paper_id
stringlengths
9
13
title
stringlengths
15
153
paper_url
stringlengths
41
45
authors
listlengths
1
21
type
stringclasses
3 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
310
2.43k
keywords
listlengths
1
32
TL;DR
large_stringlengths
23
250
submission_number
int64
1
11.7k
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
d7skOEQClK
Handling Long-tailed Feature Distribution in AdderNets
https://openreview.net/forum?id=d7skOEQClK
[ "Minjing Dong", "Yunhe Wang", "Xinghao Chen", "Chang Xu" ]
Poster
null
Adder neural networks (ANNs) are designed for low energy cost which replace expensive multiplications in convolutional neural networks (CNNs) with cheaper additions to yield energy-efficient neural networks and hardware accelerations. Although ANNs achieve satisfactory efficiency, there exist gaps between ANNs and CNNs...
[ "AdderNet", "Efficient Networks", "Multivariate Skew Laplace" ]
Alleviate gaps between ANNs and CNNs through formulating feature distribution of ANNs.
3,801
null
null
yNzF41lHYV
Believe What You See: Implicit Constraint Approach for Offline Multi-Agent Reinforcement Learning
https://openreview.net/forum?id=yNzF41lHYV
[ "Yiqin Yang", "Xiaoteng Ma", "Chenghao Li", "Zewu Zheng", "Qiyuan Zhang", "Gao Huang", "Jun Yang", "Qianchuan Zhao" ]
Spotlight
null
Learning from datasets without interaction with environments (Offline Learning) is an essential step to apply Reinforcement Learning (RL) algorithms in real-world scenarios. However, compared with the single-agent counterpart, offline multi-agent RL introduces more agents with the larger state and action space, which i...
[ "Extrapolation Error", "Offline Reinforcement Learning", "Multi-Agent Reinforcement Learning" ]
The first study analyzing and addressing the extrapolation error in multi-agent reinforcement learning.
444
2106.03400
title_snapshot
xN3XX6pKSD5
Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis
https://openreview.net/forum?id=xN3XX6pKSD5
[ "Tianchang Shen", "Jun Gao", "Kangxue Yin", "Ming-Yu Liu", "Sanja Fidler" ]
Poster
null
We introduce DMTet, a deep 3D conditional generative model that can synthesize high-resolution 3D shapes using simple user guides such as coarse voxels. It marries the merits of implicit and explicit 3D representations by leveraging a novel hybrid 3D representation. Compared to the current implicit approaches, which ar...
[ "3D Deep Learning", "3D Super-Resolution", "3D Content Creation", "3D Shape Representation" ]
null
3,658
2111.04276
title_snapshot
hx2Ckkzdf53
Stochastic Anderson Mixing for Nonconvex Stochastic Optimization
https://openreview.net/forum?id=hx2Ckkzdf53
[ "Fuchao Wei", "Chenglong Bao", "Yang Liu" ]
Poster
null
Anderson mixing (AM) is an acceleration method for fixed-point iterations. Despite its success and wide usage in scientific computing, the convergence theory of AM remains unclear, and its applications to machine learning problems are not well explored. In this paper, by introducing damped projection and adaptive regul...
[ "Anderson mixing", "nonconvex stochastic optimization", "neural network" ]
We propose a stochastic version of Anderson mixing which has theoretical guarantees and shows promising results in training neural networks.
3,651
2110.01543
title_snapshot
r1pprsDm185
SBO-RNN: Reformulating Recurrent Neural Networks via Stochastic Bilevel Optimization
https://openreview.net/forum?id=r1pprsDm185
[ "Ziming Zhang", "Yun Yue", "Guojun Wu", "Yanhua Li", "Haichong Zhang" ]
Poster
null
In this paper we consider the training stability of recurrent neural networks (RNNs) and propose a family of RNNs, namely SBO-RNN, that can be formulated using stochastic bilevel optimization (SBO). With the help of stochastic gradient descent (SGD), we manage to convert the SBO problem into an RNN where the feedforwar...
[ "recurrent neural networks", "stochastic bilevel optimization", "training stability" ]
SBO-RNN: Reformulating Recurrent Neural Networks via Stochastic Bilevel Optimization
3,612
null
null
Q4SdMvWMxb
Slow Learning and Fast Inference: Efficient Graph Similarity Computation via Knowledge Distillation
https://openreview.net/forum?id=Q4SdMvWMxb
[ "Can Qin", "Handong Zhao", "Lichen Wang", "Huan Wang", "Yulun Zhang", "Yun Fu" ]
Poster
null
Graph Similarity Computation (GSC) is essential to wide-ranging graph applications such as retrieval, plagiarism/anomaly detection, etc. The exact computation of graph similarity, e.g., Graph Edit Distance (GED), is an NP-hard problem that cannot be exactly solved within an adequate time given large graphs. Thanks to t...
[ "Graph Similarity Computation", "Efficient Model", "Knowledge Distillation" ]
A novel efficient graph similarity computation framework based on knowledge distillation.
429
null
null
44EMx-dkQU
A Gaussian Process-Bayesian Bernoulli Mixture Model for Multi-Label Active Learning
https://openreview.net/forum?id=44EMx-dkQU
[ "Weishi Shi", "Dayou Yu", "Qi Yu" ]
Poster
null
Multi-label classification (MLC) allows complex dependencies among labels, making it more suitable to model many real-world problems. However, data annotation for training MLC models becomes much more labor-intensive due to the correlated (hence non-exclusive) labels and a potential large and sparse label space. We pr...
[ "multi-label active learning", "Bayesian Bernoulli mixture", "label correlation" ]
We propose a novel integrated Gaussian process-Bayesian Bernoulli mixture model to accurately quantify a data sample's overall contribution to a correlated label space for cost-effective multi-label active learning.
3,599
null
null
j7YA-y0P3-
Online Adaptation to Label Distribution Shift
https://openreview.net/forum?id=j7YA-y0P3-
[ "Ruihan Wu", "Chuan Guo", "Yi Su", "Kilian Q Weinberger" ]
Poster
null
Machine learning models often encounter distribution shifts when deployed in the real world. In this paper, we focus on adaptation to label distribution shift in the online setting, where the test-time label distribution is continually changing and the model must dynamically adapt to it without observing the true labe...
[ "Damain Adaptation", "Label Shift", "Online Learning" ]
We study the label distribution shifts problem in the online setting and analyze our methods from both theoretical and empirical sides.
2,473
2107.04520
title_snapshot
Goz-qsH1F14
Adaptive Machine Unlearning
https://openreview.net/forum?id=Goz-qsH1F14
[ "Varun Gupta", "Christopher Jung", "Seth Neel", "Aaron Roth", "Saeed Sharifi -Malvajerdi", "Christopher Waites" ]
Poster
null
Data deletion algorithms aim to remove the influence of deleted data points from trained models at a cheaper computational cost than fully retraining those models. However, for sequences of deletions, most prior work in the non-convex setting gives valid guarantees only for sequences that are chosen independently of th...
[ "Machine Unlearning", "Data Deletion", "Differential Privacy", "Max Information" ]
We give the first data deletion algorithms that can be used for arbitrary nonconvex models without pretraining, and give guarantees for sequences of deletions that can depend on information we have previously published.
3,503
2106.04378
title_snapshot
qZpOqPbwhy
Learning Riemannian metric for disease progression modeling
https://openreview.net/forum?id=qZpOqPbwhy
[ "Samuel Gruffaz", "Pierre-Emmanuel Poulet", "Etienne Maheux", "Bruno Michel Jedynak", "Stanley Durrleman" ]
Poster
null
Linear mixed-effect models provide a natural baseline for estimating disease progression using longitudinal data. They provide interpretable models at the cost of modeling assumptions on the progression profiles and their variability across subjects. A significant improvement is to embed the data in a Riemannian manif...
[ "Riemannian Geometry", "RKHS", "mixed-effect model", "Disease progression modelling", "Longitudinal data" ]
We propose a method to learn a Riemannian metric in the observation space to estimate disease trajectories from patient data. It allows to build interpretable disease progression models with higher predictive power than state-of-the-art.
3,488
null
null
0hJ-U3aqUDf
Deep Jump Learning for Off-Policy Evaluation in Continuous Treatment Settings
https://openreview.net/forum?id=0hJ-U3aqUDf
[ "Hengrui Cai", "Chengchun Shi", "Rui Song", "Wenbin Lu" ]
Poster
null
We consider off-policy evaluation (OPE) in continuous treatment settings, such as personalized dose-finding. In OPE, one aims to estimate the mean outcome under a new treatment decision rule using historical data generated by a different decision rule. Most existing works on OPE focus on discrete treatment settings. To...
[ "Statistical Learning", "Off-Policy Evaluation", "Deep Learning", "Continuous Treatments", "Multi-Scale Change Point Detection", "Precision medicine" ]
Develop deep jump learning for off-policy evaluation in continuous treatment settings, by adaptively discretizing the treatment space.
401
2010.15963
title_snapshot
WWRBHhH158K
Deep Learning Through the Lens of Example Difficulty
https://openreview.net/forum?id=WWRBHhH158K
[ "Robert John Nicholas Baldock", "Hartmut Maennel", "Behnam Neyshabur" ]
Poster
null
Existing work on understanding deep learning often employs measures that compress all data-dependent information into a few numbers. In this work, we adopt a perspective based on the role of individual examples. We introduce a measure of the computational difficulty of making a prediction for a given input: the (effect...
