markdown stringlengths 0 1.02M | code stringlengths 0 832k | output stringlengths 0 1.02M | license stringlengths 3 36 | path stringlengths 6 265 | repo_name stringlengths 6 127 |
|---|---|---|---|---|---|
3 Lowest HDI Countries | ax = africa_low.loc[africa_low['Country']=='Niger'].plot(x='Year', y= 'TFR', kind='scatter', c = 'cornflowerblue', label = 'Niger')
africa_low.loc[africa_low['Country']=='Central African Republic'].plot(x='Year', y= 'TFR', kind='scatter', c = 'mediumblue', ax=ax, label = 'Central African Republic')
africa_low.loc[afr... | _____no_output_____ | MIT | Python Analysis/Visualisations/Code/Visualisation_TFR_and_FLPR_of_Highest_and_Lowest_HDI_countries.ipynb | ChangHuaHua/QM2-Group-12 |
Hyperparameter tuning with Amazon SageMaker and Deep Graph Library with PyTorch backend_**Creating a Hyperparameter tuning job for a DGL network**_______ Contents1. [Background](Background) 2. [Setup](Setup) 3. [Tune](Train) 4. [Wrap-up](Wrap-up) BackgroundThis example notebook shows how to generate knowledge gra... | import sagemaker
from sagemaker import get_execution_role
from sagemaker.session import Session
# Setup session
sess = sagemaker.Session()
# S3 bucket for saving code and model artifacts.
# Feel free to specify a different bucket here if you wish.
bucket = sess.default_bucket()
# Location to put your custom code.
cu... | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
Now we'll import the Python libraries we'll need. | import boto3
from sagemaker.tuner import IntegerParameter, CategoricalParameter, ContinuousParameter, HyperparameterTuner | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
TuneSimilar to training a single training job in Amazon SageMaker, you define the training estimator passing in the code scripts, IAM role, (per job) hardware configuration, and any hyperparameters you're not tuning. | from sagemaker.pytorch import PyTorch
ENTRY_POINT = 'train.py'
CODE_PATH = './'
account = sess.boto_session.client('sts').get_caller_identity()['Account']
region = sess.boto_session.region_name
params = {}
params['dataset'] = 'FB15k'
params['model'] = 'DistMult'
params['batch_size'] = 1024
params['neg_sample_size'] ... | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
After you define your estimator, specify the hyperparameters you want to tune and their possible values. You have three different types of hyperparameters. * Categorical parameters need to take one value from a discrete set. Define this by passing the list of possible values to CategoricalParameter(list) * Continuous... | hyperparameter_ranges = {'lr': ContinuousParameter(0.01, 0.1),
'gamma': ContinuousParameter(400, 600)} | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
Next, specify the objective metric that you want to tune and its definition. This includes the regular expression needed to extract that metric from the Amazon CloudWatch logs of the training job.You can capture evalution results such as MR, MRR and Hit10. | metric = []
mr_metric = {'Name': 'final_MR', 'Regex':"Test average MR at \[\S*\]: (\S*)"}
mrr_metric = {'Name': 'final_MRR', 'Regex':"Test average MRR at \[\S*\]: (\S*)"}
hit10_metric = {'Name': 'final_Hit10', 'Regex':"Test average HITS@10 at \[\S*\]: (\S*)"}
metric.append(mr_metric)
metric.append(mrr_metric)
metric.ap... | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
Now, create a HyperparameterTuner object, which you pass. * The training estimator you created above * The hyperparameter ranges * Objective metric name and definition * Number of training jobs to run in-total and how many training jobs should be run simultaneously. More parallel jobs will finish tuning sooner, but may... | task_tags = [{'Key':'ML Task', 'Value':'DGL'}]
tuner = HyperparameterTuner(estimator,
objective_metric_name='final_MR',
objective_type='Minimize',
hyperparameter_ranges=hyperparameter_ranges,
metric_definitio... | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
And finally, you can start the tuning job by calling .fit(). | tuner.fit() | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
Run a quick check of the hyperparameter tuning jobs status to make sure it started successfully and is InProgress. | boto3.client('sagemaker').describe_hyper_parameter_tuning_job(
HyperParameterTuningJobName=tuner.latest_tuning_job.job_name)['HyperParameterTuningJobStatus'] | _____no_output_____ | Apache-2.0 | sagemaker-python-sdk/dgl_kge/kge_pytorch_hypertune.ipynb | P15241328/amazon-sagemaker-examples |
You can also run the notebook in [COLAB](https://colab.research.google.com/github/deepmipt/DeepPavlov/blob/master/examples/classification_tutorial.ipynb). | !pip3 install deeppavlov | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Classification on DeepPavlov **Task**:Intent recognition on SNIPS dataset: https://github.com/snipsco/nlu-benchmark/tree/master/2017-06-custom-intent-engines that has already been recomposed to `csv` format and can be downloaded from http://files.deeppavlov.ai/datasets/snips_intents/train.csvFastText English word embe... | from deeppavlov.core.data.utils import simple_download
#download train data file for SNIPS
simple_download(url="http://files.deeppavlov.ai/datasets/snips_intents/train.csv",
destination="./snips/train.csv")
! head -n 15 snips/train.csv | text,intents
Add another song to the Cita RomГЎntica playlist. ,AddToPlaylist
add clem burke in my playlist Pre-Party R&B Jams,AddToPlaylist
Add Live from Aragon Ballroom to Trapeo,AddToPlaylist
add Unite and Win to my night out,AddToPlaylist
Add track to my Digster Future Hits,AddToPlaylist
add the piano bar to ... | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
DatasetReaderRead data using `BasicClassificationDatasetReader` из DeepPavlov | from deeppavlov.dataset_readers.basic_classification_reader import BasicClassificationDatasetReader
