accuracy,Electricity,TTM-E,MAE,0.15,
+accuracy,Electricity,TTM-A,MAE,0.141,
+accuracy,Traffic,TTM-B,MAE,0.411,
+accuracy,Traffic,TTM-E,MAE,0.411,
+accuracy,Traffic,TTM-A,MAE,0.469,
+accuracy,ETTH1,TTM-B,MAE,0.364,
+accuracy,ETTH1,TTM-E,MAE,0.363,
+accuracy,ETTH1,TTM-A,MAE,0.359,
+accuracy,ETTH2,TTM-B,MAE,0.277,
+accuracy,ETTH2,TTM-E,MAE,0.271,
+accuracy,ETTH2,TTM-A,MAE,0.264,
+accuracy,ETTM1,TTM-B,MAE,0.322,
+accuracy,ETTM1,TTM-E,MAE,0.327,
+accuracy,ETTM1,TTM-A,MAE,0.318,
+accuracy,ETTM2,TTM-B,MAE,0.171,
+accuracy,ETTM2,TTM-E,MAE,0.178,
+accuracy,ETTM2,TTM-A,MAE,0.169,
+accuracy,Weather,TTM-B,MAE,0.158,
+accuracy,Weather,TTM-E,MAE,0.166,
+accuracy,Weather,TTM-A,MAE,0.159,
+accuracy,Electricity,TTM-B,MAE,0.166,
+accuracy,Electricity,TTM-E,MAE,0.157,
+accuracy,Electricity,TTM-A,MAE,0.152,
+accuracy,ETTH1,TTM_Q,FL,0.365,
+accuracy,ETTH1,TTM_B,FL,0.364,
+accuracy,ETTH1,TTM_E,FL,0.363,
+accuracy,ETTH1,TTM_A,FL,0.359,
+accuracy,ETTH2,TTM_Q,FL,0.285,
+accuracy,ETTH2,TTM_B,FL,0.277,
+accuracy,ETTH2,TTM_E,FL,0.271,
+accuracy,ETTH2,TTM_A,FL,0.264,
+accuracy,ETTM1,TTM_Q,FL,0.413,
+accuracy,ETTM1,TTM_B,FL,0.322,
+accuracy,ETTM1,TTM_E,FL,0.327,
+accuracy,ETTM1,TTM_A,FL,0.318,
+accuracy,ETTM2,TTM_Q,FL,0.187,
+accuracy,ETTM2,TTM_B,FL,0.171,
+accuracy,ETTM2,TTM_E,FL,0.178,
+accuracy,ETTM2,TTM_A,FL,0.169,
+accuracy,Weather,TTM_Q,FL,0.154,
+accuracy,Weather,TTM_B,FL,0.158,
+accuracy,Weather,TTM_E,FL,0.166,
+accuracy,Weather,TTM_A,FL,0.159,
+accuracy,Electricity,TTM_Q,FL,0.169,
+accuracy,Electricity,TTM_B,FL,0.166,
+accuracy,Electricity,TTM_E,FL,0.157,
+accuracy,Electricity,TTM_A,FL,0.152,
+accuracy,Traffic,TTM_Q,FL,0.518,
+accuracy,Traffic,TTM_B,FL,0.514,
+accuracy,Traffic,TTM_E,FL,0.476,
+accuracy,Traffic,TTM_A,FL,0.462,
+accuracy,Model Size,TTM_Q,FL,1,
+accuracy,Model Size,TTM_B,FL,4,
+accuracy,Model Size,TTM_E,FL,5,
+accuracy,ETTH1,TTM_Q,MAE,0.366,
+accuracy,ETTH1,TTM_B,MAE,0.364,
+accuracy,ETTH1,TTM_E,MAE,0.363,
+accuracy,ETTH1,TTM_A,MAE,0.359,
+accuracy,ETTH2,TTM_Q,MAE,0.282,
+accuracy,ETTH2,TTM_B,MAE,0.277,
+accuracy,ETTH2,TTM_E,MAE,0.271,
+accuracy,ETTH2,TTM_A,MAE,0.267,
+accuracy,ETTM1,TTM_Q,MAE,0.359,
+accuracy,ETTM1,TTM_B,MAE,0.313,
+accuracy,ETTM1,TTM_E,MAE,0.326,
+accuracy,ETTM1,TTM_A,MAE,0.317,
+accuracy,ETTM2,TTM_Q,MAE,0.174,
+accuracy,ETTM2,TTM_B,MAE,0.171,
+accuracy,ETTM2,TTM_E,MAE,0.178,
+accuracy,ETTM2,TTM_A,MAE,0.17,
+accuracy,Weather,TTM_Q,MAE,0.152,
+accuracy,Weather,TTM_B,MAE,0.154,
+accuracy,Weather,TTM_E,MAE,0.162,
+accuracy,Weather,TTM_A,MAE,0.155,
+accuracy,Electricity,TTM_Q,MAE,0.142,
+accuracy,Electricity,TTM_B,MAE,0.146,
+accuracy,Electricity,TTM_E,MAE,0.15,
+accuracy,Electricity,TTM_A,MAE,0.141,
+accuracy,Traffic,TTM_Q,MAE,0.401,
+accuracy,Traffic,TTM_B,MAE,0.411,
+accuracy,Traffic,TTM_E,MAE,0.411,
+accuracy,Traffic,TTM_A,MAE,0.469,
+accuracy,Model Size,TTM_Q,MAE,1,
+accuracy,Model Size,TTM_B,MAE,4,
+accuracy,Model Size,TTM_E,MAE,5,
+accuracy,ETTH1,TTM_Q,MAE,0.398,
+accuracy,ETTH1,TTM_B,MAE,0.397,
+accuracy,ETTH1,TTM_E,MAE,0.402,
+accuracy,ETTH1,TTM_A,MAE,0.402,
+accuracy,ETTH2,TTM_Q,MAE,0.347,
+accuracy,ETTH2,TTM_B,MAE,0.338,
+accuracy,ETTH2,TTM_E,MAE,0.339,
+accuracy,ETTH2,TTM_A,MAE,0.332,
+accuracy,ETTM1,TTM_Q,MAE,0.35,
+accuracy,ETTM1,TTM_B,MAE,0.35,
+accuracy,ETTM1,TTM_E,MAE,0.35,
+accuracy,ETTM1,TTM_A,MAE,0.34,
+accuracy,ETTM2,TTM_Q,MAE,0.253,
+accuracy,ETTM2,TTM_B,MAE,0.252,
+accuracy,ETTM2,TTM_E,MAE,0.245,
+accuracy,ETTM2,TTM_A,MAE,0.252,
+accuracy,Weather,TTM_Q,MAE,0.224,
+accuracy,Weather,TTM_B,MAE,0.225,
+accuracy,Weather,TTM_E,MAE,0.234,
+accuracy,Weather,TTM_A,MAE,0.225,
+accuracy,Electricity,TTM_Q,MAE,0.161,
+accuracy,Electricity,TTM_B,MAE,0.16,
+accuracy,Electricity,TTM_E,MAE,0.158,
+accuracy,Electricity,TTM_A,MAE,0.156,
+accuracy,Traffic,TTM_Q,MAE,0.4,
+accuracy,Traffic,TTM_B,MAE,0.399,
+accuracy,Traffic,TTM_E,MAE,0.385,
+accuracy,Traffic,TTM_A,MAE,0.376,
+accuracy,ETTH1,TTM-B,MAE,0.394,
+accuracy,ETTH1,TTM-E,MAE,0.404,
+accuracy,ETTH1,TTM-A,MAE,0.4,
+accuracy,ETTH2,TTM-B,MAE,0.345,
+accuracy,ETTH2,TTM-E,MAE,0.335,
+accuracy,ETTH2,TTM-A,MAE,0.333,
+accuracy,ETTM1,TTM-B,MAE,0.386,
+accuracy,ETTM1,TTM-E,MAE,0.38,
+accuracy,ETTM1,TTM-A,MAE,0.362,
+accuracy,ETTM2,TTM-B,MAE,0.281,
+accuracy,ETTM2,TTM-E,MAE,0.271,
+accuracy,ETTM2,TTM-A,MAE,0.252,
+accuracy,Weather,TTM-B,MAE,0.237,
+accuracy,Weather,TTM-E,MAE,0.238,
+accuracy,Weather,TTM-A,MAE,0.231,
+accuracy,Electricity,TTM-B,MAE,0.205,
+accuracy,Electricity,TTM-E,MAE,0.194,
+accuracy,Electricity,TTM-A,MAE,0.192,
+accuracy,$TTM_B$ f-imp(%),TTM-B,MAE,6,
+accuracy,$TTM_B$ f-imp(%),TTM-E,MAE,1,
+accuracy,$TTM_B$ f-imp(%),TTM-A,MAE,4,
+accuracy,$TTM_E$ f-imp(%),TTM-B,MAE,7,
+accuracy,$TTM_E$ f-imp(%),TTM-E,MAE,1,
+accuracy,$TTM_E$ f-imp(%),TTM-A,MAE,6,
+accuracy,$TTM_A$ f-imp(%),TTM-B,MAE,10,
+accuracy,$TTM_A$ f-imp(%),TTM-E,MAE,4,
+accuracy,$TTM_A$ f-imp(%),TTM-A,MAE,9,