[ "Deep Learning", "Example Difficulty", "Prediction Depth" ]
We study the number of hidden layers used inside a deep model to determine the output for each given input, and find that this quantity connects several aspects of deep learning.
391
2106.09647
title_snapshot
-KGLlWv6kIc
How Fine-Tuning Allows for Effective Meta-Learning
https://openreview.net/forum?id=-KGLlWv6kIc
[ "Kurtland Chua", "Qi Lei", "Jason D. Lee" ]
Poster
null
Representation learning has served as a key tool for meta-learning, enabling rapid learning of new tasks. Recent works like MAML learn task-specific representations by finding an initial representation requiring minimal per-task adaptation (i.e. a fine-tuning-based objective). We present a theoretical framework for ana...
[ "Meta-learning", "Few-Shot Learning", "Rademacher Complexity", "Statistical Learning Theory" ]
We provide a statistical analysis of fine-tuning-based meta-learning, and establish a sample complexity gap from a standard baseline.
3,419
2105.02221
title_snapshot
HbTzvugzOp
Chebyshev-Cantelli PAC-Bayes-Bennett Inequality for the Weighted Majority Vote
https://openreview.net/forum?id=HbTzvugzOp
[ "Yi-Shan Wu", "Andres R Masegosa", "Stephan Sloth Lorenzen", "Christian Igel", "Yevgeny Seldin" ]
Poster
null
We present a new second-order oracle bound for the expected risk of a weighted majority vote. The bound is based on a novel parametric form of the Chebyshev-Cantelli inequality (a.k.a. one-sided Chebyshev’s), which is amenable to efficient minimization. The new form resolves the optimization challenge faced by prior or...
[ "Weighted Majority Vote", "PAC-Bayes", "Learning Theory" ]
null
3,310
2106.13624
title_snapshot
eW8HEhY9F7C
Nearly-Tight and Oblivious Algorithms for Explainable Clustering
https://openreview.net/forum?id=eW8HEhY9F7C
[ "Buddhima Gamlath", "Xinrui Jia", "Adam Polak", "Ola Svensson" ]
Poster
null
We study the problem of explainable clustering in the setting first formalized by Dasgupta, Frost, Moshkovitz, and Rashtchian (ICML 2020). A $k$-clustering is said to be explainable if it is given by a decision tree where each internal node splits data points with a threshold cut in a single dimension (feature), and ea...
[ "algorithms", "clustering", "combinatorial optimization", "explainability" ]
We give near-tight algorithms for explainable k-medians and k-means clustering, improving the solution quality by polynomial factors compared to prior work.
3,303
2106.16147
title_snapshot
0wqGMmfqBw
SAPE: Spatially-Adaptive Progressive Encoding for Neural Optimization
https://openreview.net/forum?id=0wqGMmfqBw
[ "Amir Hertz", "Or Perel", "Raja Giryes", "Olga Sorkine-hornung", "Daniel Cohen-Or" ]
Poster
null
Multilayer-perceptrons (MLP) are known to struggle learning functions of high-frequencies, and in particular, instances of wide frequency bands. We present a progressive mapping scheme for input signals of MLP networks, enabling them to better fit a wide range of frequencies without sacrificing training stability or r...
[ "Machine Learning", "Deep Learning", "Computer Vision", "Implicit functions", "Positional Encoding" ]
We introduce Spatially Adaptive Progressive Encoding (SAPE) layers, which gradually unmask signal components in a NN as a function of time and space.
2,424
2104.09125
title_snapshot
p9dySshcS0q
STEP: Out-of-Distribution Detection in the Presence of Limited In-Distribution Labeled Data
https://openreview.net/forum?id=p9dySshcS0q
[ "Zhi Zhou", "Lan-Zhe Guo", "Zhanzhan Cheng", "Yu-Feng Li", "Shiliang Pu" ]
Poster
null
Existing semi-supervised learning (SSL) studies typically assume that unlabeled and test data are drawn from the same distribution as labeled data. However, in many real-world applications, it is desirable to have SSL algorithms that not only classify the samples drawn from the same distribution of labeled data but als...
[ "weakly supervised learning", "semi-supervised learning", "out-of-distribution detection" ]
We utilize the self-supervised representations and local topological structures to enhance the out-of-distribution detection ability of the model.
3,265
null
null
paxcakYWwIu
Simple Stochastic and Online Gradient Descent Algorithms for Pairwise Learning
https://openreview.net/forum?id=paxcakYWwIu
[ "Zhenhuan Yang", "Yunwen Lei", "Puyu Wang", "Tianbao Yang", "Yiming Ying" ]
Poster
null
Pairwise learning refers to learning tasks where the loss function depends on a pair of instances. It instantiates many important machine learning tasks such as bipartite ranking and metric learning. A popular approach to handle streaming data in pairwise learning is an online gradient descent (OGD) algorithm, where ...
[ "Pairwise Learning", "Online (Stochastic) Gradient Descent", "Stability and Generalization", "Differential Privacy" ]
Simple Stochastic and Online Gradient Descent Algorithms for Pairwise Learning
3,259
2111.12050
title_snapshot
Ltu7TOYVh_
Understanding Bandits with Graph Feedback
https://openreview.net/forum?id=Ltu7TOYVh_
[ "Houshuang Chen", "Zengfeng Huang", "Shuai Li", "Chihao Zhang" ]
Poster
null
The bandit problem with graph feedback, proposed in [Mannor and Shamir, NeurIPS 2011], is modeled by a directed graph $G=(V,E)$ where $V$ is the collection of bandit arms, and once an arm is triggered, all its incident arms are observed. A fundamental question is how the structure of the graph affects the min-max regre...
[ "Bandits", "Graph feedback", "Online Learning" ]
Improved upper and lower bounds for graph feedback bandits.
368
2105.14260
title_snapshot
ehzq1YQrucI
An Infinite-Feature Extension for Bayesian ReLU Nets That Fixes Their Asymptotic Overconfidence
https://openreview.net/forum?id=ehzq1YQrucI
[ "Agustinus Kristiadi", "Matthias Hein", "Philipp Hennig" ]
Spotlight
null
A Bayesian treatment can mitigate overconfidence in ReLU nets around the training data. But far away from them, ReLU Bayesian neural networks (BNNs) can still underestimate uncertainty and thus be asymptotically overconfident. This issue arises since the output variance of a BNN with finitely many features is quadratic...
[ "bayesian deep learning", "uncertainty quantification", "overconfidence", "gaussian processes" ]
We propose a post-hoc extension, constructed by considering infinitely many ReLU features, for standard ReLU BNNs. This extension effectively fixes ReLU BNNs' overconfidence far away from the training data.
365
2010.02709
title_snapshot
hsqZ5v8PFyQ
Emergent Discrete Communication in Semantic Spaces
https://openreview.net/forum?id=hsqZ5v8PFyQ
[ "Mycal Tucker", "Huao Li", "Siddharth Agrawal", "Dana Hughes", "Katia P. Sycara", "Michael Lewis", "Julie Shah" ]
Poster
null
Neural agents trained in reinforcement learning settings can learn to communicate among themselves via discrete tokens, accomplishing as a team what agents would be unable to do alone. However, the current standard of using one-hot vectors as discrete communication tokens prevents agents from acquiring more desirable a...
[ "Emergent communication", "zero-shot understanding", "reference games" ]
Instead of using one-hot tokens in emergent discrete communication, we enable agents to learn tokens within a semantic space, enabling greater robustness and zero-shot understanding of new communication
3,220
2108.01828
title_snapshot
2w_2PwOYJarMu
Average-Reward Learning and Planning with Options
https://openreview.net/forum?id=2w_2PwOYJarMu
[ "Yi Wan", "Abhishek Naik", "Richard S. Sutton" ]
Poster
null
We extend the options framework for temporal abstraction in reinforcement learning from discounted Markov decision processes (MDPs) to average-reward MDPs. Our contributions include general convergent off-policy inter-option learning algorithms, intra-option algorithms for learning values and models, as well as sample-...
[ "average reward", "options", "reinforcement learning" ]
This paper extends learning and planning algorithms within the options framework (Sutton et al. 1999) from discounted MDPs to average-reward MDPs.
359
2110.13855
title_snapshot
xkQ4MhLv52X
Learning-Augmented Dynamic Power Management with Multiple States via New Ski Rental Bounds
https://openreview.net/forum?id=xkQ4MhLv52X
[ "Antonios Antoniadis", "Christian Coester", "Marek Elias", "Adam Polak", "Bertrand Simon" ]
Poster
null
We study the online problem of minimizing power consumption in systems with multiple power-saving states. During idle periods of unknown lengths, an algorithm has to choose between power-saving states of different energy consumption and wake-up costs. We develop a learning-augmented online algorithm that makes decision...
[ "Learning-augmented algorithms", "online algorithms", "energy-efficient algorithms", "power management", "ski rental" ]
We design a learning-augmented algorithm for dynamic power management with multiple power-saving states, based on new & tight bounds for learning-augmented ski rental.
3,135
2110.13116
title_snapshot
U7vVeHydyR
Fast Extra Gradient Methods for Smooth Structured Nonconvex-Nonconcave Minimax Problems
https://openreview.net/forum?id=U7vVeHydyR
[ "Sucheol Lee", "Donghwan Kim" ]
Poster
null
Modern minimax problems, such as generative adversarial network and adversarial training, are often under a nonconvex-nonconcave setting, and developing an efficient method for such setting is of interest. Recently, two variants of the extragradient (EG) method are studied in that direction. First, a two-time-scale var...