# read data from particular columns of `.csv` file
dr = BasicClassificationDatasetReader().read(
data_path='./snips/',
train='train.csv',
x = 'text',
y = 'intents'
) | 2019-02-12 12:14:23.376 WARNING in 'deeppavlov.dataset_readers.basic_classification_reader'['basic_classification_reader'] at line 96: Cannot find snips/valid.csv file
2019-02-12 12:14:23.376 WARNING in 'deeppavlov.dataset_readers.basic_classification_reader'['basic_classification_reader'] at line 96: Cannot find snips... | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
We don't have a ready train/valid/test split. | # check train/valid/test sizes
[(k, len(dr[k])) for k in dr.keys()] | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
DatasetIteratorUse `BasicClassificationDatasetIterator` to split `train` on `train` and `valid` and to generate batches of samples. | from deeppavlov.dataset_iterators.basic_classification_iterator import BasicClassificationDatasetIterator
# initialize data iterator splitting `train` field to `train` and `valid` in proportion 0.8/0.2
train_iterator = BasicClassificationDatasetIterator(
data=dr,
field_to_split='train', # field that will be sp... | 2019-02-12 12:14:23.557 INFO in 'deeppavlov.dataset_iterators.basic_classification_iterator'['basic_classification_iterator'] at line 73: Splitting field <<train>> to new fields <<['train', 'valid']>>
| Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Let's look into training samples. | # one can get train instances (or any other data type including `all`)
x_train, y_train = train_iterator.get_instances(data_type='train')
for x, y in list(zip(x_train, y_train))[:5]:
print('x:', x)
print('y:', y)
print('=================') | x: Is it freezing in Offerman, California?
y: ['GetWeather']
=================
x: put this song in the playlist Trap Land
y: ['AddToPlaylist']
=================
x: show me a textbook with a rating of 2 and a maximum rating of 6 that is current
y: ['RateBook']
=================
x: Will the weather be okay in Northern Lu... | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Data preprocessing We will be using lowercasing and tokenization as data preparation. DeepPavlov also contains several other preprocessors and tokenizers. Lowercasing `str_lower` lowercases texts. | from deeppavlov.models.preprocessors.str_lower import str_lower
str_lower(['Is it freezing in Offerman, California?']) | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Tokenization`NLTKTokenizer` can split string to tokens. | from deeppavlov.models.tokenizers.nltk_moses_tokenizer import NLTKMosesTokenizer
tokenizer = NLTKMosesTokenizer()
tokenizer(['Is it freezing in Offerman, California?']) | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Let's preprocess all `train` part of the dataset. | train_x_lower_tokenized = str_lower(tokenizer(train_iterator.get_instances(data_type='train')[0])) | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
VocabularyNow we are ready to use `vocab`. They are very usefull for:* extracting class labels and converting labels to indices and vice versa,* building of characters or tokens vocabularies. | from deeppavlov.core.data.simple_vocab import SimpleVocabulary
# initialize simple vocabulary to collect all appeared in the dataset classes
classes_vocab = SimpleVocabulary(
save_path='./snips/classes.dict',
load_path='./snips/classes.dict')
classes_vocab.fit((train_iterator.get_instances(data_type='train')[1]... | 2019-02-12 12:14:25.35 INFO in 'deeppavlov.core.data.simple_vocab'['simple_vocab'] at line 89: [saving vocabulary to /home/vimary/ipavlov/Pilot/examples/tutorials/snips/classes.dict]
| Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Let's see what classes the dataset contains and their indices in the vocabulary. | list(classes_vocab.items())
# also one can collect vocabulary of textual tokens appeared 2 and more times in the dataset
token_vocab = SimpleVocabulary(
save_path='./snips/tokens.dict',
load_path='./snips/tokens.dict',
min_freq=2,
special_tokens=('<PAD>', '<UNK>',),
unk_token='<UNK>')
token_vocab.fi... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
FeaturizationThis part contains several possible ways of featurization of text samples. One can chose any appropriate vectorizer/embedder according to available resources and given task.Bag-of-words (BoW) and TF-IDF vectorizers converts text samples to vectors (one vector per sample) while fastText, GloVe, fastText we... | import numpy as np
from deeppavlov.models.embedders.bow_embedder import BoWEmbedder
# initialize bag-of-words embedder giving total number of tokens
bow = BoWEmbedder(depth=token_vocab.len)
# it assumes indexed tokenized samples
bow(token_vocab(str_lower(tokenizer(['Is it freezing in Offerman, California?']))))
# all 8... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
TF-IDF VectorizerMatches a vector to each text sample: text -> vector $v$ from $R^N$ where $N$ is a vocabulary size.$TF-IDF(token, document) = TF(token, document) * IDF(token, document)$$TF$ is a term frequency:$TF(token, document) = \frac{n_{token}}{\sum_{k}n_k}.$$IDF$ is a inverse document frequency:$IDF(token, all\... | from deeppavlov.models.sklearn import SklearnComponent
# initialize TF-IDF vectorizer sklearn component with `transform` as infer method
tfidf = SklearnComponent(
model_class="sklearn.feature_extraction.text:TfidfVectorizer",
infer_method="transform",
save_path='./tfidf_v0.pkl',
load_path='./tfidf_v0.pk... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
GloVe embedder[GloVe](https://nlp.stanford.edu/projects/glove/) is an unsupervised learning algorithm for obtaining vector representations for words. Training is performed on aggregated global word-word co-occurrence statistics from a corpus, and the resulting representations showcase interesting linear substructures ... | from deeppavlov.models.embedders.glove_embedder import GloVeEmbedder | Using TensorFlow backend.
| Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Let's download GloVe embedding file | simple_download(url="http://files.deeppavlov.ai/embeddings/glove.6B.100d.txt",
destination="./glove.6B.100d.txt")
embedder = GloVeEmbedder(load_path='./glove.6B.100d.txt',
dim=100, pad_zero=True)
# output shape is (batch_size x max_num_tokens_in_the_batch x embedding_dim)
embed... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Mean GloVe embedder Embedder returns a vector per token while we want to get a vector per text sample. Therefore, let's calculate mean vector of embeddings of tokens. For that we can either init `GloVeEmbedder` with `mean=True` parameter (`mean=false` by default), or pass `mean=true` while calling function (this way `... | # output shape is (batch_size x embedding_dim)
embedded_batch = embedder(str_lower(tokenizer(['Is it freezing in Offerman, California?'])), mean=True)
len(embedded_batch), embedded_batch[0].shape | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
GloVe weighted by TF-IDF embedderOne of the possible ways to combine TF-IDF vectorizer and any token embedder is to weigh token embeddings by TF-IDF coefficients (therefore, `mean` set to True is obligatory to obtain embeddings of interest while it still **by default** returns embeddings of tokens. | from deeppavlov.models.embedders.tfidf_weighted_embedder import TfidfWeightedEmbedder
weighted_embedder = TfidfWeightedEmbedder(
embedder=embedder, # our GloVe embedder instance
tokenizer=tokenizer, # our tokenizer instance
mean=True, # to return one vector per sample
vectorizer=tfidf # our TF-IDF v... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Models | from deeppavlov.metrics.accuracy import sets_accuracy
# get all train and valid data from iterator
x_train, y_train = train_iterator.get_instances(data_type="train")
x_valid, y_valid = train_iterator.get_instances(data_type="valid") | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Models in python SklearnComponent classifier on Tfidf-features in python | # initialize sklearn classifier, all parameters for classifier could be passed
cls = SklearnComponent(
model_class="sklearn.linear_model:LogisticRegression",
infer_method="predict",
save_path='./logreg_v0.pkl',
load_path='./logreg_v0.pkl',
C=1,
mode='train')
# fit sklearn classifier and save it
... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
KerasClassificationModel on GloVe embeddings in python | from deeppavlov.models.classifiers.keras_classification_model import KerasClassificationModel
from deeppavlov.models.preprocessors.one_hotter import OneHotter
from deeppavlov.models.classifiers.proba2labels import Proba2Labels
# Intialize `KerasClassificationModel` that composes CNN shallow-and-wide network
# (name he... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
SklearnComponent classifier on GloVe weighted by TF-IDF embeddings in python | # initialize sklearn classifier, all parameters for classifier could be passed
cls = SklearnComponent(
model_class="sklearn.linear_model:LogisticRegression",
infer_method="predict",
save_path='./logreg_v1.pkl',
load_path='./logreg_v1.pkl',
C=1,
mode='train')
# fit sklearn classifier and save it
... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Let's free our memory from embeddings and models | embedder.reset()
cls.reset() | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Models from configs | from deeppavlov import build_model
from deeppavlov import train_model | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
SklearnComponent classifier on Tfidf-features from config | logreg_config = {
"dataset_reader": {
"class_name": "basic_classification_reader",
"x": "text",
"y": "intents",
"data_path": "./snips"
},
"dataset_iterator": {
"class_name": "basic_classification_iterator",
"seed": 42,
"split_seed": 23,
"field_to_split": "train",
"split_fields"... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
KerasClassificationModel on GloVe embeddings from config | cnn_config = {
"dataset_reader": {
"class_name": "basic_classification_reader",
"x": "text",
"y": "intents",
"data_path": "snips"
},
"dataset_iterator": {
"class_name": "basic_classification_iterator",
"seed": 42,
"split_seed": 23,
"field_to_split": "train",
"split_fields": [
... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
SklearnComponent classifier on GloVe weighted by TF-IDF embeddings from config | logreg_config = {
"dataset_reader": {
"class_name": "basic_classification_reader",
"x": "text",
"y": "intents",
"data_path": "snips"
},
"dataset_iterator": {
"class_name": "basic_classification_iterator",
"seed": 42,
"split_seed": 23,
"field_to_split": "train",
"split_fields"... | _____no_output_____ | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Bonus: pre-trained CNN model in DeepPavlov Download model files (`wiki.en.bin` 8Gb embeddings): ! python -m deeppavlov download intents_snips_big Evaluate metrics on validation set (no test set provided): ! python -m deeppavlov evaluate intents_snips_big Or one can use model from python code: | from pathlib import Path
import deeppavlov
from deeppavlov import build_model, evaluate_model
from deeppavlov.download import deep_download
config_path = Path(deeppavlov.__file__).parent.joinpath('configs/classifiers/intents_snips_big.json')
# let's download all the required data - model files, embeddings, vocabulari... | 2018-12-13 18:45:33.675 WARNING in 'deeppavlov.dataset_readers.basic_classification_reader'['basic_classification_reader'] at line 97: Cannot find /home/dilyara/.deeppavlov/downloads/snips/valid.csv file
2018-12-13 18:45:33.675 WARNING in 'deeppavlov.dataset_readers.basic_classification_reader'['basic_classification_re... | Apache-2.0 | examples/classification_tutorial.ipynb | ayeffkay/DeepPavlov |
Auto MPG data | dataset_path = '/Users/mehdi/.keras/datasets/auto-mpg.data'
# read using pandas
column_names = ['MPG','Cylinders','Displacement','Horsepower','Weight',