result/per_paper/2401.03955/accuracy_efficiency_traced.csv

@@ -0,0 +1,220 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,ETTH1,TTM_B,MSE,0.36,,0,2,2
+accuracy,ETTH1,TTM_E,MSE,0.362,,0,2,3
+accuracy,ETTH1,TTM_A,MSE,0.363,,0,2,4
+accuracy,ETTH2,TTM_B,MSE,0.269,,0,4,2
+accuracy,ETTH2,TTM_E,MSE,0.273,,0,4,3
+accuracy,ETTH2,TTM_A,MSE,0.262,,0,4,4
+accuracy,ETTM1,TTM_B,MSE,0.291,,0,6,2
+accuracy,ETTM1,TTM_E,MSE,0.293,,0,6,3
+accuracy,ETTM1,TTM_A,MSE,0.283,,0,6,4
+accuracy,ETTM2,TTM_B,MSE,0.164,,0,8,2
+accuracy,ETTM2,TTM_E,MSE,0.158,,0,8,3
+accuracy,ETTM2,TTM_A,MSE,0.158,,0,8,4
+accuracy,Weather,TTM_B,MSE,0.146,,0,10,2
+accuracy,Weather,TTM_E,MSE,0.154,,0,10,3
+accuracy,Weather,TTM_A,MSE,0.149,,0,10,4
+accuracy,Electricity,TTM_B,MSE,0.129,,0,12,2
+accuracy,Electricity,TTM_E,MSE,0.129,,0,12,3
+accuracy,Electricity,TTM_A,MSE,0.128,,0,12,4
+accuracy,Traffic,TTM_B,MSE,0.368,,0,14,2
+accuracy,Traffic,TTM_E,MSE,0.372,,0,14,3
+accuracy,Traffic,TTM_A,MSE,0.352,,0,14,4
+accuracy,ETTH1,TTM_B,MSE,0.373,,0,18,2
+accuracy,ETTH1,TTM_E,MSE,0.36,,0,18,3
+accuracy,ETTH1,TTM_A,MSE,0.362,,0,18,4
+accuracy,ETTH2,TTM_B,MSE,0.283,,0,22,2
+accuracy,ETTH2,TTM_E,MSE,0.269,,0,22,3
+accuracy,ETTH2,TTM_A,MSE,0.273,,0,22,4
+accuracy,ETTM1,TTM_B,MSE,0.286,,0,26,2
+accuracy,ETTM1,TTM_E,MSE,0.291,,0,26,3
+accuracy,ETTM1,TTM_A,MSE,0.293,,0,26,4
+accuracy,ETTM2,TTM_B,MSE,0.165,,0,30,2
+accuracy,ETTM2,TTM_E,MSE,0.164,,0,30,3
+accuracy,ETTM2,TTM_A,MSE,0.158,,0,30,4
+accuracy,Weather,TTM_B,MSE,0.144,,0,34,2
+accuracy,Weather,TTM_E,MSE,0.146,,0,34,3
+accuracy,Weather,TTM_A,MSE,0.154,,0,34,4
+accuracy,Electricity,TTM_B,MSE,0.13,,0,38,2
+accuracy,Electricity,TTM_E,MSE,0.129,,0,38,3
+accuracy,Electricity,TTM_A,MSE,0.129,,0,38,4
+accuracy,Traffic,TTM_B,MSE,0.367,,0,42,2
+accuracy,Traffic,TTM_E,MSE,0.368,,0,42,3
+accuracy,Traffic,TTM_A,MSE,0.372,,0,42,4
+accuracy,ETTH1,TTM_B,MAE,0.195,,1,1,2
+accuracy,ETTH1,TTM_E,MAE,0.243,,1,1,3
+accuracy,ETTH1,TTM_A,MAE,0.217,,1,1,4
+accuracy,ETTH2,TTM_B,MAE,0.12,,1,6,2
+accuracy,ETTH2,TTM_E,MAE,0.149,,1,6,3
+accuracy,ETTH2,TTM_A,MAE,0.158,,1,6,4
+accuracy,ETTM1,TTM_B,MAE,0.141,,1,11,2
+accuracy,ETTM1,TTM_E,MAE,0.247,,1,11,3
+accuracy,ETTM1,TTM_A,MAE,0.187,,1,11,4
+accuracy,ETTM2,TTM_B,MAE,0.04,,1,16,2
+accuracy,ETTM2,TTM_E,MAE,0.063,,1,16,3
+accuracy,ETTM2,TTM_A,MAE,0.068,,1,16,4
+accuracy,Weather,TTM_B,MAE,0.015,,1,21,2
+accuracy,Weather,TTM_E,MAE,0.015,,1,21,3
+accuracy,Weather,TTM_A,MAE,0.016,,1,21,4
+accuracy,Electricity,TTM_B,MAE,0.409,,1,26,2
+accuracy,Electricity,TTM_E,MAE,0.421,,1,26,3
+accuracy,Electricity,TTM_A,MAE,0.433,,1,26,4
+accuracy,Traffic,TTM_B,MAE,0.231,,1,31,2
+accuracy,Traffic,TTM_E,MAE,0.232,,1,31,3
+accuracy,Traffic,TTM_A,MAE,0.239,,1,31,4
+accuracy,ETTH1,TTM-B,MAE,0.364,,2,2,2
+accuracy,ETTH1,TTM-E,MAE,0.363,,2,2,3
+accuracy,ETTH1,TTM-A,MAE,0.359,,2,2,4
+accuracy,ETTH2,TTM-B,MAE,0.277,,2,6,2
+accuracy,ETTH2,TTM-E,MAE,0.271,,2,6,3
+accuracy,ETTH2,TTM-A,MAE,0.267,,2,6,4
+accuracy,ETTM1,TTM-B,MAE,0.313,,2,10,2
+accuracy,ETTM1,TTM-E,MAE,0.326,,2,10,3
+accuracy,ETTM1,TTM-A,MAE,0.317,,2,10,4
+accuracy,ETTM2,TTM-B,MAE,0.171,,2,14,2
+accuracy,ETTM2,TTM-E,MAE,0.178,,2,14,3
+accuracy,ETTM2,TTM-A,MAE,0.17,,2,14,4
+accuracy,Weather,TTM-B,MAE,0.154,,2,18,2
+accuracy,Weather,TTM-E,MAE,0.162,,2,18,3
+accuracy,Weather,TTM-A,MAE,0.155,,2,18,4
+accuracy,Electricity,TTM-B,MAE,0.146,,2,22,2
+accuracy,Electricity,TTM-E,MAE,0.15,,2,22,3
+accuracy,Electricity,TTM-A,MAE,0.141,,2,22,4
+accuracy,Traffic,TTM-B,MAE,0.411,,2,26,2
+accuracy,Traffic,TTM-E,MAE,0.411,,2,26,3
+accuracy,Traffic,TTM-A,MAE,0.469,,2,26,4
+accuracy,ETTH1,TTM-B,MAE,0.364,,3,1,2
+accuracy,ETTH1,TTM-E,MAE,0.363,,3,1,3
+accuracy,ETTH1,TTM-A,MAE,0.359,,3,1,4
+accuracy,ETTH2,TTM-B,MAE,0.277,,3,5,2
+accuracy,ETTH2,TTM-E,MAE,0.271,,3,5,3
+accuracy,ETTH2,TTM-A,MAE,0.264,,3,5,4
+accuracy,ETTM1,TTM-B,MAE,0.322,,3,9,2
+accuracy,ETTM1,TTM-E,MAE,0.327,,3,9,3
+accuracy,ETTM1,TTM-A,MAE,0.318,,3,9,4
+accuracy,ETTM2,TTM-B,MAE,0.171,,3,13,2
+accuracy,ETTM2,TTM-E,MAE,0.178,,3,13,3
+accuracy,ETTM2,TTM-A,MAE,0.169,,3,13,4
+accuracy,Weather,TTM-B,MAE,0.158,,3,17,2
+accuracy,Weather,TTM-E,MAE,0.166,,3,17,3
+accuracy,Weather,TTM-A,MAE,0.159,,3,17,4
+accuracy,Electricity,TTM-B,MAE,0.166,,3,21,2
+accuracy,Electricity,TTM-E,MAE,0.157,,3,21,3
+accuracy,Electricity,TTM-A,MAE,0.152,,3,21,4
+accuracy,ETTH1,TTM_Q,FL,0.365,,4,1,2
+accuracy,ETTH1,TTM_B,FL,0.364,,4,1,3
+accuracy,ETTH1,TTM_E,FL,0.363,,4,1,4
+accuracy,ETTH1,TTM_A,FL,0.359,,4,1,5
+accuracy,ETTH2,TTM_Q,FL,0.285,,4,5,2
+accuracy,ETTH2,TTM_B,FL,0.277,,4,5,3
+accuracy,ETTH2,TTM_E,FL,0.271,,4,5,4
+accuracy,ETTH2,TTM_A,FL,0.264,,4,5,5
+accuracy,ETTM1,TTM_Q,FL,0.413,,4,9,2
+accuracy,ETTM1,TTM_B,FL,0.322,,4,9,3
+accuracy,ETTM1,TTM_E,FL,0.327,,4,9,4
+accuracy,ETTM1,TTM_A,FL,0.318,,4,9,5
+accuracy,ETTM2,TTM_Q,FL,0.187,,4,13,2
+accuracy,ETTM2,TTM_B,FL,0.171,,4,13,3
+accuracy,ETTM2,TTM_E,FL,0.178,,4,13,4
+accuracy,ETTM2,TTM_A,FL,0.169,,4,13,5
+accuracy,Weather,TTM_Q,FL,0.154,,4,17,2
+accuracy,Weather,TTM_B,FL,0.158,,4,17,3
+accuracy,Weather,TTM_E,FL,0.166,,4,17,4
+accuracy,Weather,TTM_A,FL,0.159,,4,17,5
+accuracy,Electricity,TTM_Q,FL,0.169,,4,21,2
+accuracy,Electricity,TTM_B,FL,0.166,,4,21,3
+accuracy,Electricity,TTM_E,FL,0.157,,4,21,4
+accuracy,Electricity,TTM_A,FL,0.152,,4,21,5
+accuracy,Traffic,TTM_Q,FL,0.518,,4,25,2
+accuracy,Traffic,TTM_B,FL,0.514,,4,25,3
+accuracy,Traffic,TTM_E,FL,0.476,,4,25,4
+accuracy,Traffic,TTM_A,FL,0.462,,4,25,5
+accuracy,Model Size,TTM_Q,FL,1,,4,29,2
+accuracy,Model Size,TTM_B,FL,4,,4,29,3
+accuracy,Model Size,TTM_E,FL,5,,4,29,4
+accuracy,ETTH1,TTM_Q,MAE,0.366,,5,1,2
+accuracy,ETTH1,TTM_B,MAE,0.364,,5,1,3
+accuracy,ETTH1,TTM_E,MAE,0.363,,5,1,4
+accuracy,ETTH1,TTM_A,MAE,0.359,,5,1,5
+accuracy,ETTH2,TTM_Q,MAE,0.282,,5,5,2
+accuracy,ETTH2,TTM_B,MAE,0.277,,5,5,3
+accuracy,ETTH2,TTM_E,MAE,0.271,,5,5,4
+accuracy,ETTH2,TTM_A,MAE,0.267,,5,5,5
+accuracy,ETTM1,TTM_Q,MAE,0.359,,5,9,2
+accuracy,ETTM1,TTM_B,MAE,0.313,,5,9,3
+accuracy,ETTM1,TTM_E,MAE,0.326,,5,9,4
+accuracy,ETTM1,TTM_A,MAE,0.317,,5,9,5
+accuracy,ETTM2,TTM_Q,MAE,0.174,,5,13,2
+accuracy,ETTM2,TTM_B,MAE,0.171,,5,13,3
+accuracy,ETTM2,TTM_E,MAE,0.178,,5,13,4
+accuracy,ETTM2,TTM_A,MAE,0.17,,5,13,5
+accuracy,Weather,TTM_Q,MAE,0.152,,5,17,2
+accuracy,Weather,TTM_B,MAE,0.154,,5,17,3
+accuracy,Weather,TTM_E,MAE,0.162,,5,17,4
+accuracy,Weather,TTM_A,MAE,0.155,,5,17,5
+accuracy,Electricity,TTM_Q,MAE,0.142,,5,21,2
+accuracy,Electricity,TTM_B,MAE,0.146,,5,21,3
+accuracy,Electricity,TTM_E,MAE,0.15,,5,21,4
+accuracy,Electricity,TTM_A,MAE,0.141,,5,21,5
+accuracy,Traffic,TTM_Q,MAE,0.401,,5,25,2
+accuracy,Traffic,TTM_B,MAE,0.411,,5,25,3
+accuracy,Traffic,TTM_E,MAE,0.411,,5,25,4
+accuracy,Traffic,TTM_A,MAE,0.469,,5,25,5
+accuracy,Model Size,TTM_Q,MAE,1,,5,29,2
+accuracy,Model Size,TTM_B,MAE,4,,5,29,3
+accuracy,Model Size,TTM_E,MAE,5,,5,29,4
+accuracy,ETTH1,TTM_Q,MAE,0.398,,6,2,1
+accuracy,ETTH1,TTM_B,MAE,0.397,,6,2,2
+accuracy,ETTH1,TTM_E,MAE,0.402,,6,2,3
+accuracy,ETTH1,TTM_A,MAE,0.402,,6,2,4
+accuracy,ETTH2,TTM_Q,MAE,0.347,,6,3,1
+accuracy,ETTH2,TTM_B,MAE,0.338,,6,3,2
+accuracy,ETTH2,TTM_E,MAE,0.339,,6,3,3
+accuracy,ETTH2,TTM_A,MAE,0.332,,6,3,4
+accuracy,ETTM1,TTM_Q,MAE,0.35,,6,4,1
+accuracy,ETTM1,TTM_B,MAE,0.35,,6,4,2
+accuracy,ETTM1,TTM_E,MAE,0.35,,6,4,3
+accuracy,ETTM1,TTM_A,MAE,0.34,,6,4,4
+accuracy,ETTM2,TTM_Q,MAE,0.253,,6,5,1
+accuracy,ETTM2,TTM_B,MAE,0.252,,6,5,2
+accuracy,ETTM2,TTM_E,MAE,0.245,,6,5,3
+accuracy,ETTM2,TTM_A,MAE,0.252,,6,5,4
+accuracy,Weather,TTM_Q,MAE,0.224,,6,6,1
+accuracy,Weather,TTM_B,MAE,0.225,,6,6,2
+accuracy,Weather,TTM_E,MAE,0.234,,6,6,3
+accuracy,Weather,TTM_A,MAE,0.225,,6,6,4
+accuracy,Electricity,TTM_Q,MAE,0.161,,6,7,1
+accuracy,Electricity,TTM_B,MAE,0.16,,6,7,2
+accuracy,Electricity,TTM_E,MAE,0.158,,6,7,3
+accuracy,Electricity,TTM_A,MAE,0.156,,6,7,4
+accuracy,Traffic,TTM_Q,MAE,0.4,,6,8,1
+accuracy,Traffic,TTM_B,MAE,0.399,,6,8,2
+accuracy,Traffic,TTM_E,MAE,0.385,,6,8,3
+accuracy,Traffic,TTM_A,MAE,0.376,,6,8,4
+accuracy,ETTH1,TTM-B,MAE,0.394,,7,2,1
+accuracy,ETTH1,TTM-E,MAE,0.404,,7,2,2
+accuracy,ETTH1,TTM-A,MAE,0.4,,7,2,3
+accuracy,ETTH2,TTM-B,MAE,0.345,,7,3,1
+accuracy,ETTH2,TTM-E,MAE,0.335,,7,3,2
+accuracy,ETTH2,TTM-A,MAE,0.333,,7,3,3
+accuracy,ETTM1,TTM-B,MAE,0.386,,7,4,1
+accuracy,ETTM1,TTM-E,MAE,0.38,,7,4,2
+accuracy,ETTM1,TTM-A,MAE,0.362,,7,4,3
+accuracy,ETTM2,TTM-B,MAE,0.281,,7,5,1
+accuracy,ETTM2,TTM-E,MAE,0.271,,7,5,2
+accuracy,ETTM2,TTM-A,MAE,0.252,,7,5,3
+accuracy,Weather,TTM-B,MAE,0.237,,7,6,1
+accuracy,Weather,TTM-E,MAE,0.238,,7,6,2
+accuracy,Weather,TTM-A,MAE,0.231,,7,6,3
+accuracy,Electricity,TTM-B,MAE,0.205,,7,7,1
+accuracy,Electricity,TTM-E,MAE,0.194,,7,7,2
+accuracy,Electricity,TTM-A,MAE,0.192,,7,7,3
+accuracy,$TTM_B$ f-imp(%),TTM-B,MAE,6,,7,8,1
+accuracy,$TTM_B$ f-imp(%),TTM-E,MAE,1,,7,8,2
+accuracy,$TTM_B$ f-imp(%),TTM-A,MAE,4,,7,8,3
+accuracy,$TTM_E$ f-imp(%),TTM-B,MAE,7,,7,9,1
+accuracy,$TTM_E$ f-imp(%),TTM-E,MAE,1,,7,9,2
+accuracy,$TTM_E$ f-imp(%),TTM-A,MAE,6,,7,9,3
+accuracy,$TTM_A$ f-imp(%),TTM-B,MAE,10,,7,10,1
+accuracy,$TTM_A$ f-imp(%),TTM-E,MAE,4,,7,10,2
+accuracy,$TTM_A$ f-imp(%),TTM-A,MAE,9,,7,10,3