[ "Minimax problems", "Nonconvex-nonconcave Problems", "Extragradient Method", "Acceleration" ]
We propose a fast extragradient method, with an accelerated O(1/k^2) rate, for smooth structured nonconvex-nonconcave problems.
3,072
2106.02326
title_snapshot
O8uSRrmTeSQ
Continuous Doubly Constrained Batch Reinforcement Learning
https://openreview.net/forum?id=O8uSRrmTeSQ
[ "Rasool Fakoor", "Jonas Mueller", "Kavosh Asadi", "Pratik Chaudhari", "Alex Smola" ]
Poster
null
Reliant on too many experiments to learn good actions, current Reinforcement Learning (RL) algorithms have limited applicability in real-world settings, which can be too expensive to allow exploration. We propose an algorithm for batch RL, where effective policies are learned using only a fixed offline dataset instead ...
[ "batch reinforcement learning", "overestimation bias", "extrapolation", "offline reinforcement learning", "batch rl", "offline rl" ]
CDC is an effective batch RL algorithm that addresses overestimation bias and extrapolation errors by a simple pair of regularizers.
3,048
2102.09225
title_snapshot
kFJoj7zuDVi
Towards a Theoretical Framework of Out-of-Distribution Generalization
https://openreview.net/forum?id=kFJoj7zuDVi
[ "Haotian Ye", "Chuanlong Xie", "Tianle Cai", "Ruichen Li", "Zhenguo Li", "Liwei Wang" ]
Poster
null
Generalization to out-of-distribution (OOD) data is one of the central problems in modern machine learning. Recently, there is a surge of attempts to propose algorithms that mainly build upon the idea of extracting invariant features. Although intuitively reasonable, theoretical understanding of what kind of invariance...
[ "OOD Generalization", "Theory", "Model Selection" ]
null
2,991
2106.04496
title_snapshot
Z9Kpr38Kx_
Model-Based Episodic Memory Induces Dynamic Hybrid Controls
https://openreview.net/forum?id=Z9Kpr38Kx_
[ "Hung Le", "Thommen Karimpanal George", "Majid Abdolshah", "Truyen Tran", "Svetha Venkatesh" ]
Poster
null
Episodic control enables sample efficiency in reinforcement learning by recalling past experiences from an episodic memory. We propose a new model-based episodic memory of trajectories addressing current limitations of episodic control. Our memory estimates trajectory values, guiding the agent towards good policies. Bu...
[ "Reinforcement learning", "episodic memory", "sample-efficiency" ]
A model-based episodic control facilitating complementary learning systems
330
2111.02104
title_snapshot
0BHU7WvZ29
Adaptive Denoising via GainTuning
https://openreview.net/forum?id=0BHU7WvZ29
[ "Sreyas Mohan", "Joshua L. Vincent", "Ramon Manzorro", "Peter Crozier", "Carlos Fernandez-Granda", "Eero P Simoncelli" ]
Poster
null
Deep convolutional neural networks (CNNs) for image denoising are typically trained on large datasets. These models achieve the current state of the art, but they do not generalize well to data that deviate from the training distribution. Recent work has shown that it is possible to train denoisers on a single noisy im...
[ "denoising", "adaptation", "out of distribution", "generalization", "gain" ]
We propose a novel framework for adapting denoising neural networks at test time by adjusting a small number of parameters using an unsupervised loss, which substantially improves the generalization abilities of these models.
2,907
2107.12815
title_snapshot
QwNLVId9Df
Program Synthesis Guided Reinforcement Learning for Partially Observed Environments
https://openreview.net/forum?id=QwNLVId9Df
[ "Yichen Yang", "Jeevana Priya Inala", "Osbert Bastani", "Yewen Pu", "Armando Solar-Lezama", "Martin Rinard" ]
Spotlight
null
A key challenge for reinforcement learning is solving long-horizon planning problems. Recent work has leveraged programs to guide reinforcement learning in these settings. However, these approaches impose a high manual burden on the user since they must provide a guiding program for every new task. Partially observed e...
[ "program synthesis", "reinforcement learning", "partial observation" ]
null
2,364
2102.11137
title_snapshot
gRqHB07GGz3
DIB-R++: Learning to Predict Lighting and Material with a Hybrid Differentiable Renderer
https://openreview.net/forum?id=gRqHB07GGz3
[ "Wenzheng Chen", "Joey Litalien", "Jun Gao", "Zian Wang", "Clement Fuji Tsang", "Sameh Khamis", "Or Litany", "Sanja Fidler" ]
Poster
null
We consider the challenging problem of predicting intrinsic object properties from a single image by exploiting differentiable renderers. Many previous learning-based approaches for inverse graphics adopt rasterization-based renderers and assume naive lighting and material models, which often fail to account for non-La...
[ "differentiable rendering", "inverse graphics", "material estimation", "lighting estimation" ]
A hybrid differentiable renderer that supports advanced lighting and material effects and can be embedded in deep learning to jointly predict geometry, light and material from a single image with 2D image supervision only.
2,806
2111.00140
title_snapshot
8fztRILSxL
Machine versus Human Attention in Deep Reinforcement Learning Tasks
https://openreview.net/forum?id=8fztRILSxL
[ "Sihang Guo", "Ruohan Zhang", "Bo Liu", "Yifeng Zhu", "Dana Ballard", "Mary Hayhoe", "Peter Stone" ]
Poster
null
Deep reinforcement learning (RL) algorithms are powerful tools for solving visuomotor decision tasks. However, the trained models are often difficult to interpret, because they are represented as end-to-end deep neural networks. In this paper, we shed light on the inner workings of such trained models by analyzing the...
[ "Deep Reinforcement Learning", "Interpretability", "Attention", "Eye Tracking" ]
We compare deep reinforcement learning agent's attention map with human gaze data in Atari games in order to understand these RL agents better.
2,757
2010.15942
title_snapshot
cqD6imhK1TX
Conditional Generation Using Polynomial Expansions
https://openreview.net/forum?id=cqD6imhK1TX
[ "Grigorios Chrysos", "Markos Georgopoulos", "Yannis Panagakis" ]
Poster
null
Generative modeling has evolved to a notable field of machine learning. Deep polynomial neural networks (PNNs) have demonstrated impressive results in unsupervised image generation, where the task is to map an input vector (i.e., noise) to a synthesized image. However, the success of PNNs has not been replicated in con...
[ "polynomial expansions", "polynomial neural networks", "GAN", "VAE", "unseen combinations", "conditional data generation" ]
We condition the generator of a genertive model using a polynomial expansion
299
null
null
BEVDmheFG0
Generalized Linear Bandits with Local Differential Privacy
https://openreview.net/forum?id=BEVDmheFG0
[ "Yuxuan Han", "Zhipeng Liang", "Yang Wang", "Jiheng Zhang" ]
Poster
null
Contextual bandit algorithms are useful in personalized online decision-making. However, many applications such as personalized medicine and online advertising require the utilization of individual-specific information for effective learning, while user's data should remain private from the server due to privacy concer...
[ "Local Differential Privacy", "Contextual Bandits", "Online Decision Making" ]
null
2,701
2106.03365
title_snapshot
ba27-RzNaIv
Scalars are universal: Equivariant machine learning, structured like classical physics
https://openreview.net/forum?id=ba27-RzNaIv
[ "Soledad Villar", "David W Hogg", "Kate Storey-Fisher", "Weichi Yao", "Ben Blum-Smith" ]
Poster
null
There has been enormous progress in the last few years in designing neural networks that respect the fundamental symmetries and coordinate freedoms of physical law. Some of these frameworks make use of irreducible representations, some make use of high-order tensor objects, and some apply symmetry-enforcing constraint...
[ "invariant theory", "group theory", "physics", "regression", "machine learning", "symmetry", "scalars" ]
Simple and efficient parameterization of all group invariant and equivariant functions for physically relevant groups.
2,700
2106.06610
title_snapshot
vrXuRmaU_jM
Weak-shot Fine-grained Classification via Similarity Transfer
https://openreview.net/forum?id=vrXuRmaU_jM
[ "Junjie Chen", "Li Niu", "Liu Liu", "Liqing Zhang" ]
Poster
null
Recognizing fine-grained categories remains a challenging task, due to the subtle distinctions among different subordinate categories, which results in the need of abundant annotated samples. To alleviate the data-hungry problem, we consider the problem of learning novel categories from web data with the support of a c...
[ "transfer learning", "similarity transfer", "weak-shot learning" ]
null
2,693
2009.09197
title_snapshot
0vaPiltED1N
Sequence-to-Sequence Learning with Latent Neural Grammars
https://openreview.net/forum?id=0vaPiltED1N
[ "Yoon Kim" ]
Spotlight
null
Sequence-to-sequence learning with neural networks has become the de facto standard for sequence modeling. This approach typically models the local distribution over the next element with a powerful neural network that can condition on arbitrary context. While flexible and performant, these models often require large d...
[ "sequence-to-sequence learning", "compositional generalization", "synchronous grammars" ]
null
2,630
2109.01135
title_snapshot
7_M2f2DEIEK
Beyond Tikhonov: faster learning with self-concordant losses, via iterative regularization
https://openreview.net/forum?id=7_M2f2DEIEK
[ "Gaspard Beugnot", "Julien Mairal", "Alessandro Rudi" ]
Spotlight
null
The theory of spectral filtering is a remarkable tool to understand the statistical properties of learning with kernels. For least squares, it allows to derive various regularization schemes that yield faster convergence rates of the excess risk than with Tikhonov regularization. This is typically achieved by leveragin...