'Acceleration', 'Model Year', 'Origin']
raw_dataset = pd.read_csv(dataset_path, names=column_names,
na_values = "?", comment='\t',... | _____no_output_____ | MIT | notebooks/trunk/regression-v2.ipynb | mehdirezaie/LSSutils |
The aim of this notebook is to explore the following questions: - [ ] Does CSR ongevellan have similar numbers as the incident data that has been provided by RWS - [ ] Is there a common key between the 2 datasets such that we can beef up RWS using Ongavellen. | rws = pd.read_sql('select * from rws_schema.ongevallen_raw;', con=conn)
csr = pd.read_sql('select * from rws_schema.incidents;', con=conn)
csr.head()
rws.columns
csr.columns
csr.inc_type.value_counts(normalize=True)
csr.loc[:,'inc_start'] = pd.to_datetime(csr.inc_start)
csr.loc[:,'date'] = csr.inc_start.apply(lambda x... | _____no_output_____ | MIT | notebooks/EDA - Incident data from CSR vs Ongevallen from RWS.ipynb | G-Simeone/Learning_Accident_Occurence_on_Dutch_Highways |
Do they have a common key? | # what are the common columns
c = set(csr.columns)
r = set(rws.columns)
c.intersection(r)
r.intersection(c) | _____no_output_____ | MIT | notebooks/EDA - Incident data from CSR vs Ongevallen from RWS.ipynb | G-Simeone/Learning_Accident_Occurence_on_Dutch_Highways |
Because column names have been edited in english, so there is no direct intersection | csr.loc[csr.inc_type=='Ongeval']
rws.head()
csr.shape
pd.to_numeric(rws.id_jaar.map(lambda x: x.split('.')[0])).describe() | _____no_output_____ | MIT | notebooks/EDA - Incident data from CSR vs Ongevallen from RWS.ipynb | G-Simeone/Learning_Accident_Occurence_on_Dutch_Highways |
Non Numerical datanon_numerical_data = train.select_dtypes(include="object")non_numerical_data.head(3)train.head() | #Numerical data
numerical_data = train.select_dtypes(exclude="object")
numerical_data.head(3)
train.head()
#Sub every empty postion with smtg
numericals = train.select_dtypes(include=[np.number]).columns.tolist()
numericals.remove("TomorrowRainForecast")
#Get categoricals
categoricals = train.select_dtypes(exclude=[np... | _____no_output_____ | MIT | Binary-Classification/it will rain tomorrow/notebooks/Binary Classification-Random Forest.ipynb | mamonteiro-brg/Lisbon-Data-Science-Academy |
Parcels Experiment:Expanding the polyline code to release particles at density based on local velocity normal to section._(Based on an experiment originally designed by Christina Schmidt.)__(Runs on GEOMAR Jupyter Server at https://schulung3.geomar.de/user/workshop007/lab)_ To do- Check/ask how OceanParcels deals wit... | %matplotlib inline
from parcels import (
AdvectionRK4_3D,
ErrorCode,
FieldSet,
JITParticle,
ParticleSet,
Variable
)
# from operator import attrgetter
from datetime import datetime, timedelta
import numpy as np
from pathlib import Path
import matplotlib.pyplot as plt
import cmocean as co
import... | INFO: Compiled ParcelsRandom ==> /tmp/parcels-62665/libparcels_random_657e0035-5181-471b-9b3b-09640069ddf8.so
| MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Experiment settings (user input) ParametersThese can be set in papermill | # OSNAP multiline details
sectionPathname = '../data/external/'
sectionFilename = 'osnap_pos_wp.txt'
sectionname = 'osnap'
# location of input data
path_name = '/data/iAtlantic/ocean-only/VIKING20X.L46-KKG36107B/nemo/output/'
experiment_name = 'VIKING20X.L46-KKG36107B'
data_resolution = '1m'
w_name_extension = '_repai... | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Derived variables | # times
t_0 = datetime.fromisoformat(t_0_str) # using monthly mean fields. Check dates.
t_start = datetime.fromisoformat(t_start_str)
# RNG seed based on release day (days since 1980-01-03)
RNG_seed = int((t_start - t_0).total_seconds() / (60*60*24))
# names of files to load
fname_U = f'1_{experiment_name}_{data_res... | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Construct input / output paths etc. | mesh_mask = mask_path / mesh_mask_filename
| _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Load input datasets | def fieldset_defintions(
list_of_filenames_U, list_of_filenames_V,
list_of_filenames_W, list_of_filenames_T,
mesh_mask
):
ds_mask = xr.open_dataset(mesh_mask)
filenames = {'U': {'lon': (mesh_mask),
'lat': (mesh_mask),
'depth': list_of_filenames_W[0]... | [PosixPath('/gxfs_work1/geomar/smomw355/model_data/ocean-only/VIKING20X.L46-KKG36107B/nemo/output/1_VIKING20X.L46-KKG36107B_5d_19800101_19801231_grid_U.nc'), PosixPath('/gxfs_work1/geomar/smomw355/model_data/ocean-only/VIKING20X.L46-KKG36107B/nemo/output/1_VIKING20X.L46-KKG36107B_5d_19810101_19811231_grid_U.nc'), Posix... | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Create Virtual Particles add a couple of simple plotting routines | def plot_section_sdist():
plt.figure(figsize=(10,5))
u = np.array([p.uvel for p in pset]) * degree2km * 1000.0 * np.cos(np.radians(pset.lat))
v = np.array([p.vvel for p in pset]) * degree2km * 1000.0
section_index = np.searchsorted(lonlat.lon,pset.lon)-1
u_normal = v * lonlatdiff.costheta[section_i... | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Create a set of particles with random initial positionsWe seed the RNG to be reproducible (and to be able to quickly create a second equivalent experiment with differently chosen compatible initial positions), and create arrays of random starting times, lats, lons, depths, and speed parameters (see kernel definitions ... | lonlat = xr.Dataset(pd.read_csv(sectionPath / sectionFilename,delim_whitespace=True))
lonlat.lon.attrs['long_name']='Longitude'
lonlat.lat.attrs['long_name']='Latitude'
lonlat.lon.attrs['standard_name']='longitude'
lonlat.lat.attrs['standard_name']='latitude'
lonlat.lon.attrs['units']='degrees_east'
lonlat.lat.attrs['u... | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Seed particles uniform random along OSNAP section | np.random.seed(RNG_seed)