result/per_paper/2401.03955/components_architecture.csv

@@ -0,0 +1,7 @@
+component,what_it_is,provenance,citation,evidence
+MLPMixer Blocks,"A lightweight architectural component using multi-layer perceptron (MLP) operations to process time series data, replacing self-attention mechanisms in Transformers.",reused_cited,"TSMixer, 2021",TTM is based on the light-weight TSMixer architecture that uses MLPMixer blocks interleaved with simple gated attention as alternatives to the quadratic time-consuming self-attention blocks in Transformers.
+Gated Attention,A simplified attention mechanism that combines gating functions with attention to reduce computational complexity while maintaining performance.,reused_cited,"TSMixer, 2021",TTM is based on the light-weight TSMixer architecture that uses MLPMixer blocks interleaved with simple gated attention as alternatives to the quadratic time-consuming self-attention blocks in Transformers.
+Adaptive Patching,"A technique to dynamically partition time series data into variable-length patches during pre-training, enabling the model to handle diverse temporal resolutions.",proposed_here,,"TTM incorporates innovations like adaptive patching, diverse resolution sampling, and resolution prefix tuning to handle pre-training on varied dataset resolutions with minimal model capacity."
+Diverse Resolution Sampling,"A method to train the model on time series data with varying temporal resolutions (e.g., seconds to days) to improve generalization across different domains.",proposed_here,,"TTM incorporates innovations like adaptive patching, diverse resolution sampling, and resolution prefix tuning to handle pre-training on varied dataset resolutions with minimal model capacity."
+Resolution Prefix Tuning,A parameter-efficient technique to adapt the model to specific temporal resolutions during fine-tuning by modifying prefix embeddings.,proposed_here,,"TTM incorporates innovations like adaptive patching, diverse resolution sampling, and resolution prefix tuning to handle pre-training on varied dataset resolutions with minimal model capacity."
+Multi-level Modeling,A framework to capture cross-channel correlations (between target variables) and integrate exogenous signals (external influencing variables) during fine-tuning.,proposed_here,,"Additionally, it employs multi-level modeling to capture channel correlations and infuse exogenous signals during fine-tuning."

result/per_paper/2401.03955/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,CPU or GPU,not_applicable,stated,Outline of TTM's key capabilities: ... zero-shot inference and fine-tuning of TTM models can be easily executed with just one GPU or in CPU-only environments.
+num_devices,1,not_applicable,stated,Outline of TTM's key capabilities: ... zero-shot inference and fine-tuning of TTM models can be easily executed with just one GPU or in CPU-only environments.
+training_cost,,not_applicable,not_reported,Not mentioned in the text.
+training_batch_size,,not_applicable,not_reported,Not mentioned in the text.
+training_steps_or_epochs,,not_applicable,not_reported,Not mentioned in the text.
+precision,,not_applicable,not_reported,Not mentioned in the text.
+inference_latency,0.01,s,stated,Table: TTM_B | ... | CPU TIME (s): 0.01
+inference_throughput,,not_applicable,not_reported,Not mentioned in the text.
+peak_memory,0.06,GB,stated,Table: TTM_B | ... | MEM (GB): 0.06
+flops_or_macs,,not_applicable,not_reported,Not mentioned in the text.
+num_inference_samples,,not_applicable,not_reported,Not mentioned in the text.
+params,1,M,stated,Table: Size | 1M | 4M | 5M | ... for TTM_B
+context_lengths_evaluated,,not_applicable,not_reported,Not mentioned in the text.
+horizon_lengths_evaluated,,not_applicable,not_reported,Not mentioned in the text.
+inference_batch_size,,not_applicable,not_reported,Not mentioned in the text.

result/per_paper/2402.02368/accuracy_efficiency.csv

@@ -0,0 +1,108 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,
+accuracy,SOTA,Timer-12G,MAE,0.129,
+accuracy,SOTA,Timer-12G,MAE,0.149,
+accuracy,SOTA,Timer-12G,MAE,0.115,
+efficiency,MASK RATIO,Timer-12G,Δ%,25.0,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,ETTH1,Timer-12G,Δ%,0.278,
+efficiency,ETTH1,Timer-12G,Δ%,0.287,
+efficiency,ETTH1,Timer-12G,Δ%,0.297,
+efficiency,ETTH1,Timer-12G,Δ%,0.314,
+efficiency,ETTH2,Timer-12G,Δ%,0.166,
+efficiency,ETTH2,Timer-12G,Δ%,0.178,
+efficiency,ETTH2,Timer-12G,Δ%,0.190,
+efficiency,ETTH2,Timer-12G,Δ%,0.208,
+efficiency,ETTM1,Timer-12G,Δ%,0.328,
+efficiency,ETTM1,Timer-12G,Δ%,0.326,
+efficiency,ETTM1,Timer-12G,Δ%,0.360,
+efficiency,ETTM1,Timer-12G,Δ%,0.407,
+efficiency,ETTM2,Timer-12G,Δ%,0.133,
+efficiency,ETTM2,Timer-12G,Δ%,0.123,
+efficiency,ETTM2,Timer-12G,Δ%,0.136,
+efficiency,ETTM2,Timer-12G,Δ%,0.143,
+efficiency,ECL,Timer-12G,Δ%,0.130,
+efficiency,ECL,Timer-12G,Δ%,0.138,
+efficiency,ECL,Timer-12G,Δ%,0.149,
+efficiency,ECL,Timer-12G,Δ%,0.162,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.420,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.446,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.477,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.511,
+efficiency,WEATHER,Timer-12G,Δ%,0.129,
+efficiency,WEATHER,Timer-12G,Δ%,0.147,
+efficiency,WEATHER,Timer-12G,Δ%,0.125,
+efficiency,WEATHER,Timer-12G,Δ%,0.153,
+efficiency,PEMS03,Timer-12G,Δ%,0.120,
+efficiency,PEMS03,Timer-12G,Δ%,0.150,
+efficiency,PEMS03,Timer-12G,Δ%,0.198,
+efficiency,PEMS03,Timer-12G,Δ%,0.273,
+efficiency,PEMS04,Timer-12G,Δ%,0.146,
+efficiency,PEMS04,Timer-12G,Δ%,0.184,
+efficiency,PEMS04,Timer-12G,Δ%,0.236,
+efficiency,PEMS04,Timer-12G,Δ%,0.320,
+efficiency,PEMS07,Timer-12G,Δ%,0.125,
+efficiency,PEMS07,Timer-12G,Δ%,0.162,
+efficiency,PEMS07,Timer-12G,Δ%,0.214,
+efficiency,PEMS07,Timer-12G,Δ%,0.290,
+efficiency,PEMS08,Timer-12G,Δ%,0.139,
+efficiency,PEMS08,Timer-12G,Δ%,0.174,
+efficiency,PEMS08,Timer-12G,Δ%,0.236,
+efficiency,PEMS08,Timer-12G,Δ%,0.324,
+efficiency,MASK RATIO,Timer-12G,Δ%,25.0,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,
+efficiency,ETTH1,Timer-12G,Δ%,0.273,
+efficiency,ETTH1,Timer-12G,Δ%,0.283,
+efficiency,ETTH1,Timer-12G,Δ%,0.294,
+efficiency,ETTH1,Timer-12G,Δ%,0.312,
+efficiency,ETTH2,Timer-12G,Δ%,0.177,
+efficiency,ETTH2,Timer-12G,Δ%,0.186,
+efficiency,ETTH2,Timer-12G,Δ%,0.195,
+efficiency,ETTH2,Timer-12G,Δ%,0.209,
+efficiency,ETTM1,Timer-12G,Δ%,0.352,
+efficiency,ETTM1,Timer-12G,Δ%,0.345,
+efficiency,ETTM1,Timer-12G,Δ%,0.371,
+efficiency,ETTM1,Timer-12G,Δ%,0.413,
+efficiency,ETTM2,Timer-12G,Δ%,0.161,
+efficiency,ETTM2,Timer-12G,Δ%,0.171,
+efficiency,ETTM2,Timer-12G,Δ%,0.176,
+efficiency,ETTM2,Timer-12G,Δ%,0.158,
+efficiency,ECL,Timer-12G,Δ%,0.122,
+efficiency,ECL,Timer-12G,Δ%,0.130,
+efficiency,ECL,Timer-12G,Δ%,0.139,
+efficiency,ECL,Timer-12G,Δ%,0.152,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.392,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.414,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.443,
+efficiency,TRAFFIC,Timer-12G,Δ%,0.477,
+efficiency,WEATHER,Timer-12G,Δ%,0.157,
+efficiency,WEATHER,Timer-12G,Δ%,0.146,
+efficiency,WEATHER,Timer-12G,Δ%,0.147,
+efficiency,WEATHER,Timer-12G,Δ%,0.158,
+efficiency,PEMS03,Timer-12G,Δ%,0.108,
+efficiency,PEMS03,Timer-12G,Δ%,0.135,
+efficiency,PEMS03,Timer-12G,Δ%,0.179,
+efficiency,PEMS03,Timer-12G,Δ%,0.248,
+efficiency,PEMS04,Timer-12G,Δ%,0.134,
+efficiency,PEMS04,Timer-12G,Δ%,0.166,
+efficiency,PEMS04,Timer-12G,Δ%,0.216,
+efficiency,PEMS04,Timer-12G,Δ%,0.296,
+efficiency,PEMS07,Timer-12G,Δ%,0.114,
+efficiency,PEMS07,Timer-12G,Δ%,0.144,
+efficiency,PEMS07,Timer-12G,Δ%,0.189,
+efficiency,PEMS07,Timer-12G,Δ%,0.256,
+efficiency,PEMS08,Timer-12G,Δ%,0.129,
+efficiency,PEMS08,Timer-12G,Δ%,0.157,
+efficiency,PEMS08,Timer-12G,Δ%,0.206,
+efficiency,PEMS08,Timer-12G,Δ%,0.288,