[ "Kernel methods", "learning theory", "self concordance", "iterated tikhonov", "proximal point" ]
The iterative Thikonov regularization scheme achieves optimal sample complexity on self concordant losses.
267
2106.08855
title_snapshot
AqprMSXI1Wn
SSMF: Shifting Seasonal Matrix Factorization
https://openreview.net/forum?id=AqprMSXI1Wn
[ "Koki Kawabata", "Siddharth Bhatia", "Rui Liu", "Mohit Wadhwa", "Bryan Hooi" ]
Poster
null
Given taxi-ride counts information between departure and destination locations, how can we forecast their future demands? In general, given a data stream of events with seasonal patterns that innovate over time, how can we effectively and efficiently forecast future events? In this paper, we propose Shifting Seasonal M...
[ "Matrix Factorization", "Seasonality", "Regime Switching", "Online Algorithm" ]
null
2,430
2110.12763
title_snapshot
wKf9iSu_TEm
Multi-Objective Meta Learning
https://openreview.net/forum?id=wKf9iSu_TEm
[ "Feiyang Ye", "Baijiong Lin", "Zhixiong Yue", "Pengxin Guo", "Qiao Xiao", "Yu Zhang" ]
Poster
null
Meta learning with multiple objectives has been attracted much attention recently since many applications need to consider multiple factors when designing learning models. Existing gradient-based works on meta learning with multiple objectives mainly combine multiple objectives into a single objective in a weighted sum...
[ "Meta Learning", "Multi-objective Optimization", "Bi-level Optimization" ]
null
2,347
2102.07121
title_snapshot
rJq1SdaNPX4
Disentangled Contrastive Learning on Graphs
https://openreview.net/forum?id=rJq1SdaNPX4
[ "Haoyang Li", "Xin Wang", "Ziwei Zhang", "Zehuan Yuan", "Hang Li", "Wenwu Zhu" ]
Poster
null
Recently, self-supervised learning for graph neural networks (GNNs) has attracted considerable attention because of their notable successes in learning the representation of graph-structure data. However, the formation of a real-world graph typically arises from the highly complex interaction of many latent factors. Th...
[ "Graph Neural Network", "Contrastive Learning", "Self-supervised Learning", "Disentangled Representation Learning" ]
We introduce the Disentangled Graph Contrastive Learning (DGCL) method, which is able to learn disentangled graph-level representations with self-supervision.
2,338
null
null
0lzmTb4LGd3F
VoiceMixer: Adversarial Voice Style Mixup
https://openreview.net/forum?id=0lzmTb4LGd3F
[ "Sang-Hoon Lee", "Ji-Hoon Kim", "Hyunseung Chung", "Seong-Whan Lee" ]
Poster
null
Although recent advances in voice conversion have shown significant improvement, there still remains a gap between the converted voice and target voice. A key factor that maintains this gap is the insufficient decomposition of content and voice style from the source speech. This insufficiency leads to the converted spe...
[ "Voice Conversion", "Style Transfer", "Speech Processing" ]
Voice conversion through similarity-based information bottleneck and adversarial feedback using self-supervised representation learning
245
null
null
kSR-_SVzDR-
Gaussian Kernel Mixture Network for Single Image Defocus Deblurring
https://openreview.net/forum?id=kSR-_SVzDR-
[ "Yuhui Quan", "Zicong Wu", "Hui Ji" ]
Poster
null
Defocus blur is one kind of blur effects often seen in images, which is challenging to remove due to its spatially variant amount. This paper presents an end-to-end deep learning approach for removing defocus blur from a single image, so as to have an all-in-focus image for consequent vision tasks. First, a pixel-wis...
[ "Defocus Deblurring", "Deep Learning", "Image Recovery", "Unrolling", "Attention" ]
A lightweight DNN for single image defocus deblurring with SOTA performance
2,285
2111.00454
title_snapshot
OItvP2-i9j
Closing the Gap: Tighter Analysis of Alternating Stochastic Gradient Methods for Bilevel Problems
https://openreview.net/forum?id=OItvP2-i9j
[ "Tianyi Chen", "Yuejiao Sun", "Wotao Yin" ]
Spotlight
null
Stochastic nested optimization, including stochastic compositional, min-max, and bilevel optimization, is gaining popularity in many machine learning applications. While the three problems share a nested structure, existing works often treat them separately, thus developing problem-specific algorithms and analyses. A...
[ "Stochastic bilevel optimization", "min-max optimization", "compositional optimization", "convergence analysis" ]
Our results explain why simple SGD-type algorithms all work very well in practical bilevel problems without the need for further modifications.
2,240
null
null
pBKOx_dxYAN
SNIPS: Solving Noisy Inverse Problems Stochastically
https://openreview.net/forum?id=pBKOx_dxYAN
[ "Bahjat Kawar", "Gregory Vaksman", "Michael Elad" ]
Poster
null
In this work we introduce a novel stochastic algorithm dubbed SNIPS, which draws samples from the posterior distribution of any linear inverse problem, where the observation is assumed to be contaminated by additive white Gaussian noise. Our solution incorporates ideas from Langevin dynamics and Newton's method, and ex...
[ "Inverse Problems", "Image Restoration", "Denoising", "Langevin Dynamics", "Diffusion", "Super Resolution", "Deblurring", "Compressive Sensing" ]
null
2,217
2105.14951
title_snapshot
GvU4RvMwlGo
Coresets for Decision Trees of Signals
https://openreview.net/forum?id=GvU4RvMwlGo
[ "Ibrahim Jubran", "Ernesto Evgeniy Sanches Shayda", "Ilan Newman", "Dan Feldman" ]
Spotlight
null
A $k$-decision tree $t$ (or $k$-tree) is a recursive partition of a matrix (2D-signal) into $k\geq 1$ block matrices (axis-parallel rectangles, leaves) where each rectangle is assigned a real label. Its regression or classification loss to a given matrix $D$ of $N$ entries (labels) is the sum of squared differences ove...
[ "Coresets", "Machine Learning", "Decision Trees", "Random Forests" ]
Coresets for Decision Trees of Signals
2,206
2110.03195
title_snapshot
Ba3odanehCw
Regret Minimization Experience Replay in Off-Policy Reinforcement Learning
https://openreview.net/forum?id=Ba3odanehCw
[ "Xu-Hui Liu", "Zhenghai Xue", "Jing-Cheng Pang", "Shengyi Jiang", "Feng Xu", "Yang Yu" ]
Poster
null
In reinforcement learning, experience replay stores past samples for further reuse. Prioritized sampling is a promising technique to better utilize these samples. Previous criteria of prioritization include TD error, recentness and corrective feedback, which are mostly heuristically designed. In this work, we start fro...
[ "reinforcement learning", "experience replay" ]
We start from the regret minimization objective, and obtain an optimal prioritization strategy for Bellman update that can directly maximize the return of the policy.
2,108
2105.07253
title_snapshot
pbAmqUUHsQ
Continuous Mean-Covariance Bandits
https://openreview.net/forum?id=pbAmqUUHsQ
[ "Yihan Du", "Siwei Wang", "Zhixuan Fang", "Longbo Huang" ]
Poster
null
Existing risk-aware multi-armed bandit models typically focus on risk measures of individual options such as variance. As a result, they cannot be directly applied to important real-world online decision making problems with correlated options. In this paper, we propose a novel Continuous Mean-Covariance Bandit (CMCB)...
[ "risk-aware bandits", "mean-covariance metric" ]
null
2,024
2102.12090
title_snapshot
dxaINwQdXh1
Risk-Averse Bayes-Adaptive Reinforcement Learning
https://openreview.net/forum?id=dxaINwQdXh1
[ "Marc Rigter", "Bruno Lacerda", "Nick Hawes" ]
Poster
null
In this work, we address risk-averse Bayes-adaptive reinforcement learning. We pose the problem of optimising the conditional value at risk (CVaR) of the total return in Bayes-adaptive Markov decision processes (MDPs). We show that a policy optimising CVaR in this setting is risk-averse to both the epistemic uncertain...
[ "reinforcement learning", "planning", "model-based bayesian reinforcement learning", "risk" ]
Addresses risk sensitive optimisation in the model-based Bayesian reinforcement learning context.
2,000
2102.05762
title_snapshot
pvjfA4wogD6
Video Instance Segmentation using Inter-Frame Communication Transformers
https://openreview.net/forum?id=pvjfA4wogD6
[ "Sukjun Hwang", "Miran Heo", "Seoung Wug Oh", "Seon Joo Kim" ]
Poster
null
We propose a novel end-to-end solution for video instance segmentation (VIS) based on transformers. Recently, the per-clip pipeline shows superior performance over per-frame methods leveraging richer information from multiple frames. However, previous per-clip models require heavy computation and memory usage to achi...
[ "video", "instance segmentation", "video instance segmentation", "tracking", "transformers" ]
null
1,944
2106.03299
title_snapshot
chwaxchpG3
Stronger NAS with Weaker Predictors
https://openreview.net/forum?id=chwaxchpG3
[ "Junru Wu", "Xiyang Dai", "Dongdong Chen", "Yinpeng Chen", "Mengchen Liu", "Ye Yu", "Zhangyang Wang", "Zicheng Liu", "Mei Chen", "Lu Yuan" ]
Poster
null
Neural Architecture Search (NAS) often trains and evaluates a large number of architectures. Recent predictor-based NAS approaches attempt to alleviate such heavy computation costs with two key steps: sampling some architecture-performance pairs and fitting a proxy accuracy predictor. Given limited samples, these predi...