# define time of release for each particle relative to t0
# can start each particle at a different time if required
# here all start at time t_start.
times = []
lons = []
lats = []
depths = []
# for subsect in range(lonlatdiff.length.shape[0]):
for subsect in range(start_vertex,end_vertex):
... | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Build particle set | %%time
pset = ParticleSet(
fieldset=fieldset,
pclass=SampleParticle,
lat=lat,
lon=lon,
# speed_param=speed_param,
depth=depth,
time=time
# repeatdt = repeatdt
)
print(f"Created {len(pset)} particles.")
# display(pset[:5])
# display(pset[-5:]) | Created 2643886 particles.
| MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Compose custom kernelWe'll create three additional kernels:- One Kernel adds velocity sampling- One Kernel adds temperature sampling- One kernel adds salinity samplingThen, we combine the builtin `AdvectionRK4_3D` kernel with these additional kernels. | def velocity_sampling(particle, fieldset, time):
'''Sample velocity.'''
(particle.uvel,particle.vvel) = fieldset.UV[time, particle.depth, particle.lat, particle.lon]
def temperature_sampling(particle, fieldset, time):
'''Sample temperature.'''
particle.temp = fieldset.T[time, particle.dep... | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Be able to handle errors during integrationWe have restricted our domain so in principle, particles could reach undefined positions.In that case, we want to just delete the particle (without forgetting its history). | def DeleteParticle(particle, fieldset, time):
particle.delete()
recovery_cases = {
ErrorCode.ErrorOutOfBounds: DeleteParticle,
ErrorCode.Error: DeleteParticle,
ErrorCode.ErrorInterpolation: DeleteParticle
} | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Run with runtime=0 to initialise fields | %%time
# with dask.config.set(**{'array.slicing.split_large_chunks': False}):
pset.execute(
custom_kernel,
runtime=0,
# dt=timedelta(minutes=0),
# output_file=outputfile,
recovery=recovery_cases
)
plot_section_sdist() | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Trim unwanted points from ParticleSetUse initialised fields to remove land points. We test `temp == 0.0` (the mask value over land). | t = np.array([p.temp for p in pset])
# u = np.array([p.uvel for p in pset])
# v = np.array([p.vvel for p in pset])
pset.remove_indices(np.argwhere(t == 0).flatten())
# pset.remove(np.argwhere(x * y * z == 0).flatten())
print(len(pset))
plot_section_sdist() | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Test velocity normal to section Velocity conversions from degrees lat/lon per second to m/s | u = np.array([p.uvel for p in pset])
v = np.array([p.vvel for p in pset])
u=u * degree2km * 1000.0 * np.cos(np.radians(pset.lat))
v=v * degree2km * 1000.0 | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
normal velocities | section_index = np.searchsorted(lonlat.lon,pset.lon)-1
u_normal = v * lonlatdiff.costheta[section_index].data - u * lonlatdiff.sintheta[section_index].data
abs(u_normal).max() | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
remove particles randomly with probability proportional to normal speed | u_random = np.random.rand(len(u_normal))*max_current
pset.remove_indices(np.argwhere(abs(u_normal) < u_random).flatten())
print(len(pset))
plot_section_sdist() | _____no_output_____ | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Prepare outputWe define an output file and specify the desired output frequency. | # output_filename = 'Parcels_IFFForwards_1m_June2016_2000.nc'
npart = str(len(pset))
output_filename = 'tracks_randomvel_mxl_'+sectionname+direction+year_str+month_str+day_str+'_N'+npart+'_D'+days+'_Rnd'+ seed+'.nc'
outfile = outpath / output_filename
print(outfile)
outputfile = pset.ParticleFile(
name=outfile,
... | ../data/raw/tracks_randomvel_mxl_osnap_backward_20191020_N59894_D3650_Rnd14535.nc
| MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Execute the experimentWe'll evolve particles, log their positions and variables to the output buffer and finally export the output to a the file. Run the experiment | %%time
# with dask.config.set(**{'array.slicing.split_large_chunks': False}):
pset.execute(
custom_kernel,
runtime=timedelta(days=runtime_in_days),
dt=timedelta(minutes=dt_in_minutes),
output_file=outputfile,
recovery=recovery_cases
)
# outputfile.export()
outputfile.close()
conda list
p... | Package Version
----------------------------- --------------------------
alembic 1.5.5
ansiwrap 0.8.4
anyio 2.2.0
appdirs 1.4.4
argon2-cffi 20.1.0
asciitree 0.3.3
... | MIT | notebooks/executed/037_afox_RunParcels_TS_MXL_Multiline_Randomvel_Papermill_executed_2019-10-20.ipynb | alanfox/spg_fresh_blob_202104 |
Plotly - Create Candlestick chart **Tags:** plotly chart candlestick trading dataviz Input Import libraries | import plotly.graph_objects as go
import pandas as pd
from datetime import datetime | _____no_output_____ | BSD-3-Clause | Plotly/Create Candlestick chart.ipynb | vivard/awesome-notebooks |
Model Read a csv and map the plot | df = pd.read_csv('https://raw.githubusercontent.com/plotly/datasets/master/finance-charts-apple.csv')
fig = go.Figure(data=[go.Candlestick(x=df['Date'],
open=df['AAPL.Open'],
high=df['AAPL.High'],
low=df['AAPL.Low'],
close=df['AAPL.Close'])]) | _____no_output_____ | BSD-3-Clause | Plotly/Create Candlestick chart.ipynb | vivard/awesome-notebooks |
Output Display result | fig.show() | _____no_output_____ | BSD-3-Clause | Plotly/Create Candlestick chart.ipynb | vivard/awesome-notebooks |