result/per_paper/2402.02368/accuracy_efficiency_traced.csv

@@ -0,0 +1,108 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,,1,1,2
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,,1,1,4
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,,1,1,6
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,,1,1,8
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,,1,1,10
+accuracy,PRE-TRAINED,Timer-12G,MAE,12,,1,1,12
+accuracy,SOTA,Timer-12G,MAE,0.129,,1,14,2
+accuracy,SOTA,Timer-12G,MAE,0.149,,1,14,4
+accuracy,SOTA,Timer-12G,MAE,0.115,,1,14,6
+efficiency,MASK RATIO,Timer-12G,Δ%,25.0,,4,0,2
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,4,1,2
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,4,1,5
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,4,1,8
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,4,1,11
+efficiency,ETTH1,Timer-12G,Δ%,0.278,,4,2,2
+efficiency,ETTH1,Timer-12G,Δ%,0.287,,4,2,5
+efficiency,ETTH1,Timer-12G,Δ%,0.297,,4,2,8
+efficiency,ETTH1,Timer-12G,Δ%,0.314,,4,2,11
+efficiency,ETTH2,Timer-12G,Δ%,0.166,,4,3,2
+efficiency,ETTH2,Timer-12G,Δ%,0.178,,4,3,5
+efficiency,ETTH2,Timer-12G,Δ%,0.190,,4,3,8
+efficiency,ETTH2,Timer-12G,Δ%,0.208,,4,3,11
+efficiency,ETTM1,Timer-12G,Δ%,0.328,,4,4,2
+efficiency,ETTM1,Timer-12G,Δ%,0.326,,4,4,5
+efficiency,ETTM1,Timer-12G,Δ%,0.360,,4,4,8
+efficiency,ETTM1,Timer-12G,Δ%,0.407,,4,4,11
+efficiency,ETTM2,Timer-12G,Δ%,0.133,,4,5,2
+efficiency,ETTM2,Timer-12G,Δ%,0.123,,4,5,5
+efficiency,ETTM2,Timer-12G,Δ%,0.136,,4,5,8
+efficiency,ETTM2,Timer-12G,Δ%,0.143,,4,5,11
+efficiency,ECL,Timer-12G,Δ%,0.130,,4,6,2
+efficiency,ECL,Timer-12G,Δ%,0.138,,4,6,5
+efficiency,ECL,Timer-12G,Δ%,0.149,,4,6,8
+efficiency,ECL,Timer-12G,Δ%,0.162,,4,6,11
+efficiency,TRAFFIC,Timer-12G,Δ%,0.420,,4,7,2
+efficiency,TRAFFIC,Timer-12G,Δ%,0.446,,4,7,5
+efficiency,TRAFFIC,Timer-12G,Δ%,0.477,,4,7,8
+efficiency,TRAFFIC,Timer-12G,Δ%,0.511,,4,7,11
+efficiency,WEATHER,Timer-12G,Δ%,0.129,,4,8,2
+efficiency,WEATHER,Timer-12G,Δ%,0.147,,4,8,5
+efficiency,WEATHER,Timer-12G,Δ%,0.125,,4,8,8
+efficiency,WEATHER,Timer-12G,Δ%,0.153,,4,8,11
+efficiency,PEMS03,Timer-12G,Δ%,0.120,,4,9,2
+efficiency,PEMS03,Timer-12G,Δ%,0.150,,4,9,5
+efficiency,PEMS03,Timer-12G,Δ%,0.198,,4,9,8
+efficiency,PEMS03,Timer-12G,Δ%,0.273,,4,9,11
+efficiency,PEMS04,Timer-12G,Δ%,0.146,,4,10,2
+efficiency,PEMS04,Timer-12G,Δ%,0.184,,4,10,5
+efficiency,PEMS04,Timer-12G,Δ%,0.236,,4,10,8
+efficiency,PEMS04,Timer-12G,Δ%,0.320,,4,10,11
+efficiency,PEMS07,Timer-12G,Δ%,0.125,,4,11,2
+efficiency,PEMS07,Timer-12G,Δ%,0.162,,4,11,5
+efficiency,PEMS07,Timer-12G,Δ%,0.214,,4,11,8
+efficiency,PEMS07,Timer-12G,Δ%,0.290,,4,11,11
+efficiency,PEMS08,Timer-12G,Δ%,0.139,,4,12,2
+efficiency,PEMS08,Timer-12G,Δ%,0.174,,4,12,5
+efficiency,PEMS08,Timer-12G,Δ%,0.236,,4,12,8
+efficiency,PEMS08,Timer-12G,Δ%,0.324,,4,12,11
+efficiency,MASK RATIO,Timer-12G,Δ%,25.0,,5,0,2
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,5,1,2
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,5,1,5
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,5,1,8
+efficiency,PRE-TRAINED,Timer-12G,Δ%,12,,5,1,11
+efficiency,ETTH1,Timer-12G,Δ%,0.273,,5,2,2
+efficiency,ETTH1,Timer-12G,Δ%,0.283,,5,2,5
+efficiency,ETTH1,Timer-12G,Δ%,0.294,,5,2,8
+efficiency,ETTH1,Timer-12G,Δ%,0.312,,5,2,11
+efficiency,ETTH2,Timer-12G,Δ%,0.177,,5,3,2
+efficiency,ETTH2,Timer-12G,Δ%,0.186,,5,3,5
+efficiency,ETTH2,Timer-12G,Δ%,0.195,,5,3,8
+efficiency,ETTH2,Timer-12G,Δ%,0.209,,5,3,11
+efficiency,ETTM1,Timer-12G,Δ%,0.352,,5,4,2
+efficiency,ETTM1,Timer-12G,Δ%,0.345,,5,4,5
+efficiency,ETTM1,Timer-12G,Δ%,0.371,,5,4,8
+efficiency,ETTM1,Timer-12G,Δ%,0.413,,5,4,11
+efficiency,ETTM2,Timer-12G,Δ%,0.161,,5,5,2
+efficiency,ETTM2,Timer-12G,Δ%,0.171,,5,5,5
+efficiency,ETTM2,Timer-12G,Δ%,0.176,,5,5,8
+efficiency,ETTM2,Timer-12G,Δ%,0.158,,5,5,11
+efficiency,ECL,Timer-12G,Δ%,0.122,,5,6,2
+efficiency,ECL,Timer-12G,Δ%,0.130,,5,6,5
+efficiency,ECL,Timer-12G,Δ%,0.139,,5,6,8
+efficiency,ECL,Timer-12G,Δ%,0.152,,5,6,11
+efficiency,TRAFFIC,Timer-12G,Δ%,0.392,,5,7,2
+efficiency,TRAFFIC,Timer-12G,Δ%,0.414,,5,7,5
+efficiency,TRAFFIC,Timer-12G,Δ%,0.443,,5,7,8
+efficiency,TRAFFIC,Timer-12G,Δ%,0.477,,5,7,11
+efficiency,WEATHER,Timer-12G,Δ%,0.157,,5,8,2
+efficiency,WEATHER,Timer-12G,Δ%,0.146,,5,8,5
+efficiency,WEATHER,Timer-12G,Δ%,0.147,,5,8,8
+efficiency,WEATHER,Timer-12G,Δ%,0.158,,5,8,11
+efficiency,PEMS03,Timer-12G,Δ%,0.108,,5,9,2
+efficiency,PEMS03,Timer-12G,Δ%,0.135,,5,9,5
+efficiency,PEMS03,Timer-12G,Δ%,0.179,,5,9,8
+efficiency,PEMS03,Timer-12G,Δ%,0.248,,5,9,11
+efficiency,PEMS04,Timer-12G,Δ%,0.134,,5,10,2
+efficiency,PEMS04,Timer-12G,Δ%,0.166,,5,10,5
+efficiency,PEMS04,Timer-12G,Δ%,0.216,,5,10,8
+efficiency,PEMS04,Timer-12G,Δ%,0.296,,5,10,11
+efficiency,PEMS07,Timer-12G,Δ%,0.114,,5,11,2
+efficiency,PEMS07,Timer-12G,Δ%,0.144,,5,11,5
+efficiency,PEMS07,Timer-12G,Δ%,0.189,,5,11,8
+efficiency,PEMS07,Timer-12G,Δ%,0.256,,5,11,11
+efficiency,PEMS08,Timer-12G,Δ%,0.129,,5,12,2
+efficiency,PEMS08,Timer-12G,Δ%,0.157,,5,12,5
+efficiency,PEMS08,Timer-12G,Δ%,0.206,,5,12,8
+efficiency,PEMS08,Timer-12G,Δ%,0.288,,5,12,11

result/per_paper/2402.02368/components_architecture.csv

@@ -0,0 +1,7 @@
+component,what_it_is,provenance,citation,evidence
+GPT-style architecture,"A transformer-based architecture inspired by large language models, designed for autoregressive generation and flexible context length.",reused_cited,"Bengio et al., 2000","we adopt the GPT-style objective that predicts the next token (Bengio et al., 2000)"
+Single-series sequence (S3) format,A data format that converts heterogeneous multivariate time series into unified token sequences for pre-training.,proposed_here,,we propose the single-series sequence (S3) format that converts multivariate series with reserved patterns into unified token sequences
+Time Series Transformer (Timer),"A large-scale pre-trained model for time series analysis, adapted to forecasting, imputation, and anomaly detection.",proposed_here,,"we present Timer, a large-scale pre-trained Time Series Transformer"
+Next token prediction objective,"A pre-training task where the model predicts the next token in a sequence, similar to language models.",reused_cited,"Bengio et al., 2000","we adopt the GPT-style objective that predicts the next token (Bengio et al., 2000)"
+Autoregressive generation,"A generation mechanism where the model predicts subsequent tokens based on prior context, enabling multi-step forecasting.",reused_cited,"Bengio et al., 2000",Timer exhibits similar characteristics as large language models such as... autoregressive generation
+Flexible context length,A design feature allowing the model to handle variable-length input sequences during pre-training and adaptation.,reused_cited,"Bengio et al., 2000",Timer exhibits similar characteristics as large language models such as... flexible context length