[ "Neural Architecture Search", "Predictor", "Bayesian Optimization" ]
We present a novel method to estimate weak predictors progressively in predictor-based neural architecture search. By coarse-to-fine iteration, the ranking of sampling space is refined gradually which helps find the optimal architectures eventually.
1,939
2102.10490
title_snapshot
kLJjmSrRB3S
Continuous vs. Discrete Optimization of Deep Neural Networks
https://openreview.net/forum?id=kLJjmSrRB3S
[ "Omer Elkabetz", "Nadav Cohen" ]
Spotlight
null
Existing analyses of optimization in deep learning are either continuous, focusing on (variants of) gradient flow, or discrete, directly treating (variants of) gradient descent. Gradient flow is amenable to theoretical analysis, but is stylized and disregards computational efficiency. The extent to which it represent...
[ "Deep Learning", "Non-Convex Optimization", "Gradient Flow", "Gradient Descent" ]
We present a theory quantifying the discrepancy between gradient flow and gradient descent over deep neural networks, and use it to translate an analysis of gradient flow into a new convergence guarantee for gradient descent.
204
2107.06608
title_snapshot
sfzseGUqFrd
POODLE: Improving Few-shot Learning via Penalizing Out-of-Distribution Samples
https://openreview.net/forum?id=sfzseGUqFrd
[ "Duong Hoang Le", "Khoi Duc Nguyen", "Khoi Nguyen", "Quoc-Huy Tran", "Rang Nguyen", "Binh-Son Hua" ]
Poster
null
In this work, we propose to use out-of-distribution samples, i.e., unlabeled samples coming from outside the target classes, to improve few-shot learning. Specifically, we exploit the easily available out-of-distribution samples to drive the classifier to avoid irrelevant features by maximizing the distance from protot...
[ "few-shot learning" ]
We leverage samples from distractor classes or randomly generated noise to improve the generalization of few-shot learner
1,904
2206.04679
title_snapshot
KNSi_LqyORt
Reliable Decisions with Threshold Calibration
https://openreview.net/forum?id=KNSi_LqyORt
[ "Roshni Sahoo", "Shengjia Zhao", "Alyssa Chen", "Stefano Ermon" ]
Poster
null
Decision makers rely on probabilistic forecasts to predict the loss of different decision rules before deployment. When the forecasted probabilities match the true frequencies, predicted losses will be accurate. Although perfect forecasts are typically impossible, probabilities can be calibrated to match the true frequ...
[ "uncertainty quantification", "calibration" ]
null
198
null
null
AAWuCvzaVt
Diffusion Models Beat GANs on Image Synthesis
https://openreview.net/forum?id=AAWuCvzaVt
[ "Prafulla Dhariwal", "Alexander Quinn Nichol" ]
Spotlight
null
We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier g...
[ "generative models", "diffusion models", "score-based models", "denoising diffusion probabilistic models", "image generation", "neural networks", "attention" ]
We achieve state-of-the-art image generation on ImageNet and several LSUN classes with diffusion models.
1,854
2105.05233
title_snapshot
srHp6A1c2z-
Adversarially Robust 3D Point Cloud Recognition Using Self-Supervisions
https://openreview.net/forum?id=srHp6A1c2z-
[ "Jiachen Sun", "Yulong Cao", "Christopher Choy", "Zhiding Yu", "Anima Anandkumar", "Zhuoqing Mao", "Chaowei Xiao" ]
Poster
null
3D point cloud data is increasingly used in safety-critical applications such as autonomous driving. Thus, the robustness of 3D deep learning models against adversarial attacks becomes a major consideration. In this paper, we systematically study the impact of various self-supervised learning proxy tasks on different a...
[ "Adversarial Training", "Point Cloud Recognition", "Self-supervised Learning" ]
In this paper, we demonstrate that appropriate self-supervisions in adversarial training can significantly enhance the robustness in 3D point cloud recognition.
1,847
null
null
6x8tcREIL2W
Gradient-based Hyperparameter Optimization Over Long Horizons
https://openreview.net/forum?id=6x8tcREIL2W
[ "Paul Micaelli", "Amos Storkey" ]
Poster
null
Gradient-based hyperparameter optimization has earned a widespread popularity in the context of few-shot meta-learning, but remains broadly impractical for tasks with long horizons (many gradient steps), due to memory scaling and gradient degradation issues. A common workaround is to learn hyperparameters online, but t...
[ "Meta-learning", "Hyperparameter Optimization", "Gradient-based", "Gradient Degradation", "Forward-mode differentiation", "AutoML" ]
Hyperparameter sharing combined with forward mode differentiation enables effective HPO over long horizons
1,821
2007.07869
title_snapshot
AVvcLO2UYGA
Beyond Bandit Feedback in Online Multiclass Classification
https://openreview.net/forum?id=AVvcLO2UYGA
[ "Dirk van der Hoeven", "Federico Fusco", "Nicolò Cesa-Bianchi" ]
Poster
null
We study the problem of online multiclass classification in a setting where the learner's feedback is determined by an arbitrary directed graph. While including bandit feedback as a special case, feedback graphs allow a much richer set of applications, including filtering and label efficient classification. We introduc...
[ "Online learning", "multiclass classification", "bandit algorithms", "surrogate losses", "feedback graphs" ]
We present new and improved results for the online multiclass classification setting.
1,810
2106.03596
title_snapshot
9QwPhXWmuRp
Interventional Sum-Product Networks: Causal Inference with Tractable Probabilistic Models
https://openreview.net/forum?id=9QwPhXWmuRp
[ "Matej Zecevic", "Devendra Singh Dhami", "Athresh Karanam", "Sriraam Natarajan", "Kristian Kersting" ]
Poster
null
While probabilistic models are an important tool for studying causality, doing so suffers from the intractability of inference. As a step towards tractable causal models, we consider the problem of learning interventional distributions using sum-product networks (SPNs) that are over-parameterized by gate functions, e.g...
[ "Interventions", "Causality", "Tractablity", "Probabilistic Models" ]
We consider the problem of learning interventional distributions (i.e., answering causal queries) with tractable probabilistic models (gated SPN).
1,769
2102.10440
title_snapshot
vKxFYApxBjr
Exploiting Domain-Specific Features to Enhance Domain Generalization
https://openreview.net/forum?id=vKxFYApxBjr
[ "Ha Manh Bui", "Toan Tran", "Anh Tuan Tran", "Dinh Phung" ]
Poster
null
Domain Generalization (DG) aims to train a model, from multiple observed source domains, in order to perform well on unseen target domains. To obtain the generalization capability, prior DG approaches have focused on extracting domain-invariant information across sources to generalize on target domains, while useful do...
[ "Representation Learning", "Domain Generalization" ]
mDSDI: a theoretical sound framework in DG
1,668
2110.09410
title_snapshot
_RnHyIeu5Y5
ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias
https://openreview.net/forum?id=_RnHyIeu5Y5
[ "Yufei Xu", "Qiming ZHANG", "Jing Zhang", "Dacheng Tao" ]
Poster
null
Transformers have shown great potential in various computer vision tasks owing to their strong capability in modeling long-range dependency using the self-attention mechanism. Nevertheless, vision transformers treat an image as 1D sequence of visual tokens, lacking an intrinsic inductive bias (IB) in modeling local vis...
[ "Vision Transformer", "Classification", "Convolution", "Inductive Bias" ]
null
1,628
2106.03348
title_snapshot
ZsGg52s-cQZ
Topology-Imbalance Learning for Semi-Supervised Node Classification
https://openreview.net/forum?id=ZsGg52s-cQZ
[ "Deli Chen", "Yankai Lin", "Guangxiang Zhao", "Xuancheng Ren", "Peng Li", "Jie Zhou", "Xu Sun" ]
Poster
null
The class imbalance problem, as an important issue in learning node representations, has drawn increasing attention from the community. Although the imbalance considered by existing studies roots from the unequal quantity of labeled examples in different classes (quantity imbalance), we argue that graph data expose a u...
[ "node classification", "topology imbalance learning", "semi-supervised learning", "graph neural network" ]
This paper study a graph-specific imbalance issue: topology imbalance and the relative solution.
164
2110.04099
title_snapshot
MXmmuhJYPdU
Uncertainty-Driven Loss for Single Image Super-Resolution
https://openreview.net/forum?id=MXmmuhJYPdU
[ "Qian Ning", "Weisheng Dong", "Xin Li", "Jinjian Wu", "Guangming Shi" ]
Poster
null
In low-level vision such as single image super-resolution (SISR), traditional MSE or L_1 loss function treats every pixel equally with the assumption that the importance of all pixels is the same. However, it has been long recognized that texture and edge areas carry more important visual information than smooth areas ...
[ "Uncertainty-Driven Loss", "Uncertainty Learning", "Single Image Super-Resolution" ]
To prioritize visually important image structures (e.g., texture and edge pixels), we propose a novel uncertainty-driven loss function and demonstrate its supriority to traditional MSE/L_1 losses for SISR.