The idea is to do random patches but try out different methodologies regarding the sampling procedure. First, in the form of weighted samples where ideas from Breiman's Paper (pasting) and Adaboost can be used.Second, in the form of weighted features with respect to correlation (chi square, best of k?) between the sele... | from sklearn.cross_validation import cross_val_score
from sklearn.ensemble import BaggingClassifier
from sklearn.neighbors import KNeighborsClassifier
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
import itertools
import sklearn
from sklearn.linear_model import LogisticRe... | _____no_output_____ | MIT | RANDOM PATCHES WITH NON-UNIFORM SAMPLING.ipynb | kbogas/Cascada |
CutMix Callback> Callback to apply [CutMix](https://arxiv.org/pdf/1905.04899.pdf) data augmentation technique to the training data. From the [research paper](https://arxiv.org/pdf/1905.04899.pdf), `CutMix` is a way to combine two images. It comes from `MixUp` and `Cutout`. In this data augmentation technique:> patches... | #export
class CutMix(Callback):
"Implementation of `https://arxiv.org/abs/1905.04899`"
run_after,run_valid = [Normalize],False
def __init__(self, alpha=1.): self.distrib = Beta(tensor(alpha), tensor(alpha))
def before_fit(self):
self.stack_y = getattr(self.learn.loss_func, 'y_int', False)
... | _____no_output_____ | Apache-2.0 | nbs/74_callback.cutmix.ipynb | hanshin-back/fastai |
How does the batch with `CutMix` data augmentation technique look like? First, let's quickly create the `dls` using `ImageDataLoaders.from_name_re` DataBlocks API. | path = untar_data(URLs.PETS)
pat = r'([^/]+)_\d+.*$'
fnames = get_image_files(path/'images')
item_tfms = [Resize(256, method='crop')]
batch_tfms = [*aug_transforms(size=224), Normalize.from_stats(*imagenet_stats)]
dls = ImageDataLoaders.from_name_re(path, fnames, pat, bs=64, item_tfms=item_tfms,
... | _____no_output_____ | Apache-2.0 | nbs/74_callback.cutmix.ipynb | hanshin-back/fastai |
Next, let's initialize the callback `CutMix`, create a learner, do one batch and display the images with the labels. `CutMix` inside updates the loss function based on the ratio of the cutout bbox to the complete image. | cutmix = CutMix(alpha=1.)
with Learner(dls, resnet18(), loss_func=CrossEntropyLossFlat(), cbs=cutmix) as learn:
learn.epoch,learn.training = 0,True
learn.dl = dls.train
b = dls.one_batch()
learn._split(b)
learn('before_batch')
_,axs = plt.subplots(3,3, figsize=(9,9))
dls.show_batch(b=(cutmix.x,cutm... | _____no_output_____ | Apache-2.0 | nbs/74_callback.cutmix.ipynb | hanshin-back/fastai |
Using `CutMix` in Training | learn = cnn_learner(dls, resnet18, loss_func=CrossEntropyLossFlat(), cbs=cutmix, metrics=[accuracy, error_rate])
# learn.fit_one_cycle(1) | _____no_output_____ | Apache-2.0 | nbs/74_callback.cutmix.ipynb | hanshin-back/fastai |
Export - | #hide
from nbdev.export import notebook2script
notebook2script() | Converted 00_torch_core.ipynb.
Converted 01_layers.ipynb.
Converted 01a_losses.ipynb.
Converted 02_data.load.ipynb.
Converted 03_data.core.ipynb.
Converted 04_data.external.ipynb.
Converted 05_data.transforms.ipynb.
Converted 06_data.block.ipynb.
Converted 07_vision.core.ipynb.
Converted 08_vision.data.ipynb.
Converted... | Apache-2.0 | nbs/74_callback.cutmix.ipynb | hanshin-back/fastai |
Installing required packages | from IPython.display import clear_output
!pip install --upgrade pip
!pip install findspark
!pip install pyspark
clear_output(wait=False) | _____no_output_____ | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Importing global objects | import findspark, pyspark
from pyspark.sql import SparkSession
from pyspark import SparkFiles | _____no_output_____ | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Global SettingsNeeded for environments not Databricks | findspark.init()
spark = SparkSession.builder.getOrCreate() | _____no_output_____ | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Reading data source | url = 'https://raw.githubusercontent.com/edsonlourenco/public_datasets/main/Carros.csv'
spark.sparkContext.addFile(url)
csv_cars = SparkFiles.get("Carros.csv")
df_cars = spark.read.csv(csv_cars, header=True, inferSchema=True, sep=';') | _____no_output_____ | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Checking **data** | df_cars.orderBy('Consumo').show(truncate=False) | +-------+---------+-----------+---------------+----+-----+---------+-----------+-------+-----------+---+
|Consumo|Cilindros|Cilindradas|RelEixoTraseiro|Peso|Tempo|TipoMotor|Transmissao|Marchas|Carburadors|HP |
+-------+---------+-----------+---------------+----+-----+---------+-----------+-------+-----------+---+
|15 ... | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Transform VectorAssembler Importing **VectorAssembler** class | from pyspark.ml.feature import VectorAssembler | _____no_output_____ | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Doing transformation and creating features column | vectas = VectorAssembler(inputCols=[
"Consumo",
"Cilindros",
"Cilindradas",
"RelEixoTraseiro",
"Peso",
"... | +--------------------+
| features|
+--------------------+
|[15.0,8.0,301.0,3...|
|[21.0,6.0,160.0,3...|
|[21.0,6.0,160.0,3...|
|[26.0,4.0,1203.0,...|
|[104.0,8.0,472.0,...|
|[104.0,8.0,460.0,...|
|[133.0,8.0,350.0,...|
|[143.0,8.0,360.0,...|
|[147.0,8.0,440.0,...|
|[152.0,8.0,2758.0...|
|[152.0,8.0,304.0,...... | MIT | 3_VectorAssembler_example.ipynb | edsonlourenco/pyspark_ml_examples |
Export | #default_exp templ
from nbdev.export import notebook2script
notebook2script() | Converted om.ipynb.