result/per_paper/2402.02368/computational.csv

@@ -0,0 +1,16 @@
+field,value,unit,source,evidence
+hardware_type,,,not_reported,
+num_devices,,,not_reported,
+training_cost,,,not_reported,
+training_batch_size,,,not_reported,
+training_steps_or_epochs,,,not_reported,
+precision,,,not_reported,
+inference_latency,,,not_reported,
+inference_throughput,,,not_reported,
+peak_memory,,,not_reported,
+flops_or_macs,,,not_reported,
+num_inference_samples,,,not_reported,
+params,"29M, 50M, 67M",M,stated,Compute table under 'METHOD | TIMER(OURS)'
+context_lengths_evaluated,1440,tokens,stated,Compute table under 'METHOD | TIMER(OURS)'
+horizon_lengths_evaluated,96,time points,stated,Setup text: 'forecast length as 96'
+inference_batch_size,,,not_reported,

result/per_paper/2402.02592/accuracy_efficiency.csv

@@ -0,0 +1,289 @@
+metric_kind,entity,model_variant,metric,value,uncertainty
+accuracy,ETTh1,Moirai Small,MSE,0.419,
+accuracy,ETTh1,Moirai Base,MSE,0.423,
+accuracy,ETTh1,Moirai Large,MSE,0.448,
+accuracy,ETTh2,Moirai Small,MSE,0.348,
+accuracy,ETTh2,Moirai Base,MSE,0.745,
+accuracy,ETTh2,Moirai Large,MSE,0.584,
+accuracy,ETTm1,Moirai Small,MSE,0.367,
+accuracy,ETTm1,Moirai Base,MSE,0.404,
+accuracy,ETTm1,Moirai Large,MSE,0.426,
+accuracy,ETTm2,Moirai Small,MSE,0.259,
+accuracy,ETTm2,Moirai Base,MSE,0.287,
+accuracy,ETTm2,Moirai Large,MSE,0.366,
+accuracy,Electricity,Moirai Small,MSE,0.285,
+accuracy,Electricity,Moirai Base,MSE,0.219,
+accuracy,Electricity,Moirai Large,MSE,0.314,
+accuracy,Weather,Moirai Small,MSE,0.218,
+accuracy,Weather,Moirai Base,MSE,0.158,
+accuracy,Weather,Moirai Large,MSE,0.230,
+accuracy,M1 Monthly,Moirai Small,MAE,2,
+accuracy,M1 Monthly,Moirai Base,MAE,2,
+accuracy,M1 Monthly,Moirai Large,MAE,1,
+accuracy,M3 Monthly,Moirai Small,MAE,713.41,
+accuracy,M3 Monthly,Moirai Base,MAE,658.17,
+accuracy,M3 Monthly,Moirai Large,MAE,664.03,
+accuracy,M3 Other,Moirai Small,MAE,263.54,
+accuracy,M3 Other,Moirai Base,MAE,198.62,
+accuracy,M3 Other,Moirai Large,MAE,202.41,
+accuracy,M4 Monthly,Moirai Small,MAE,597.6,
+accuracy,M4 Monthly,Moirai Base,MAE,592.09,
+accuracy,M4 Monthly,Moirai Large,MAE,584.36,
+accuracy,M4 Weekly,Moirai Small,MAE,339.76,
+accuracy,M4 Weekly,Moirai Base,MAE,328.08,
+accuracy,M4 Weekly,Moirai Large,MAE,301.52,
+accuracy,M4 Daily,Moirai Small,MAE,189.1,
+accuracy,M4 Daily,Moirai Base,MAE,192.66,
+accuracy,M4 Daily,Moirai Large,MAE,189.78,
+accuracy,M4 Hourly,Moirai Small,MAE,268.04,
+accuracy,M4 Hourly,Moirai Base,MAE,209.87,
+accuracy,M4 Hourly,Moirai Large,MAE,197.79,
+accuracy,Tourism Quarterly,Moirai Small,MAE,18,
+accuracy,Tourism Quarterly,Moirai Base,MAE,17,
+accuracy,Tourism Quarterly,Moirai Large,MAE,15,
+accuracy,Tourism Monthly,Moirai Small,MAE,3,
+accuracy,Tourism Monthly,Moirai Base,MAE,2,
+accuracy,Tourism Monthly,Moirai Large,MAE,2,
+accuracy,CIF 2016,Moirai Small,MAE,655,
+accuracy,CIF 2016,Moirai Base,MAE,539,
+accuracy,CIF 2016,Moirai Large,MAE,695,
+accuracy,Aus. Elec. Demand,Moirai Small,MAE,266.57,
+accuracy,Aus. Elec. Demand,Moirai Base,MAE,201.39,
+accuracy,Aus. Elec. Demand,Moirai Large,MAE,177.68,
+accuracy,Bitcoin,Moirai Small,MAE,1.76,
+accuracy,Bitcoin,Moirai Base,MAE,1.62,
+accuracy,Bitcoin,Moirai Large,MAE,1.87,
+accuracy,Pedestrian Counts,Moirai Small,MAE,54.88,
+accuracy,Pedestrian Counts,Moirai Base,MAE,54.08,
+accuracy,Pedestrian Counts,Moirai Large,MAE,41.66,
+accuracy,Vehicle Trips,Moirai Small,MAE,24.46,
+accuracy,Vehicle Trips,Moirai Base,MAE,23.17,
+accuracy,Vehicle Trips,Moirai Large,MAE,21.85,
+accuracy,KDD cup,Moirai Small,MAE,39.81,
+accuracy,KDD cup,Moirai Base,MAE,38.66,
+accuracy,KDD cup,Moirai Large,MAE,39.09,
+accuracy,Weather,Moirai Small,MAE,1.96,
+accuracy,Weather,Moirai Base,MAE,1.8,
+accuracy,Weather,Moirai Large,MAE,1.75,
+accuracy,NN5 Daily,Moirai Small,MAE,5.37,
+accuracy,NN5 Daily,Moirai Base,MAE,4.26,
+accuracy,NN5 Daily,Moirai Large,MAE,3.77,
+accuracy,NN5 Weekly,Moirai Small,MAE,15.07,
+accuracy,NN5 Weekly,Moirai Base,MAE,16.42,
+accuracy,NN5 Weekly,Moirai Large,MAE,15.3,
+accuracy,Carparts,Moirai Small,MAE,0.53,
+accuracy,Carparts,Moirai Base,MAE,0.47,
+accuracy,Carparts,Moirai Large,MAE,0.49,
+accuracy,FRED-MD,Moirai Small,MAE,2,
+accuracy,FRED-MD,Moirai Base,MAE,2,
+accuracy,FRED-MD,Moirai Large,MAE,2,
+accuracy,Traffic Hourly,Moirai Small,MAE,0.02,
+accuracy,Traffic Hourly,Moirai Base,MAE,0.02,
+accuracy,Traffic Hourly,Moirai Large,MAE,0.01,
+accuracy,Traffic Weekly,Moirai Small,MAE,1.17,
+accuracy,Traffic Weekly,Moirai Base,MAE,1.14,
+accuracy,Traffic Weekly,Moirai Large,MAE,1.13,
+accuracy,Rideshare,Moirai Small,MAE,1.35,
+accuracy,Rideshare,Moirai Base,MAE,1.39,
+accuracy,Rideshare,Moirai Large,MAE,1.29,
+accuracy,Hospital,Moirai Small,MAE,23,
+accuracy,Hospital,Moirai Base,MAE,19.4,
+accuracy,Hospital,Moirai Large,MAE,19.44,
+accuracy,COVID Deaths,Moirai Small,MAE,124.32,
+accuracy,COVID Deaths,Moirai Base,MAE,126.11,
+accuracy,COVID Deaths,Moirai Large,MAE,117.11,
+accuracy,Temperature Rain,Moirai Small,MAE,5.3,
+accuracy,Temperature Rain,Moirai Base,MAE,5.08,
+accuracy,Temperature Rain,Moirai Large,MAE,5.27,
+accuracy,Sunspot,Moirai Small,MAE,0.11,
+accuracy,Sunspot,Moirai Base,MAE,0.08,
+accuracy,Sunspot,Moirai Large,MAE,0.13,
+accuracy,Saugeen River Flow,Moirai Small,MAE,24.07,
+accuracy,Saugeen River Flow,Moirai Base,MAE,24.4,
+accuracy,Saugeen River Flow,Moirai Large,MAE,24.76,
+accuracy,US Births,Moirai Small,MAE,872.51,
+accuracy,US Births,Moirai Base,MAE,624.3,
+accuracy,US Births,Moirai Large,MAE,476.5,
+accuracy,Electricity,MoiraiSmall,CRPS,0.072,
+accuracy,Electricity,MoiraiBase,CRPS,0.055,
+accuracy,Electricity,MoiraiLarge,CRPS,0.050,
+accuracy,MSIS,MoiraiSmall,CRPS,6.172,
+accuracy,MSIS,MoiraiBase,CRPS,5.875,
+accuracy,MSIS,MoiraiLarge,CRPS,5.744,
+accuracy,sMAPE,MoiraiSmall,CRPS,0.111,
+accuracy,sMAPE,MoiraiBase,CRPS,0.106,
+accuracy,sMAPE,MoiraiLarge,CRPS,0.107,
+accuracy,MASE,MoiraiSmall,CRPS,0.792,
+accuracy,MASE,MoiraiBase,CRPS,0.751,
+accuracy,MASE,MoiraiLarge,CRPS,0.753,
+accuracy,ND,MoiraiSmall,CRPS,0.069,
+accuracy,ND,MoiraiBase,CRPS,0.063,
+accuracy,ND,MoiraiLarge,CRPS,0.065,
+accuracy,NRMSE,MoiraiSmall,CRPS,0.551,
+accuracy,NRMSE,MoiraiBase,CRPS,0.465,
+accuracy,NRMSE,MoiraiLarge,CRPS,0.506,
+accuracy,Solar,MoiraiSmall,CRPS,0.471,
+accuracy,Solar,MoiraiBase,CRPS,0.419,
+accuracy,Solar,MoiraiLarge,CRPS,0.406,
+accuracy,MSIS,MoiraiSmall,CRPS,7.011,
+accuracy,MSIS,MoiraiBase,CRPS,6.250,
+accuracy,MSIS,MoiraiLarge,CRPS,8.447,