1,581
null
null
ZCDObqK6ifu
SILG: The Multi-domain Symbolic Interactive Language Grounding Benchmark
https://openreview.net/forum?id=ZCDObqK6ifu
[ "Victor Zhong", "Austin W. Hanjie", "Sida Wang", "Karthik R Narasimhan", "Luke Zettlemoyer" ]
Poster
null
Existing work in language grounding typically study single environments. How do we build unified models that apply across multiple environments? We propose the multi-environment Symbolic Interactive Language Grounding benchmark (SILG), which unifies a collection of diverse grounded language learning environments under ...
[ "language grounding", "reinforcement learning" ]
We propose a symbolic interactive language grounding benchmark that unifies multiple domains under a common interface so that researchers can quickly evaluate new models and algorithms across a variety of grounding challenges.
153
2110.10661
title_judge
h3qQzodaAq7
$\alpha$-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression
https://openreview.net/forum?id=h3qQzodaAq7
[ "Jiabo He", "Sarah Monazam Erfani", "Xingjun Ma", "James Bailey", "Ying Chi", "Xian-Sheng Hua" ]
Poster
null
Bounding box (bbox) regression is a fundamental task in computer vision. So far, the most commonly used loss functions for bbox regression are the Intersection over Union (IoU) loss and its variants. In this paper, we generalize existing IoU-based losses to a new family of power IoU losses that have a power IoU term an...
[ "bounding box regression", "localization loss", "object detection", "intersection over union" ]
We propose the family of power IoU losses as new options for localization losses in object detection.
151
2110.13675
title_judge
cD2Ls4qXTc
Fast Pure Exploration via Frank-Wolfe
https://openreview.net/forum?id=cD2Ls4qXTc
[ "Po-An Wang", "Ruo-Chun Tzeng", "Alexandre Proutiere" ]
Poster
null
We study the problem of active pure exploration with fixed confidence in generic stochastic bandit environments. The goal of the learner is to answer a query about the environment with a given level of certainty while minimizing her sampling budget. For this problem, instance-specific lower bounds on the expected sampl...
[ "pure exploration", "Frank-Wolfe", "structured bandits" ]
null
1,449
null
null
OU4LL1qP3Dg
Probability Paths and the Structure of Predictions over Time
https://openreview.net/forum?id=OU4LL1qP3Dg
[ "Zhiyuan Jerry Lin", "Hao Sheng", "Sharad Goel" ]
Poster
null
In settings ranging from weather forecasts to political prognostications to financial projections, probability estimates of future binary outcomes often evolve over time. For example, the estimated likelihood of rain on a specific day changes by the hour as new information becomes available. Given a collection of such ...
[ "Probability Path", "Bayesian Time Series Modeling", "Evolving Prediction", "Martingale" ]
We introduce an interesting yet understudied problem that concerns the dynamic structure of evolving predictions over time and propose a Bayesian framework that can coherently model such structure.
1,391
2106.06515
title_snapshot
U9NNzquYEHC
Linear-Time Probabilistic Solution of Boundary Value Problems
https://openreview.net/forum?id=U9NNzquYEHC
[ "Nicholas Krämer", "Philipp Hennig" ]
Poster
null
We propose a fast algorithm for the probabilistic solution of boundary value problems (BVPs), which are ordinary differential equations subject to boundary conditions. In contrast to previous work, we introduce a Gauss-Markov prior and tailor it specifically to BVPs, which allows computing a posterior distribution ove...
[ "Differential equations", "Gaussian processes", "probabilistic numerics" ]
We introduce a Markov prior and other practical considerations to probabilistic solvers for boundary value problems.
1,384
null
null
0lz4QxW2tDf
Balanced Chamfer Distance as a Comprehensive Metric for Point Cloud Completion
https://openreview.net/forum?id=0lz4QxW2tDf
[ "Tong Wu", "Liang Pan", "Junzhe Zhang", "Tai WANG", "Ziwei Liu", "Dahua Lin" ]
Poster
null
Chamfer Distance (CD) and Earth Mover’s Distance (EMD) are two broadly adopted metrics for measuring the similarity between two point sets. However, CD is usually insensitive to mismatched local density, and EMD is usually dominated by global distribution while overlooks the fidelity of detailed structures. Besides, th...
[ "Point Cloud Completion", "Similarity Metrics" ]
We propose a comprehensive metric for point cloud similarity and examine it in the task of point cloud completion.
1,372
2111.12702
title_judge
J4gRj6d5Qm
Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting
https://openreview.net/forum?id=J4gRj6d5Qm
[ "Haixu Wu", "Jiehui Xu", "Jianmin Wang", "Mingsheng Long" ]
Poster
null
Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-ran...
[ "Time Series Forecasting", "Transformers", "Deep Decomposition Model", "Auto-Correlation" ]
We renovate Transformer to a deep decomposition model Autoformer and propose a series-wise Auto-Correlation mechanism based on series periodicity to replace self-attention. Autoformer surpasses SOTA by 38% relative accuracy promotion on six datasets.
1,349
2106.13008
title_snapshot
UKoV0-BamX4
On Locality of Local Explanation Models
https://openreview.net/forum?id=UKoV0-BamX4
[ "Sahra Ghalebikesabi", "Lucile Ter-Minassian", "Karla DiazOrdaz", "Christopher C. Holmes" ]
Poster
null
Shapley values provide model agnostic feature attributions for model outcome at a particular instance by simulating feature absence under a global population distribution. The use of a global population can lead to potentially misleading results when local model behaviour is of interest. Hence we consider the formulat...
[ "Shapley values", "model interpretability", "reference distribution", "neighbourhood sampling" ]
We consider the formulation of neighbourhood reference distributions that improve the local interpretability of Shapley values.
1,279
2106.14648
title_snapshot
xmMHxfE1qS6
Adversarially Robust Change Point Detection
https://openreview.net/forum?id=xmMHxfE1qS6
[ "Mengchu Li", "Yi Yu" ]
Poster
null
Change point detection is becoming increasingly popular in many application areas. On one hand, most of the theoretically-justified methods are investigated in an ideal setting without model violations, or merely robust against identical heavy-tailed noise distribution across time and/or against isolate outliers; on th...
[ "change point detection", "robust statistics" ]
null
1,271
2105.10417
title_snapshot
tSfud5OOqR
Edge Representation Learning with Hypergraphs
https://openreview.net/forum?id=tSfud5OOqR
[ "Jaehyeong Jo", "Jinheon Baek", "Seul Lee", "Dongki Kim", "Minki Kang", "Sung Ju Hwang" ]
Poster
null
Graph neural networks have recently achieved remarkable success in representing graph-structured data, with rapid progress in both the node embedding and graph pooling methods. Yet, they mostly focus on capturing information from the nodes considering their connectivity, and not much work has been done in representing ...
[ "Graph Neural Network", "Edge Representation Learning", "Graph Pooling", "Hypergraph" ]
We propose a novel edge representation learning scheme with hypergraphs, which can be further exploited for graph pooling.
125
2106.15845
title_snapshot
0xs40KGnsq3
Learning Student-Friendly Teacher Networks for Knowledge Distillation
https://openreview.net/forum?id=0xs40KGnsq3
[ "Dae Young Park", "Moon-Hyun Cha", "Changwook Jeong", "Daesin Kim", "Bohyung Han" ]
Poster
null
We propose a novel knowledge distillation approach to facilitate the transfer of dark knowledge from a teacher to a student. Contrary to most of the existing methods that rely on effective training of student models given pretrained teachers, we aim to learn the teacher models that are friendly to students and, consequ...
[ "deep learning", "transfer learning", "knowledge distillation" ]
We aim to learn the teacher models that are friendly to students and, consequently, more appropriate for knowledge transfer.
1,163
2102.07650
title_snapshot
_Eo8bl4MpT3
Fast Algorithms for $L_\infty$-constrained S-rectangular Robust MDPs
https://openreview.net/forum?id=_Eo8bl4MpT3
[ "Bahram Behzadian", "Marek Petrik", "Chin Pang Ho" ]
Poster
null
Robust Markov decision processes (RMDPs) are a useful building block of robust reinforcement learning algorithms but can be hard to solve. This paper proposes a fast, exact algorithm for computing the Bellman operator for S-rectangular robust Markov decision processes with $L_\infty$-constrained rectangular ambiguity s...
[ "reinforcement learning", "robust Markov decision processing", "rectangular ambiguity sets" ]
null
120
null
null
oqKC5A7iq_k
Low-Fidelity Video Encoder Optimization for Temporal Action Localization
https://openreview.net/forum?id=oqKC5A7iq_k
[ "Mengmeng Xu", "Juan-Manuel Perez-Rua", "Xiatian Zhu", "Bernard Ghanem", "Brais Martinez" ]
Poster
null
Most existing temporal action localization (TAL) methods rely on a transfer learning pipeline: by first optimizing a video encoder on a large action classification dataset (i.e., source domain), followed by freezing the encoder and training a TAL head on the action localization dataset (i.e., target domain). This resul...
[ "End-to-End Pre-training", "Temporal Action Localization" ]
We tackle the problem of temporal action localization and propose an intermediate low-fidelity optimization stage to jointly optimize the video encoder and the localization head on the target dataset and task, outperforming SOTA performance.
1,078
2103.15233
title_judge
ilVv1LO0Ew
Exploring Cross-Video and Cross-Modality Signals for Weakly-Supervised Audio-Visual Video Parsing
https://openreview.net/forum?id=ilVv1LO0Ew
[ "Yan-Bo Lin", "Hung-Yu Tseng", "Hsin-Ying Lee", "Yen-Yu Lin", "Ming-Hsuan Yang" ]
Poster
null
The audio-visual video parsing task aims to temporally parse a video into audio or visual event categories. However, it is labor intensive to temporally annotate audio and visual events and thus hampers the learning of a parsing model. To this end, we propose to explore additional cross-video and cross-modality supervi...