Converted pspace.ipynb.
Converted templ.ipynb.
| Apache-2.0 | templ.ipynb | mirkoklukas/nbx |
Additional dependenciesYou will need to have tensorflow keras pydot and graphviz in your OS installed and added to the path ```bashpython -m pip install pydot``````bashyay graphviz ```bashsudo apt install python-pydot python-pydot-ng graphviz``` | import os
import sys
import time
import warnings
warnings.filterwarnings("ignore")
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import numpy as np
import pandas as pd
from pyspark.sql import SparkSession
packages = [
'JohnSnowLabs:spark-nlp: 2.4.2'
]
spark = SparkSession \
.builder \
.appName("ML SQL sessio... | _____no_output_____ | Apache-2.0 | tutorials/jupyter/8- Sarcasm Classifiers (GloVe and CNN).ipynb | nabinkhadka/spark-nlp-workshop |
Train your Unet with membrane datamembrane data is in folder membrane/, it is a binary classification task.The input shape of image and mask are the same :(batch_size,rows,cols,channel = 1) Train with data generator | data_gen_args = dict(rotation_range=0.2,
width_shift_range=0.05,
height_shift_range=0.05,
shear_range=0.05,
zoom_range=0.05,
horizontal_flip=True,
fill_mode='nearest')
myGene = trainGenerator(2,'data/... | _____no_output_____ | MIT | trainUnet.ipynb | chitrakumarsai/Semantic-segmentation---Unet- |
Train with npy file | #imgs_train,imgs_mask_train = geneTrainNpy("data/membrane/train/aug/","data/membrane/train/aug/")
#model.fit(imgs_train, imgs_mask_train, batch_size=2, nb_epoch=10, verbose=1,validation_split=0.2, shuffle=True, callbacks=[model_checkpoint]) | _____no_output_____ | MIT | trainUnet.ipynb | chitrakumarsai/Semantic-segmentation---Unet- |
test your model and save predicted results | testGene = testGenerator("data/membrane/test")
model = unet()
model.load_weights("unet_membrane.hdf5")
results = model.predict_generator(testGene,30,verbose=1)
saveResult("data/membrane/test",results) | C:\Users\xuhaozhi\Documents\Study\unet\model.py:34: UserWarning: The `merge` function is deprecated and will be removed after 08/2017. Use instead layers from `keras.layers.merge`, e.g. `add`, `concatenate`, etc.
merge6 = merge([drop4,up6], mode = 'concat', concat_axis = 3)
C:\SoftWare\Anaconda2\envs\python3\lib\site... | MIT | trainUnet.ipynb | chitrakumarsai/Semantic-segmentation---Unet- |
**Note: Please use the [pyEOF](https://pyeof.readthedocs.io/en/latest/installation.html) environment for this script** This script is used to implement EOF, REOF and k-means clustering to get regions | from pyEOF import *
import numpy as np
import pandas as pd
import xarray as xr
import matplotlib.pyplot as plt
import gc
import warnings
import pickle
from matplotlib.ticker import MaxNLocator
from tqdm import tqdm
from sklearn.cluster import KMeans
import cartopy.crs as ccrs
import cartopy.feature as cfeature
warning... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
plot the time series of the mean PM2.5 We decided to use April and August as the testing data | mask = xr.open_dataset(mask_path)
ds = sel_extent(xr.open_dataset(ds_path)).where(mask["mask"])
ds["PM25"].groupby('time.month').mean(dim=["lon","lat","time"]).plot()
plt.show() | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
get the data and implement EOFsBased on the results, we will select 4 PCs (n=4) and implemented varimax rotated EOFs (REOFs) | n=28
# remove months "4" (April) and "8" (August), to be consistant with training data
ds = ds.sel(time=ds.time.dt.month.isin([1,2,3,
5,6,7,
9,10,11,12]))
df = ds["PM25"].to_dataframe().reset_index() # get df from ds
# process the data fo... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
implement REOFs | n=4
# implement REOF
pca = df_eof(df_data,pca_type="varimax",n_components=n)
eofs = pca.eofs(s=2, n=n)
eofs_ds = eofs.stack(["lat","lon"], dropna=False).to_xarray()
pcs = pca.pcs(s=2, n=n)
evf = pca.evf(n=n)
df = pd.DataFrame({"n":np.arange(1,n+1),
"evf":pca.evf(n)*100.0,
"accum"... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
weighted EOFs loading | eofs_w = pd.DataFrame(data = (eofs.values * evf.reshape(n,1)),
index = eofs.index,
columns = eofs.columns)
eofs_w_ds = eofs_w.stack(["lat","lon"], dropna=False).to_xarray()
fig = plt.figure(figsize=(10,2))
for i in range(1,n+1):
ax = fig.add_subplot(1,4,i)
eofs_w_ds[... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
implement k-Means | # get the index which is not "nan"
placeholder_idx = np.argwhere(~np.isnan((eofs_w.values)[0])).reshape(-1)
# get the matrix without missing values: locations (row) * EOFs (columns)
m = eofs_w.values[:,placeholder_idx].transpose()
# clustering and calculate the Sum_of_squared_distances
Sum_of_squared_distances = []
K ... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
we select n_cluster = 6 to further the analysis | n_cluster = 6
fig = plt.figure(figsize=(8,3))
ax = fig.add_subplot(121)
ax.plot(K, Sum_of_squared_distances, 'bx-')