+accuracy,sMAPE,MoiraiSmall,CRPS,1.410,
+accuracy,sMAPE,MoiraiBase,CRPS,1.400,
+accuracy,sMAPE,MoiraiLarge,CRPS,1.501,
+accuracy,MASE,MoiraiSmall,CRPS,1.292,
+accuracy,MASE,MoiraiBase,CRPS,1.237,
+accuracy,MASE,MoiraiLarge,CRPS,1.607,
+accuracy,ND,MoiraiSmall,CRPS,0.551,
+accuracy,ND,MoiraiBase,CRPS,0.528,
+accuracy,ND,MoiraiLarge,CRPS,0.685,
+accuracy,NRMSE,MoiraiSmall,CRPS,1.034,
+accuracy,NRMSE,MoiraiBase,CRPS,1.014,
+accuracy,NRMSE,MoiraiLarge,CRPS,1.408,
+accuracy,Walmart,MoiraiSmall,CRPS,0.103,
+accuracy,Walmart,MoiraiBase,CRPS,0.093,
+accuracy,Walmart,MoiraiLarge,CRPS,0.098,
+accuracy,MSIS,MoiraiSmall,CRPS,8.421,
+accuracy,MSIS,MoiraiBase,CRPS,8.520,
+accuracy,MSIS,MoiraiLarge,CRPS,6.005,
+accuracy,sMAPE,MoiraiSmall,CRPS,0.168,
+accuracy,sMAPE,MoiraiBase,CRPS,0.174,
+accuracy,sMAPE,MoiraiLarge,CRPS,0.150,
+accuracy,MASE,MoiraiSmall,CRPS,0.964,
+accuracy,MASE,MoiraiBase,CRPS,1.007,
+accuracy,MASE,MoiraiLarge,CRPS,0.867,
+accuracy,ND,MoiraiSmall,CRPS,0.117,
+accuracy,ND,MoiraiBase,CRPS,0.124,
+accuracy,ND,MoiraiLarge,CRPS,0.105,
+accuracy,NRMSE,MoiraiSmall,CRPS,0.291,
+accuracy,NRMSE,MoiraiBase,CRPS,0.332,
+accuracy,NRMSE,MoiraiLarge,CRPS,0.218,
+accuracy,Weather,MoiraiSmall,CRPS,0.049,
+accuracy,Weather,MoiraiBase,CRPS,0.041,
+accuracy,Weather,MoiraiLarge,CRPS,0.051,
+accuracy,MSIS,MoiraiSmall,CRPS,5.136,
+accuracy,MSIS,MoiraiBase,CRPS,4.962,
+accuracy,MSIS,MoiraiLarge,CRPS,7.759,
+accuracy,sMAPE,MoiraiSmall,CRPS,0.623,
+accuracy,sMAPE,MoiraiBase,CRPS,0.688,
+accuracy,sMAPE,MoiraiLarge,CRPS,0.668,
+accuracy,MASE,MoiraiSmall,CRPS,0.487,
+accuracy,MASE,MoiraiBase,CRPS,0.515,
+accuracy,MASE,MoiraiLarge,CRPS,0.844,
+accuracy,ND,MoiraiSmall,CRPS,0.048,
+accuracy,ND,MoiraiBase,CRPS,0.063,
+accuracy,ND,MoiraiLarge,CRPS,0.072,
+accuracy,NRMSE,MoiraiSmall,CRPS,0.417,
+accuracy,NRMSE,MoiraiBase,CRPS,0.331,
+accuracy,NRMSE,MoiraiLarge,CRPS,0.260,
+accuracy,Istanbul Traffic,MoiraiSmall,CRPS,0.173,
+accuracy,Istanbul Traffic,MoiraiBase,CRPS,0.116,
+accuracy,Istanbul Traffic,MoiraiLarge,CRPS,0.112,
+accuracy,MSIS,MoiraiSmall,CRPS,4.461,
+accuracy,MSIS,MoiraiBase,CRPS,4.277,
+accuracy,MSIS,MoiraiLarge,CRPS,3.813,
+accuracy,sMAPE,MoiraiSmall,CRPS,0.284,
+accuracy,sMAPE,MoiraiBase,CRPS,0.288,
+accuracy,sMAPE,MoiraiLarge,CRPS,0.287,
+accuracy,MASE,MoiraiSmall,CRPS,0.644,
+accuracy,MASE,MoiraiBase,CRPS,0.631,
+accuracy,MASE,MoiraiLarge,CRPS,0.653,
+accuracy,ND,MoiraiSmall,CRPS,0.146,
+accuracy,ND,MoiraiBase,CRPS,0.143,
+accuracy,ND,MoiraiLarge,CRPS,0.148,
+accuracy,NRMSE,MoiraiSmall,CRPS,0.194,
+accuracy,NRMSE,MoiraiBase,CRPS,0.186,
+accuracy,NRMSE,MoiraiLarge,CRPS,0.190,
+accuracy,Turkey Power,MoiraiSmall,CRPS,0.048,
+accuracy,Turkey Power,MoiraiBase,CRPS,0.040,
+accuracy,Turkey Power,MoiraiLarge,CRPS,0.036,
+accuracy,MSIS,MoiraiSmall,CRPS,6.766,
+accuracy,MSIS,MoiraiBase,CRPS,6.341,
+accuracy,MSIS,MoiraiLarge,CRPS,8.978,
+accuracy,sMAPE,MoiraiSmall,CRPS,0.378,
+accuracy,sMAPE,MoiraiBase,CRPS,0.375,
+accuracy,sMAPE,MoiraiLarge,CRPS,0.416,
+accuracy,MASE,MoiraiSmall,CRPS,0.888,
+accuracy,MASE,MoiraiBase,CRPS,0.870,
+accuracy,MASE,MoiraiLarge,CRPS,1.234,
+accuracy,ND,MoiraiSmall,CRPS,0.051,
+accuracy,ND,MoiraiBase,CRPS,0.046,
+accuracy,ND,MoiraiLarge,CRPS,0.071,
+accuracy,NRMSE,MoiraiSmall,CRPS,0.118,
+accuracy,NRMSE,MoiraiBase,CRPS,0.102,
+accuracy,NRMSE,MoiraiLarge,CRPS,0.158,
+accuracy,Electricity,Moirai Small,CRPS,0.072,
+accuracy,Electricity,Moirai Base,CRPS,0.055,
+accuracy,Electricity,Moirai Large,CRPS,0.050,
+accuracy,MSIS,Moirai Small,CRPS,6.172,
+accuracy,MSIS,Moirai Base,CRPS,5.875,
+accuracy,MSIS,Moirai Large,CRPS,5.744,0.12
+accuracy,Solar,Moirai Small,CRPS,0.471,
+accuracy,Solar,Moirai Base,CRPS,0.419,
+accuracy,Solar,Moirai Large,CRPS,0.406,
+accuracy,MSIS,Moirai Small,CRPS,7.011,
+accuracy,MSIS,Moirai Base,CRPS,6.250,
+accuracy,MSIS,Moirai Large,CRPS,8.447,1.59
+accuracy,Walmart,Moirai Small,CRPS,0.103,
+accuracy,Walmart,Moirai Base,CRPS,0.093,
+accuracy,Walmart,Moirai Large,CRPS,0.098,
+accuracy,MSIS,Moirai Small,CRPS,8.421,
+accuracy,MSIS,Moirai Base,CRPS,8.520,
+accuracy,MSIS,Moirai Large,CRPS,6.005,0.21
+accuracy,Weather,Moirai Small,CRPS,0.049,
+accuracy,Weather,Moirai Base,CRPS,0.041,
+accuracy,Weather,Moirai Large,CRPS,0.051,
+accuracy,MSIS,Moirai Small,CRPS,5.136,
+accuracy,MSIS,Moirai Base,CRPS,4.962,
+accuracy,MSIS,Moirai Large,CRPS,7.759,0.49
+accuracy,Istanbul Traffic,Moirai Small,CRPS,0.173,
+accuracy,Istanbul Traffic,Moirai Base,CRPS,0.116,
+accuracy,Istanbul Traffic,Moirai Large,CRPS,0.112,
+accuracy,MSIS,Moirai Small,CRPS,4.461,
+accuracy,MSIS,Moirai Base,CRPS,4.277,
+accuracy,MSIS,Moirai Large,CRPS,3.813,0.09
+accuracy,Turkey Power,Moirai Small,CRPS,0.048,
+accuracy,Turkey Power,Moirai Base,CRPS,0.040,
+accuracy,Turkey Power,Moirai Large,CRPS,0.036,
+accuracy,MSIS,Moirai Small,CRPS,6.766,
+accuracy,MSIS,Moirai Base,CRPS,6.341,
+accuracy,MSIS,Moirai Large,CRPS,8.978,0.51
+accuracy,ETTh1,Moirai Small,MSE,0.400,
+accuracy,ETTh1,Moirai Base,MSE,0.434,
+accuracy,ETTh1,Moirai Large,MSE,0.510,
+accuracy,MAE,Moirai Small,MSE,0.438,
+accuracy,MAE,Moirai Base,MSE,0.469,
+accuracy,MAE,Moirai Large,MSE,0.448,
+accuracy,ETTh2,Moirai Small,MSE,0.341,
+accuracy,ETTh2,Moirai Base,MSE,0.345,
+accuracy,ETTh2,Moirai Large,MSE,0.354,
+accuracy,MAE,Moirai Small,MSE,0.382,
+accuracy,MAE,Moirai Base,MSE,0.376,
+accuracy,MAE,Moirai Large,MSE,0.407,
+accuracy,ETTm1,Moirai Small,MSE,0.448,
+accuracy,ETTm1,Moirai Base,MSE,0.381,
+accuracy,ETTm1,Moirai Large,MSE,0.390,
+accuracy,MAE,Moirai Small,MSE,0.388,
+accuracy,MAE,Moirai Base,MSE,0.389,
+accuracy,MAE,Moirai Large,MSE,0.410,
+accuracy,ETTm2,Moirai Small,MSE,0.300,
+accuracy,ETTm2,Moirai Base,MSE,0.272,
+accuracy,ETTm2,Moirai Large,MSE,0.276,
+accuracy,MAE,Moirai Small,MSE,0.321,
+accuracy,MAE,Moirai Base,MSE,0.320,
+accuracy,MAE,Moirai Large,MSE,0.332,
+accuracy,Electricity,Moirai Small,MSE,0.233,
+accuracy,Electricity,Moirai Base,MSE,0.188,
+accuracy,Electricity,Moirai Large,MSE,0.188,
+accuracy,MAE,Moirai Small,MSE,0.274,
+accuracy,MAE,Moirai Base,MSE,0.273,
+accuracy,MAE,Moirai Large,MSE,0.270,
+accuracy,Weather,Moirai Small,MSE,0.242,
+accuracy,Weather,Moirai Base,MSE,0.238,
+accuracy,Weather,Moirai Large,MSE,0.259,
+accuracy,MAE,Moirai Small,MSE,0.261,
+accuracy,MAE,Moirai Base,MSE,0.275,
+accuracy,MAE,Moirai Large,MSE,0.278,
+efficiency,$MOIRAI_{Small} (32)$,Moirai-Small,seconds,32,
+efficiency,$MOIRAI_{Base} (32)$,Moirai-Small,seconds,32,
+efficiency,$MOIRAI_{Large} (32)$,Moirai-Small,seconds,32,