[ "Audio-visual Video Parsing", "Weakly-supervised Learning", "Vision Applications and Systems" ]
null
110
null
null
KfC0i9Hjvl2
Locally private online change point detection
https://openreview.net/forum?id=KfC0i9Hjvl2
[ "Thomas Berrett", "Yi Yu" ]
Poster
null
We study online change point detection problems under the constraint of local differential privacy (LDP) where, in particular, the statistician does not have access to the raw data. As a concrete problem, we study a multivariate nonparametric regression problem. At each time point $t$, the raw data are assumed to be ...
[ "Locally private data", "online change point detection", "multivariate nonparametric regression" ]
Optimal online change point detection in multivariate nonparametric estimation problem with locally private data.
951
2105.10675
title_snapshot
WxH774N0mEu
FACMAC: Factored Multi-Agent Centralised Policy Gradients
https://openreview.net/forum?id=WxH774N0mEu
[ "Bei Peng", "Tabish Rashid", "Christian Schroeder de Witt", "Pierre-Alexandre Kamienny", "Philip Torr", "Wendelin Boehmer", "Shimon Whiteson" ]
Poster
null
We propose FACtored Multi-Agent Centralised policy gradients (FACMAC), a new method for cooperative multi-agent reinforcement learning in both discrete and continuous action spaces. Like MADDPG, a popular multi-agent actor-critic method, our approach uses deep deterministic policy gradients to learn policies. However, ...
[ "multi-agent reinforcement learning", "multi-agent policy gradients" ]
We propose a multi-agent actor-critic method that fully benefits from a centralised critic by using a centralised policy gradient, scales gracefully by factoring the critic, and achieves SOTA performance on discrete and continuous cooperative tasks.
105
2003.06709
title_snapshot
f5liPryFRoA
Adversarial Reweighting for Partial Domain Adaptation
https://openreview.net/forum?id=f5liPryFRoA
[ "Xiang Gu", "Xi Yu", "Yan Yang", "Jian Sun", "Zongben Xu" ]
Poster
null
Partial domain adaptation (PDA) has gained much attention due to its practical setting. The current PDA methods usually adapt the feature extractor by aligning the target and reweighted source domain distributions. In this paper, we experimentally find that the feature adaptation by the reweighted distribution alignmen...
[ "Partial Domain Adaptation", "Adversarial Reweighting", "Negative Domain Transfer", "Wasserstein" ]
We investigated the limitations of feature adaptation for partial domain adaptation (PDA), and proposed a novel adversarial reweighting method for PDA, and achieved SOTA results on challenging benchmark datasets.
887
null
null
9TX5OsKJvm
Post-Training Quantization for Vision Transformer
https://openreview.net/forum?id=9TX5OsKJvm
[ "Zhenhua Liu", "Yunhe Wang", "Kai Han", "Wei Zhang", "Siwei Ma", "Wen Gao" ]
Poster
null
Recently, transformer has achieved remarkable performance on a variety of computer vision applications. Compared with mainstream convolutional neural networks, vision transformers are often of sophisticated architectures for extracting powerful feature representations, which are more difficult to be developed on mobile...
[ "Post-training", "vision transformer", "ranking loss of self-attention", "mixed-precision" ]
We propose a post-training quantization scheme for visual transformer, which consider the ranking loss of self-attention and take the nuclear norm of the features as the evaluation of the sensitivity of the transformer layer,
102
2106.14156
title_snapshot
haSQRA5RnuM
Which Mutual-Information Representation Learning Objectives are Sufficient for Control?
https://openreview.net/forum?id=haSQRA5RnuM
[ "Kate Rakelly", "Abhishek Gupta", "Carlos Florensa", "Sergey Levine" ]
Poster
null
Mutual information (MI) maximization provides an appealing formalism for learning representations of data. In the context of reinforcement learning (RL), such representations can accelerate learning by discarding irrelevant and redundant information, while retaining the information necessary for control. Much prior wor...
[ "reinforcement learning", "representation learning", "mutual information" ]
We theoretically analyze whether popular MI-based representation learning objectives for RL yield state representations sufficient for learning and representing optimal control policies, and illustrate our findings with deep RL experiments.
868
2106.07278
title_snapshot
P-if5sUWBn
Deformable Butterfly: A Highly Structured and Sparse Linear Transform
https://openreview.net/forum?id=P-if5sUWBn
[ "Rui Lin", "Jie Ran", "King Hung Chiu", "Graziano Chesi", "Ngai Wong" ]
Poster
null
We introduce a new kind of linear transform named Deformable Butterfly (DeBut) that generalizes the conventional butterfly matrices and can be adapted to various input-output dimensions. It inherits the fine-to-coarse-grained learnable hierarchy of traditional butterflies and when deployed to neural networks, the promi...
[ "Deformable Butterfly", "Linear transform", "Model compression" ]
A new kind of linear transform named deformable butterfly (DeBut) that generalizes the conventional butterfly matrices and can be adopted to various input-output dimensions.
773
2203.13556
title_snapshot
eXlxB3aLOe
GradInit: Learning to Initialize Neural Networks for Stable and Efficient Training
https://openreview.net/forum?id=eXlxB3aLOe
[ "Chen Zhu", "Renkun Ni", "Zheng Xu", "Kezhi Kong", "W Ronny Huang", "Tom Goldstein" ]
Poster
null
Innovations in neural architectures have fostered significant breakthroughs in language modeling and computer vision. Unfortunately, novel architectures often result in challenging hyper-parameter choices and training instability if the network parameters are not properly initialized. A number of architecture-specific ...
[ "Initialization", "Transformers", "Convolutional Networks" ]
We propose an automated and architecture agnostic method for initializing neural networks that works well for CNNs and Transformers.
761
2102.08098
title_snapshot
e5vrkfc5aau
Towards Multi-Grained Explainability for Graph Neural Networks
https://openreview.net/forum?id=e5vrkfc5aau
[ "Xiang Wang", "Yingxin Wu", "An Zhang", "Xiangnan He", "Tat-Seng Chua" ]
Poster
null
When a graph neural network (GNN) made a prediction, one raises question about explainability: “Which fraction of the input graph is most influential to the model’s decision?” Producing an answer requires understanding the model’s inner workings in general and emphasizing the insights on the decision for the instance at...
[ "Graph Neural Networks", "Multi-grained Explainability", "Feature Attribution" ]
Towards multi-grained explainability of graph neural networks, we integrate the idea of pre-training and fine-tuning in the explainers.
698
null
null
Cn7d_BHE-s
Compressed Video Contrastive Learning
https://openreview.net/forum?id=Cn7d_BHE-s
[ "Yuqi Huo", "Mingyu Ding", "Haoyu Lu", "Nanyi Fei", "Zhiwu Lu", "Ji-Rong Wen", "Ping Luo" ]
Poster
null
This work concerns self-supervised video representation learning (SSVRL), one topic that has received much attention recently. Since videos are storage-intensive and contain a rich source of visual content, models designed for SSVRL are expected to be storage- and computation-efficient, as well as effective. However, m...
[ "compressed video", "self-supervised learning", "motion vector", "contrastive learning" ]
We propose an efficient and effective framework for self-supervised representation learning from compressed videos (without decompressing off-the-fly).
89
null
null
GSHFVNejxs7
PlayVirtual: Augmenting Cycle-Consistent Virtual Trajectories for Reinforcement Learning
https://openreview.net/forum?id=GSHFVNejxs7
[ "Tao Yu", "Cuiling Lan", "Wenjun Zeng", "Mingxiao Feng", "Zhizheng Zhang", "Zhibo Chen" ]
Poster
null
Learning good feature representations is important for deep reinforcement learning (RL). However, with limited experience, RL often suffers from data inefficiency for training. For un-experienced or less-experienced trajectories (i.e., state-action sequences), the lack of data limits the use of them for better feature ...
[ "Reinforcement learning", "data efficiency", "representation learning", "augmentation", "virtual trajectory", "cycle consistency" ]
We demonstrate that augmenting cycle-consistent virtual trajectories can significantly improve RL data efficiency.
88
2106.04152
title_snapshot
yhjpeuWepoj
Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation
https://openreview.net/forum?id=yhjpeuWepoj
[ "Shiqi Yang", "Yaxing Wang", "Joost van de weijer", "Luis Herranz", "SHANGLING JUI" ]
Poster
null
Domain adaptation (DA) aims to alleviate the domain shift between source domain and target domain. Most DA methods require access to the source data, but often that is not possible (e.g. due to data privacy or intellectual property). In this paper, we address the challenging source-free domain adaptation (SFDA) problem...
[ "source-free domain adaptation", "reciprocal nearest neighbors" ]
We tackle source free domain adaptation problem by exploiting the intrinsic neighborhood structure of target data.
584
2110.04202
title_snapshot
nRBZWEUhIhW
GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinement
https://openreview.net/forum?id=nRBZWEUhIhW
[ "Martin Engelcke", "Oiwi Parker Jones", "Ingmar Posner" ]
Poster
null
Advances in unsupervised learning of object-representations have culminated in the development of a broad range of methods for unsupervised object segmentation and interpretable object-centric scene generation. These methods, however, are limited to simulated and real-world datasets with limited visual complexity. More...