ax.plot(n_cluster,Sum_of_squared_distances[n_cluster-2],"rx")
ax.set_xlabel('number of clusters')
ax.set_ylabel('sum of squared distances')
# ax.set_title('Elbow method for optimal number of clusters')
... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
use n_cluster = 6 to implement the clusters | clusters = KMeans(n_clusters=n_cluster, random_state=66).fit_predict(m)
df_f = eofs.copy()
df_f.loc[str(n+1),:] = np.nan
df_f.iloc[n,placeholder_idx] = clusters
ds_f = df_f.stack(["lat","lon"], dropna=False).to_xarray()
fig = plt.figure(figsize=(3,3))
ax = fig.add_subplot(111, projection=ccrs.PlateCarree())
ax.set_ex... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
save the clusters masks | fig = plt.figure(figsize=(2*n_cluster,2))
for i in range(n_cluster):
ax = fig.add_subplot(1,n_cluster,i+1)
ds_f["mask_"+str(i)] = ds_regions.where(ds_regions==i).notnull().squeeze()
ds_f["mask_"+str(i)].plot(ax=ax,cbar_kwargs={"label":""})
ax.set_title("mask_"+str(i))
plt.tight_layout()
plt.show()
mask... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
save the regional masks | cluster_mask = xr.open_dataset("./data/cluster_mask_"+str(n_cluster)+".nc",engine="scipy")
fig = plt.figure(figsize=(2*n_cluster,2))
for i in range(n_cluster):
ax = fig.add_subplot(1,n_cluster,i+1)
ds.mean(dim="time").where(cluster_mask["mask_"+str(i)])["PM25"].plot(ax=ax,cbar_kwargs={"label":""})
ax.set_ti... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
save and load the regional mask | # save the regional mask
cluster_mask.to_netcdf("./data/r_mask.nc",engine="scipy")
# load the regional mask
test = xr.open_dataset("./data/r_mask.nc",engine="scipy")
fig = plt.figure(figsize=(n_cluster*2,2))
for i in range(len(loc_name)):
ax = fig.add_subplot(1,n_cluster,i+1)
ds.mean(dim="time").where(test[loc... | _____no_output_____ | MIT | 1_get_regions_eof_reofs.ipynb | zzheng93/code_DSI_India_AutoML |
 Link Prediction - IntroductionIn this Notebook we are going to examine the process of using Amazon Neptune ML feature to perform link prediction in a property graph. Note: This notebook take approximately 1 hour to complete[Neptune ML](https://docs.aws.amazon.com/neptune/latest/user... | import neptune_ml_utils as neptune_ml
neptune_ml.check_ml_enabled() | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
If the check above did not say that this cluster is ready to run Neptune ML jobs then please check that the cluster meets all the pre-requisites defined [here](https://docs.aws.amazon.com/neptune/latest/userguide/machine-learning.htmlmachine-learning-overview). Load the dataThe first step in building a Neptune ML model... | s3_bucket_uri="s3://<INSERT S3 BUCKET OR PATH>"
# remove trailing slashes
s3_bucket_uri = s3_bucket_uri[:-1] if s3_bucket_uri.endswith('/') else s3_bucket_uri | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
Now that you have provided an S3 bucket, run the cell below which will download and format the MovieLens data into a format compatible with Neptune's bulk loader. | response = neptune_ml.prepare_movielens_data(s3_bucket_uri) | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
This process only takes a few minutes and once it has completed you can load the data using the `%load` command in the cell below. | %load -s {response} -f csv -p OVERSUBSCRIBE --run | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
Check to make sure the data is loadedOnce the cell has completed, the data has been loaded into the cluster. We verify the data loaded correctly by running the traversals below to see the count of nodes by label: Note: The numbers below assume no other data is in the cluster | %%gremlin
g.V().groupCount().by(label).unfold() | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
If our nodes loaded correctly then the output is:* 19 genres* 1682 movies* 100000 rating* 943 usersTo check that our edges loaded correctly we check the edge counts: | %%gremlin
g.E().groupCount().by(label).unfold() | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
If our edges loaded correctly then the output is:* 100000 about* 2893 included_in* 100000 rated* 100000 wrote Preparing for exportWith our data validated let's remove a few `rated` vertices so that we can build a model that predicts these missing connections. In a normal scenario, the data you would like to predict is... | %%gremlin
g.V('user_1').outE('rated') | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
Now let's remove these edges to simulate them missing from our data. | %%gremlin
g.V('user_1').outE('rated').drop() | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
Checking our data again we see that the edges have now been removed. | %%gremlin
g.V('user_1').outE('rated') | _____no_output_____ | ISC | src/graph_notebook/notebooks/04-Machine-Learning/Neptune-ML-03-Introduction-to-Link-Prediction-Gremlin.ipynb | zacharyrs/graph-notebook |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.