result/per_paper/2402.02592/accuracy_efficiency_traced.csv

@@ -0,0 +1,289 @@
+metric_kind,entity,model_variant,metric,value,uncertainty,table,row,col
+accuracy,ETTh1,Moirai Small,MSE,0.419,,0,3,13
+accuracy,ETTh1,Moirai Base,MSE,0.423,,0,3,14
+accuracy,ETTh1,Moirai Large,MSE,0.448,,0,3,15
+accuracy,ETTh2,Moirai Small,MSE,0.348,,0,7,13
+accuracy,ETTh2,Moirai Base,MSE,0.745,,0,7,14
+accuracy,ETTh2,Moirai Large,MSE,0.584,,0,7,15
+accuracy,ETTm1,Moirai Small,MSE,0.367,,0,11,13
+accuracy,ETTm1,Moirai Base,MSE,0.404,,0,11,14
+accuracy,ETTm1,Moirai Large,MSE,0.426,,0,11,15
+accuracy,ETTm2,Moirai Small,MSE,0.259,,0,15,13
+accuracy,ETTm2,Moirai Base,MSE,0.287,,0,15,14
+accuracy,ETTm2,Moirai Large,MSE,0.366,,0,15,15
+accuracy,Electricity,Moirai Small,MSE,0.285,,0,19,13
+accuracy,Electricity,Moirai Base,MSE,0.219,,0,19,14
+accuracy,Electricity,Moirai Large,MSE,0.314,,0,19,15
+accuracy,Weather,Moirai Small,MSE,0.218,,0,23,13
+accuracy,Weather,Moirai Base,MSE,0.158,,0,23,14
+accuracy,Weather,Moirai Large,MSE,0.230,,0,23,15
+accuracy,M1 Monthly,Moirai Small,MAE,2,,1,1,1
+accuracy,M1 Monthly,Moirai Base,MAE,2,,1,1,2
+accuracy,M1 Monthly,Moirai Large,MAE,1,,1,1,3
+accuracy,M3 Monthly,Moirai Small,MAE,713.41,,1,2,1
+accuracy,M3 Monthly,Moirai Base,MAE,658.17,,1,2,2
+accuracy,M3 Monthly,Moirai Large,MAE,664.03,,1,2,3
+accuracy,M3 Other,Moirai Small,MAE,263.54,,1,3,1
+accuracy,M3 Other,Moirai Base,MAE,198.62,,1,3,2
+accuracy,M3 Other,Moirai Large,MAE,202.41,,1,3,3
+accuracy,M4 Monthly,Moirai Small,MAE,597.6,,1,4,1
+accuracy,M4 Monthly,Moirai Base,MAE,592.09,,1,4,2
+accuracy,M4 Monthly,Moirai Large,MAE,584.36,,1,4,3
+accuracy,M4 Weekly,Moirai Small,MAE,339.76,,1,5,1
+accuracy,M4 Weekly,Moirai Base,MAE,328.08,,1,5,2
+accuracy,M4 Weekly,Moirai Large,MAE,301.52,,1,5,3
+accuracy,M4 Daily,Moirai Small,MAE,189.1,,1,6,1
+accuracy,M4 Daily,Moirai Base,MAE,192.66,,1,6,2
+accuracy,M4 Daily,Moirai Large,MAE,189.78,,1,6,3
+accuracy,M4 Hourly,Moirai Small,MAE,268.04,,1,7,1
+accuracy,M4 Hourly,Moirai Base,MAE,209.87,,1,7,2
+accuracy,M4 Hourly,Moirai Large,MAE,197.79,,1,7,3
+accuracy,Tourism Quarterly,Moirai Small,MAE,18,,1,8,1
+accuracy,Tourism Quarterly,Moirai Base,MAE,17,,1,8,2
+accuracy,Tourism Quarterly,Moirai Large,MAE,15,,1,8,3
+accuracy,Tourism Monthly,Moirai Small,MAE,3,,1,9,1
+accuracy,Tourism Monthly,Moirai Base,MAE,2,,1,9,2
+accuracy,Tourism Monthly,Moirai Large,MAE,2,,1,9,3
+accuracy,CIF 2016,Moirai Small,MAE,655,,1,10,1
+accuracy,CIF 2016,Moirai Base,MAE,539,,1,10,2
+accuracy,CIF 2016,Moirai Large,MAE,695,,1,10,3
+accuracy,Aus. Elec. Demand,Moirai Small,MAE,266.57,,1,11,1
+accuracy,Aus. Elec. Demand,Moirai Base,MAE,201.39,,1,11,2
+accuracy,Aus. Elec. Demand,Moirai Large,MAE,177.68,,1,11,3
+accuracy,Bitcoin,Moirai Small,MAE,1.76,,1,12,1
+accuracy,Bitcoin,Moirai Base,MAE,1.62,,1,12,2
+accuracy,Bitcoin,Moirai Large,MAE,1.87,,1,12,3
+accuracy,Pedestrian Counts,Moirai Small,MAE,54.88,,1,13,1
+accuracy,Pedestrian Counts,Moirai Base,MAE,54.08,,1,13,2
+accuracy,Pedestrian Counts,Moirai Large,MAE,41.66,,1,13,3
+accuracy,Vehicle Trips,Moirai Small,MAE,24.46,,1,14,1
+accuracy,Vehicle Trips,Moirai Base,MAE,23.17,,1,14,2
+accuracy,Vehicle Trips,Moirai Large,MAE,21.85,,1,14,3
+accuracy,KDD cup,Moirai Small,MAE,39.81,,1,15,1
+accuracy,KDD cup,Moirai Base,MAE,38.66,,1,15,2
+accuracy,KDD cup,Moirai Large,MAE,39.09,,1,15,3
+accuracy,Weather,Moirai Small,MAE,1.96,,1,16,1
+accuracy,Weather,Moirai Base,MAE,1.8,,1,16,2
+accuracy,Weather,Moirai Large,MAE,1.75,,1,16,3
+accuracy,NN5 Daily,Moirai Small,MAE,5.37,,1,17,1
+accuracy,NN5 Daily,Moirai Base,MAE,4.26,,1,17,2
+accuracy,NN5 Daily,Moirai Large,MAE,3.77,,1,17,3
+accuracy,NN5 Weekly,Moirai Small,MAE,15.07,,1,18,1
+accuracy,NN5 Weekly,Moirai Base,MAE,16.42,,1,18,2
+accuracy,NN5 Weekly,Moirai Large,MAE,15.3,,1,18,3
+accuracy,Carparts,Moirai Small,MAE,0.53,,1,19,1
+accuracy,Carparts,Moirai Base,MAE,0.47,,1,19,2
+accuracy,Carparts,Moirai Large,MAE,0.49,,1,19,3
+accuracy,FRED-MD,Moirai Small,MAE,2,,1,20,1
+accuracy,FRED-MD,Moirai Base,MAE,2,,1,20,2
+accuracy,FRED-MD,Moirai Large,MAE,2,,1,20,3
+accuracy,Traffic Hourly,Moirai Small,MAE,0.02,,1,21,1
+accuracy,Traffic Hourly,Moirai Base,MAE,0.02,,1,21,2
+accuracy,Traffic Hourly,Moirai Large,MAE,0.01,,1,21,3
+accuracy,Traffic Weekly,Moirai Small,MAE,1.17,,1,22,1
+accuracy,Traffic Weekly,Moirai Base,MAE,1.14,,1,22,2
+accuracy,Traffic Weekly,Moirai Large,MAE,1.13,,1,22,3
+accuracy,Rideshare,Moirai Small,MAE,1.35,,1,23,1
+accuracy,Rideshare,Moirai Base,MAE,1.39,,1,23,2
+accuracy,Rideshare,Moirai Large,MAE,1.29,,1,23,3
+accuracy,Hospital,Moirai Small,MAE,23,,1,24,1
+accuracy,Hospital,Moirai Base,MAE,19.4,,1,24,2
+accuracy,Hospital,Moirai Large,MAE,19.44,,1,24,3
+accuracy,COVID Deaths,Moirai Small,MAE,124.32,,1,25,1
+accuracy,COVID Deaths,Moirai Base,MAE,126.11,,1,25,2
+accuracy,COVID Deaths,Moirai Large,MAE,117.11,,1,25,3
+accuracy,Temperature Rain,Moirai Small,MAE,5.3,,1,26,1
+accuracy,Temperature Rain,Moirai Base,MAE,5.08,,1,26,2
+accuracy,Temperature Rain,Moirai Large,MAE,5.27,,1,26,3
+accuracy,Sunspot,Moirai Small,MAE,0.11,,1,27,1
+accuracy,Sunspot,Moirai Base,MAE,0.08,,1,27,2
+accuracy,Sunspot,Moirai Large,MAE,0.13,,1,27,3
+accuracy,Saugeen River Flow,Moirai Small,MAE,24.07,,1,28,1
+accuracy,Saugeen River Flow,Moirai Base,MAE,24.4,,1,28,2
+accuracy,Saugeen River Flow,Moirai Large,MAE,24.76,,1,28,3
+accuracy,US Births,Moirai Small,MAE,872.51,,1,29,1
+accuracy,US Births,Moirai Base,MAE,624.3,,1,29,2
+accuracy,US Births,Moirai Large,MAE,476.5,,1,29,3
+accuracy,Electricity,MoiraiSmall,CRPS,0.072,,2,2,2
+accuracy,Electricity,MoiraiBase,CRPS,0.055,,2,2,3
+accuracy,Electricity,MoiraiLarge,CRPS,0.050,,2,2,4
+accuracy,MSIS,MoiraiSmall,CRPS,6.172,,2,3,2
+accuracy,MSIS,MoiraiBase,CRPS,5.875,,2,3,3
+accuracy,MSIS,MoiraiLarge,CRPS,5.744,,2,3,4
+accuracy,sMAPE,MoiraiSmall,CRPS,0.111,,2,4,2
+accuracy,sMAPE,MoiraiBase,CRPS,0.106,,2,4,3
+accuracy,sMAPE,MoiraiLarge,CRPS,0.107,,2,4,4
+accuracy,MASE,MoiraiSmall,CRPS,0.792,,2,5,2
+accuracy,MASE,MoiraiBase,CRPS,0.751,,2,5,3
+accuracy,MASE,MoiraiLarge,CRPS,0.753,,2,5,4
+accuracy,ND,MoiraiSmall,CRPS,0.069,,2,6,2
+accuracy,ND,MoiraiBase,CRPS,0.063,,2,6,3
+accuracy,ND,MoiraiLarge,CRPS,0.065,,2,6,4
+accuracy,NRMSE,MoiraiSmall,CRPS,0.551,,2,7,2
+accuracy,NRMSE,MoiraiBase,CRPS,0.465,,2,7,3
+accuracy,NRMSE,MoiraiLarge,CRPS,0.506,,2,7,4
+accuracy,Solar,MoiraiSmall,CRPS,0.471,,2,8,2
+accuracy,Solar,MoiraiBase,CRPS,0.419,,2,8,3
+accuracy,Solar,MoiraiLarge,CRPS,0.406,,2,8,4
+accuracy,MSIS,MoiraiSmall,CRPS,7.011,,2,9,2
+accuracy,MSIS,MoiraiBase,CRPS,6.250,,2,9,3
+accuracy,MSIS,MoiraiLarge,CRPS,8.447,,2,9,4
+accuracy,sMAPE,MoiraiSmall,CRPS,1.410,,2,10,2
+accuracy,sMAPE,MoiraiBase,CRPS,1.400,,2,10,3
+accuracy,sMAPE,MoiraiLarge,CRPS,1.501,,2,10,4
+accuracy,MASE,MoiraiSmall,CRPS,1.292,,2,11,2
+accuracy,MASE,MoiraiBase,CRPS,1.237,,2,11,3
+accuracy,MASE,MoiraiLarge,CRPS,1.607,,2,11,4
+accuracy,ND,MoiraiSmall,CRPS,0.551,,2,12,2
+accuracy,ND,MoiraiBase,CRPS,0.528,,2,12,3
+accuracy,ND,MoiraiLarge,CRPS,0.685,,2,12,4
+accuracy,NRMSE,MoiraiSmall,CRPS,1.034,,2,13,2