[ "Generative models", "object-centric representations", "variational inference" ]
We present an improved object-centric generative model of visual scenes that uses a stochastic clustering algorithm for inferring object representations without imposing a fixed ordering on objects or using iterative refinement.
81
2104.09958
title_snapshot
zAuDbrHC6fq
Aligned Structured Sparsity Learning for Efficient Image Super-Resolution
https://openreview.net/forum?id=zAuDbrHC6fq
[ "Yulun Zhang", "Huan Wang", "Can Qin", "Yun Fu" ]
Spotlight
null
Lightweight image super-resolution (SR) networks have obtained promising results with moderate model size. Many SR methods have focused on designing lightweight architectures, which neglect to further reduce the redundancy of network parameters. On the other hand, model compression techniques, like neural architecture ...
[ "Neural Network Pruning", "Lightweight Image Super-Resolution", "Aligned Structured Sparsity Learning" ]
Optimize image SR networks with network pruning simultaneously and achieve SOTA results
555
null
null
ar85GL0N11
Visual Search Asymmetry: Deep Nets and Humans Share Similar Inherent Biases
https://openreview.net/forum?id=ar85GL0N11
[ "Shashi Kant Gupta", "Mengmi Zhang", "Chia-Chien Wu", "Jeremy M. Wolfe", "Gabriel Kreiman" ]
Poster
null
Visual search is a ubiquitous and often challenging daily task, exemplified by looking for the car keys at home or a friend in a crowd. An intriguing property of some classical search tasks is an asymmetry such that finding a target A among distractors B can be easier than finding B among A. To elucidate the mechanisms...
[ "visual search", "asymmetry", "deep learning", "recognition", "neuroscience", "eye movement", "visual cortex", "psychophyiscs", "human behavior", "visual cognition", "inductive bias", "natural statistics" ]
Deep nets and humans share similar inherent biases in visual search asymmetry
529
2106.02953
title_snapshot
9N_vdopOU0h
PolarStream: Streaming Object Detection and Segmentation with Polar Pillars
https://openreview.net/forum?id=9N_vdopOU0h
[ "Qi Chen", "Sourabh Vora", "Oscar Beijbom" ]
Poster
null
Recent works recognized lidars as an inherently streaming data source and showed that the end-to-end latency of lidar perception models can be reduced significantly by operating on wedge-shaped point cloud sectors rather then the full point cloud. However, due to use of cartesian coordinate systems these methods repre...
[ "Streaming Object Detection", "3D Object Detection", "LiDAR Segmentation", "3D Panoptic Segmentation", "Polar Grid" ]
This paper proposes a streaming model for simultaneous 3D object detection, Lidar segmentation and panoptic segmentation based on polar representation
483
2106.07545
title_judge
5-GXHFNbq_U
Prototypical Cross-Attention Networks for Multiple Object Tracking and Segmentation
https://openreview.net/forum?id=5-GXHFNbq_U
[ "Lei Ke", "Xia Li", "Martin Danelljan", "Yu-Wing Tai", "Chi-Keung Tang", "Fisher Yu" ]
Spotlight
null
Multiple object tracking and segmentation requires detecting, tracking, and segmenting objects belonging to a set of given classes. Most approaches only exploit the temporal dimension to address the association problem, while relying on single frame predictions for the segmentation mask itself. We propose Prototypical ...
[ "multiple object tracking and segmentation", "video instance segmentation", "efficient cross-attention networks", "space-time memory" ]
We design efficient cross-attention networks (PCAN) for multiple object tracking and segmentation.
59
2106.11958
title_snapshot
-K4tIyQLaY
Dual Progressive Prototype Network for Generalized Zero-Shot Learning
https://openreview.net/forum?id=-K4tIyQLaY
[ "Chaoqun Wang", "Shaobo Min", "Xuejin Chen", "Xiaoyan Sun", "Houqiang Li" ]
Poster
null
Generalized Zero-Shot Learning (GZSL) aims to recognize new categories with auxiliary semantic information, e.g., category attributes. In this paper, we handle the critical issue of domain shift problem, i.e., confusion between seen and unseen categories, by progressively improving cross-domain transferability and cate...
[ "generalized zero-shot learning", "progressive prototypes" ]
We propose a novel Dual Progressive Prototype Network (DPPN) to alleviate domain shift problem in GZSL by progressively improving cross-domain transferability and category discriminability of visual representations.
292
2111.02073
title_snapshot
NO_cSsVghGb
Neural Architecture Dilation for Adversarial Robustness
https://openreview.net/forum?id=NO_cSsVghGb
[ "Yanxi Li", "Zhaohui Yang", "Yunhe Wang", "Chang Xu" ]
Poster
null
With the tremendous advances in the architecture and scale of convolutional neural networks (CNNs) over the past few decades, they can easily reach or even exceed the performance of humans in certain tasks. However, a recently discovered shortcoming of CNNs is that they are vulnerable to adversarial attacks. Although t...
[ "Deep Learning", "Neural Architecture Search", "Adversarial Robustness" ]
null
266
2108.06885
title_snapshot
lVBu4PqM9HU
Localization with Sampling-Argmax
https://openreview.net/forum?id=lVBu4PqM9HU
[ "Jiefeng Li", "Tong Chen", "Ruiqi Shi", "Yujing Lou", "Yong-Lu Li", "Cewu Lu" ]
Poster
null
Soft-argmax operation is commonly adopted in detection-based methods to localize the target position in a differentiable manner. However, training the neural network with soft-argmax makes the shape of the probability map unconstrained. Consequently, the model lacks pixel-wise supervision through the map during trainin...
[ "soft-argmax", "probability map", "localization" ]
We propose sampling-argmax, a differentiable training method that imposes implicit constraints to the shape of the probability map and facilitates localization.
242
2110.08825
title_snapshot
kwN2xvZ2XZ9
Learning Frequency Domain Approximation for Binary Neural Networks
https://openreview.net/forum?id=kwN2xvZ2XZ9
[ "Yixing Xu", "Kai Han", "Chang Xu", "Yehui Tang", "Chunjing Xu", "Yunhe Wang" ]
Oral
null
Binary neural networks (BNNs) represent original full-precision weights and activations into 1-bit with sign function. Since the gradient of the conventional sign function is almost zero everywhere which cannot be used for back-propagation, several attempts have been proposed to alleviate the optimization difficulty by...
[ "binary neural network", "frequency domain approximation", "fourier series", "noise" ]
Using sine module and noise adaptation module to approximate sign function in BNN.
40
2103.00841
title_snapshot
NlLynLBBi01
DP-SSL: Towards Robust Semi-supervised Learning with A Few Labeled Samples
https://openreview.net/forum?id=NlLynLBBi01
[ "Yi Xu", "Jiandong Ding", "Lu Zhang", "Shuigeng Zhou" ]
Poster
null
The scarcity of labeled data is a critical obstacle to deep learning. Semi-supervised learning (SSL) provides a promising way to leverage unlabeled data by pseudo labels. However, when the size of labeled data is very small (say a few labeled samples per class), SSL performs poorly and unstably, possibly due to the low...
[ "semi-supervised learning", "data programming", "classification" ]
A novel semi-supervised learning method that is effective and robust even only a few labeled samples per class are available.
202
2110.13740
title_snapshot
HRE7guiwMgG
An Empirical Study of Adder Neural Networks for Object Detection
https://openreview.net/forum?id=HRE7guiwMgG
[ "Xinghao Chen", "Chang Xu", "Minjing Dong", "Chunjing Xu", "Yunhe Wang" ]
Poster
null
Adder neural networks (AdderNets) have shown impressive performance on image classification with only addition operations, which are more energy efficient than traditional convolutional neural networks built with multiplications. Compared with classification, there is a strong demand on reducing the energy consumption ...
[ "Object Detection", "Adder Neural Networks" ]
An emperical study of how to use adder neural networks to build accurate object detectors.
23
2112.13608
title_snapshot
sjRlHsawmRf
On the Representation Power of Set Pooling Networks
https://openreview.net/forum?id=sjRlHsawmRf
[ "Christian Bueno", "Alan Hylton" ]
Poster
null
Point clouds and sets are input data-types which pose unique problems to deep learning. Since sets can have variable cardinality and are unchanged by permutation, the input space for these problems naturally form infinite-dimensional non-Euclidean spaces. Despite these mathematical difficulties, PointNet (Qi et al. 201...
[ "universal approximation", "point clouds", "sets", "topology", "functional analysis", "non-euclidean", "invariance" ]
Proves new cardinality-agnostic universality results for deep learning on point clouds, explores cases where approximation fails, and compares the representational power of DeepSets and PointNet.
11,720
null
null
0IqTX6FcZWv
Capacity and Bias of Learned Geometric Embeddings for Directed Graphs
https://openreview.net/forum?id=0IqTX6FcZWv
[ "Michael Boratko", "Dongxu Zhang", "Nicholas Monath", "Luke Vilnis", "Kenneth L. Clarkson", "Andrew McCallum" ]
Poster
null
A wide variety of machine learning tasks such as knowledge base completion, ontology alignment, and multi-label classification can benefit from incorporating into learning differentiable representations of graphs or taxonomies. While vectors in Euclidean space can theoretically represent any graph, much recent work sh...
[ "graph embeddings", "representation learning", "knowledge graphs", "structured prediction" ]
We introduce a novel geometric embedding method for capturing graph structure, prove it's ability to represent any DAG, and empirically analyze the representational capacity and bias of a large set of geometric embeddings for graph modeling.
11,709
null
null