+accuracy,NRMSE,MoiraiBase,CRPS,1.014,,2,13,3
+accuracy,NRMSE,MoiraiLarge,CRPS,1.408,,2,13,4
+accuracy,Walmart,MoiraiSmall,CRPS,0.103,,2,14,2
+accuracy,Walmart,MoiraiBase,CRPS,0.093,,2,14,3
+accuracy,Walmart,MoiraiLarge,CRPS,0.098,,2,14,4
+accuracy,MSIS,MoiraiSmall,CRPS,8.421,,2,15,2
+accuracy,MSIS,MoiraiBase,CRPS,8.520,,2,15,3
+accuracy,MSIS,MoiraiLarge,CRPS,6.005,,2,15,4
+accuracy,sMAPE,MoiraiSmall,CRPS,0.168,,2,16,2
+accuracy,sMAPE,MoiraiBase,CRPS,0.174,,2,16,3
+accuracy,sMAPE,MoiraiLarge,CRPS,0.150,,2,16,4
+accuracy,MASE,MoiraiSmall,CRPS,0.964,,2,17,2
+accuracy,MASE,MoiraiBase,CRPS,1.007,,2,17,3
+accuracy,MASE,MoiraiLarge,CRPS,0.867,,2,17,4
+accuracy,ND,MoiraiSmall,CRPS,0.117,,2,18,2
+accuracy,ND,MoiraiBase,CRPS,0.124,,2,18,3
+accuracy,ND,MoiraiLarge,CRPS,0.105,,2,18,4
+accuracy,NRMSE,MoiraiSmall,CRPS,0.291,,2,19,2
+accuracy,NRMSE,MoiraiBase,CRPS,0.332,,2,19,3
+accuracy,NRMSE,MoiraiLarge,CRPS,0.218,,2,19,4
+accuracy,Weather,MoiraiSmall,CRPS,0.049,,2,20,2
+accuracy,Weather,MoiraiBase,CRPS,0.041,,2,20,3
+accuracy,Weather,MoiraiLarge,CRPS,0.051,,2,20,4
+accuracy,MSIS,MoiraiSmall,CRPS,5.136,,2,21,2
+accuracy,MSIS,MoiraiBase,CRPS,4.962,,2,21,3
+accuracy,MSIS,MoiraiLarge,CRPS,7.759,,2,21,4
+accuracy,sMAPE,MoiraiSmall,CRPS,0.623,,2,22,2
+accuracy,sMAPE,MoiraiBase,CRPS,0.688,,2,22,3
+accuracy,sMAPE,MoiraiLarge,CRPS,0.668,,2,22,4
+accuracy,MASE,MoiraiSmall,CRPS,0.487,,2,23,2
+accuracy,MASE,MoiraiBase,CRPS,0.515,,2,23,3
+accuracy,MASE,MoiraiLarge,CRPS,0.844,,2,23,4
+accuracy,ND,MoiraiSmall,CRPS,0.048,,2,24,2
+accuracy,ND,MoiraiBase,CRPS,0.063,,2,24,3
+accuracy,ND,MoiraiLarge,CRPS,0.072,,2,24,4
+accuracy,NRMSE,MoiraiSmall,CRPS,0.417,,2,25,2
+accuracy,NRMSE,MoiraiBase,CRPS,0.331,,2,25,3
+accuracy,NRMSE,MoiraiLarge,CRPS,0.260,,2,25,4
+accuracy,Istanbul Traffic,MoiraiSmall,CRPS,0.173,,2,26,2
+accuracy,Istanbul Traffic,MoiraiBase,CRPS,0.116,,2,26,3
+accuracy,Istanbul Traffic,MoiraiLarge,CRPS,0.112,,2,26,4
+accuracy,MSIS,MoiraiSmall,CRPS,4.461,,2,27,2
+accuracy,MSIS,MoiraiBase,CRPS,4.277,,2,27,3
+accuracy,MSIS,MoiraiLarge,CRPS,3.813,,2,27,4
+accuracy,sMAPE,MoiraiSmall,CRPS,0.284,,2,28,2
+accuracy,sMAPE,MoiraiBase,CRPS,0.288,,2,28,3
+accuracy,sMAPE,MoiraiLarge,CRPS,0.287,,2,28,4
+accuracy,MASE,MoiraiSmall,CRPS,0.644,,2,29,2
+accuracy,MASE,MoiraiBase,CRPS,0.631,,2,29,3
+accuracy,MASE,MoiraiLarge,CRPS,0.653,,2,29,4
+accuracy,ND,MoiraiSmall,CRPS,0.146,,2,30,2
+accuracy,ND,MoiraiBase,CRPS,0.143,,2,30,3
+accuracy,ND,MoiraiLarge,CRPS,0.148,,2,30,4
+accuracy,NRMSE,MoiraiSmall,CRPS,0.194,,2,31,2
+accuracy,NRMSE,MoiraiBase,CRPS,0.186,,2,31,3
+accuracy,NRMSE,MoiraiLarge,CRPS,0.190,,2,31,4
+accuracy,Turkey Power,MoiraiSmall,CRPS,0.048,,2,32,2
+accuracy,Turkey Power,MoiraiBase,CRPS,0.040,,2,32,3
+accuracy,Turkey Power,MoiraiLarge,CRPS,0.036,,2,32,4
+accuracy,MSIS,MoiraiSmall,CRPS,6.766,,2,33,2
+accuracy,MSIS,MoiraiBase,CRPS,6.341,,2,33,3
+accuracy,MSIS,MoiraiLarge,CRPS,8.978,,2,33,4
+accuracy,sMAPE,MoiraiSmall,CRPS,0.378,,2,34,2
+accuracy,sMAPE,MoiraiBase,CRPS,0.375,,2,34,3
+accuracy,sMAPE,MoiraiLarge,CRPS,0.416,,2,34,4
+accuracy,MASE,MoiraiSmall,CRPS,0.888,,2,35,2
+accuracy,MASE,MoiraiBase,CRPS,0.870,,2,35,3
+accuracy,MASE,MoiraiLarge,CRPS,1.234,,2,35,4
+accuracy,ND,MoiraiSmall,CRPS,0.051,,2,36,2
+accuracy,ND,MoiraiBase,CRPS,0.046,,2,36,3
+accuracy,ND,MoiraiLarge,CRPS,0.071,,2,36,4
+accuracy,NRMSE,MoiraiSmall,CRPS,0.118,,2,37,2
+accuracy,NRMSE,MoiraiBase,CRPS,0.102,,2,37,3
+accuracy,NRMSE,MoiraiLarge,CRPS,0.158,,2,37,4
+accuracy,Electricity,Moirai Small,CRPS,0.072,,3,2,2
+accuracy,Electricity,Moirai Base,CRPS,0.055,,3,2,3
+accuracy,Electricity,Moirai Large,CRPS,0.050,,3,2,4
+accuracy,MSIS,Moirai Small,CRPS,6.172,,3,3,2
+accuracy,MSIS,Moirai Base,CRPS,5.875,,3,3,3
+accuracy,MSIS,Moirai Large,CRPS,5.744,0.12,3,3,4
+accuracy,Solar,Moirai Small,CRPS,0.471,,3,4,2
+accuracy,Solar,Moirai Base,CRPS,0.419,,3,4,3
+accuracy,Solar,Moirai Large,CRPS,0.406,,3,4,4
+accuracy,MSIS,Moirai Small,CRPS,7.011,,3,5,2
+accuracy,MSIS,Moirai Base,CRPS,6.250,,3,5,3
+accuracy,MSIS,Moirai Large,CRPS,8.447,1.59,3,5,4
+accuracy,Walmart,Moirai Small,CRPS,0.103,,3,6,2
+accuracy,Walmart,Moirai Base,CRPS,0.093,,3,6,3
+accuracy,Walmart,Moirai Large,CRPS,0.098,,3,6,4
+accuracy,MSIS,Moirai Small,CRPS,8.421,,3,7,2
+accuracy,MSIS,Moirai Base,CRPS,8.520,,3,7,3
+accuracy,MSIS,Moirai Large,CRPS,6.005,0.21,3,7,4
+accuracy,Weather,Moirai Small,CRPS,0.049,,3,8,2
+accuracy,Weather,Moirai Base,CRPS,0.041,,3,8,3
+accuracy,Weather,Moirai Large,CRPS,0.051,,3,8,4
+accuracy,MSIS,Moirai Small,CRPS,5.136,,3,9,2
+accuracy,MSIS,Moirai Base,CRPS,4.962,,3,9,3
+accuracy,MSIS,Moirai Large,CRPS,7.759,0.49,3,9,4
+accuracy,Istanbul Traffic,Moirai Small,CRPS,0.173,,3,10,2
+accuracy,Istanbul Traffic,Moirai Base,CRPS,0.116,,3,10,3
+accuracy,Istanbul Traffic,Moirai Large,CRPS,0.112,,3,10,4
+accuracy,MSIS,Moirai Small,CRPS,4.461,,3,11,2
+accuracy,MSIS,Moirai Base,CRPS,4.277,,3,11,3
+accuracy,MSIS,Moirai Large,CRPS,3.813,0.09,3,11,4
+accuracy,Turkey Power,Moirai Small,CRPS,0.048,,3,12,2
+accuracy,Turkey Power,Moirai Base,CRPS,0.040,,3,12,3
+accuracy,Turkey Power,Moirai Large,CRPS,0.036,,3,12,4
+accuracy,MSIS,Moirai Small,CRPS,6.766,,3,13,2
+accuracy,MSIS,Moirai Base,CRPS,6.341,,3,13,3
+accuracy,MSIS,Moirai Large,CRPS,8.978,0.51,3,13,4
+accuracy,ETTh1,Moirai Small,MSE,0.400,,4,2,2
+accuracy,ETTh1,Moirai Base,MSE,0.434,,4,2,3
+accuracy,ETTh1,Moirai Large,MSE,0.510,,4,2,4
+accuracy,MAE,Moirai Small,MSE,0.438,,4,3,2
+accuracy,MAE,Moirai Base,MSE,0.469,,4,3,3
+accuracy,MAE,Moirai Large,MSE,0.448,,4,3,4
+accuracy,ETTh2,Moirai Small,MSE,0.341,,4,4,2
+accuracy,ETTh2,Moirai Base,MSE,0.345,,4,4,3
+accuracy,ETTh2,Moirai Large,MSE,0.354,,4,4,4
+accuracy,MAE,Moirai Small,MSE,0.382,,4,5,2
+accuracy,MAE,Moirai Base,MSE,0.376,,4,5,3
+accuracy,MAE,Moirai Large,MSE,0.407,,4,5,4
+accuracy,ETTm1,Moirai Small,MSE,0.448,,4,6,2
+accuracy,ETTm1,Moirai Base,MSE,0.381,,4,6,3
+accuracy,ETTm1,Moirai Large,MSE,0.390,,4,6,4
+accuracy,MAE,Moirai Small,MSE,0.388,,4,7,2
+accuracy,MAE,Moirai Base,MSE,0.389,,4,7,3
+accuracy,MAE,Moirai Large,MSE,0.410,,4,7,4
+accuracy,ETTm2,Moirai Small,MSE,0.300,,4,8,2
+accuracy,ETTm2,Moirai Base,MSE,0.272,,4,8,3
+accuracy,ETTm2,Moirai Large,MSE,0.276,,4,8,4
+accuracy,MAE,Moirai Small,MSE,0.321,,4,9,2
+accuracy,MAE,Moirai Base,MSE,0.320,,4,9,3
+accuracy,MAE,Moirai Large,MSE,0.332,,4,9,4
+accuracy,Electricity,Moirai Small,MSE,0.233,,4,10,2
+accuracy,Electricity,Moirai Base,MSE,0.188,,4,10,3
+accuracy,Electricity,Moirai Large,MSE,0.188,,4,10,4
+accuracy,MAE,Moirai Small,MSE,0.274,,4,11,2
+accuracy,MAE,Moirai Base,MSE,0.273,,4,11,3
+accuracy,MAE,Moirai Large,MSE,0.270,,4,11,4
+accuracy,Weather,Moirai Small,MSE,0.242,,4,12,2
+accuracy,Weather,Moirai Base,MSE,0.238,,4,12,3
+accuracy,Weather,Moirai Large,MSE,0.259,,4,12,4
+accuracy,MAE,Moirai Small,MSE,0.261,,4,13,2
+accuracy,MAE,Moirai Base,MSE,0.275,,4,13,3
+accuracy,MAE,Moirai Large,MSE,0.278,,4,13,4
+efficiency,$MOIRAI_{Small} (32)$,Moirai-Small,seconds,32,,5,2,0
+efficiency,$MOIRAI_{Base} (32)$,Moirai-Small,seconds,32,,5,3,0
+efficiency,$MOIRAI_{Large} (32)$,Moirai-Small,seconds,32,,5,4,0