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
9jRH00HT4-4
Distilling Meta Knowledge on Heterogeneous Graph for Illicit Drug Trafficker Detection on Social Media
https://openreview.net/forum?id=9jRH00HT4-4
[ "Yiyue Qian", "Yiming Zhang", "Yanfang Ye", "Chuxu Zhang" ]
Poster
null
Driven by the considerable profits, the crime of drug trafficking (a.k.a. illicit drug trading) has co-evolved with modern technologies, e.g., social media such as Instagram has become a popular platform for marketing and selling illicit drugs. The activities of online drug trafficking are nimble and resilient, which c...
[ "graph representation learning", "few-shot learning", "drug trafficker detection" ]
develop a novel heterogeneous graph model for illicit drug trafficker detection.
9,359
null
null
9FREJhzo1q
A sampling-based circuit for optimal decision making
https://openreview.net/forum?id=9FREJhzo1q
[ "Camille E. Rullán Buxó", "Cristina Savin" ]
Spotlight
null
Many features of human and animal behavior can be understood in the framework of Bayesian inference and optimal decision making, but the biological substrate of such processes is not fully understood. Neural sampling provides a flexible code for probabilistic inference in high dimensions and explains key features of se...
[ "computational neuroscience", "neural sampling", "spiking network", "optimal decision making" ]
We describe a spiking neural circuit that performs approximately optimal decision making and analyze its computational and representational properties.
9,345
null
null
0FDxsIEv9G
Deep Proxy Causal Learning and its Application to Confounded Bandit Policy Evaluation
https://openreview.net/forum?id=0FDxsIEv9G
[ "Liyuan Xu", "Heishiro Kanagawa", "Arthur Gretton" ]
Poster
null
Proxy causal learning (PCL) is a method for estimating the causal effect of treatments on outcomes in the presence of unobserved confounding, using proxies (structured side information) for the confounder. This is achieved via two-stage regression: in the first stage, we model relations among the treatment and proxies;...
[ "Causal Inference", "Deep Learning", "Off-policy evaluation" ]
null
9,342
2106.03907
title_snapshot
V5prUHOrOP4
Revisit Multimodal Meta-Learning through the Lens of Multi-Task Learning
https://openreview.net/forum?id=V5prUHOrOP4
[ "Milad Abdollahzadeh", "Touba Malekzadeh", "Ngai-man Cheung" ]
Poster
null
Multimodal meta-learning is a recent problem that extends conventional few-shot meta-learning by generalizing its setup to diverse multimodal task distributions. This setup makes a step towards mimicking how humans make use of a diverse set of prior skills to learn new skills. Previous work has achieved encouraging per...
[ "Meta-Learning", "Few-Shot Learning", "Multi-Task Learning", "Transference Analysis" ]
This paper proposes a method to quantify knowledge transfer between few-shot tasks. It also proposes a multimodal meta-learner which advances the state-of-the-art with significant margins.
9,338
2110.14202
title_snapshot
jar9C-V8GH
Coresets for Time Series Clustering
https://openreview.net/forum?id=jar9C-V8GH
[ "Lingxiao Huang", "K. Sudhir", "Nisheeth K Vishnoi" ]
Spotlight
null
We study the problem of constructing coresets for clustering problems with time series data. This problem has gained importance across many fields including biology, medicine, and economics due to the proliferation of sensors facilitating real-time measurement and rapid drop in storage costs. In particular, we consid...
[ "coresets", "clustering", "time series data", "Gaussian mixture model", "autocorrelation" ]
We present an efficient algorithm to construct coresets for clustering Gaussian mixture time series data.
9,329
2110.15263
title_snapshot
a9WXj5XV5mK
Meta-Learning the Search Distribution of Black-Box Random Search Based Adversarial Attacks
https://openreview.net/forum?id=a9WXj5XV5mK
[ "Maksym Yatsura", "Jan Hendrik Metzen", "Matthias Hein" ]
Poster
null
Adversarial attacks based on randomized search schemes have obtained state-of-the-art results in black-box robustness evaluation recently. However, as we demonstrate in this work, their efficiency in different query budget regimes depends on manual design and heuristic tuning of the underlying proposal distributions. W...
[ "black-box adversarial attacks", "meta-learning" ]
Meta-learned adaptive controllers of the search distribution can make random search based adversarial attacks, in particular Square Attack, more efficient in different query budget regimes.
9,328
2111.01714
title_snapshot
soDi-HkzC1
Generative vs. Discriminative: Rethinking The Meta-Continual Learning
https://openreview.net/forum?id=soDi-HkzC1
[ "Mohammad Amin Banayeean Zade", "Rasoul Mirzaiezadeh", "Hosein Hasani", "Mahdieh Soleymani Baghshah" ]
Poster
null
Deep neural networks have achieved human-level capabilities in various learning tasks. However, they generally lose performance in more realistic scenarios like learning in a continual manner. In contrast, humans can incorporate their prior knowledge to learn new concepts efficiently without forgetting older ones. In t...
[ "Continual Learning", "Meta-Learning", "Meta-Continual Learning", "Generative Classifier", "Bayesian Classifier", "Neuro-Inspired Artificial Intelligence" ]
This paper addresses catastrophic forgetting in continual learning using a meta-learned feature extractor and a generative classifier.
9,313
null
null
Xl1Z1L9DBIJ
Topological Attention for Time Series Forecasting
https://openreview.net/forum?id=Xl1Z1L9DBIJ
[ "Sebastian Zeng", "Florian Graf", "Christoph Hofer", "Roland Kwitt" ]
Poster
null
The problem of (point) forecasting univariate time series is considered. Most approaches, ranging from traditional statistical methods to recent learning-based techniques with neural networks, directly operate on raw time series observations. As an extension, we study whether local topological properties, as captured v...
[ "Time series forecasting", "Persistent homology", "Attention", "Topological Data Analysis" ]
We study whether local topological properties of time series, as captured via persistent homology, can serve as a reliable signal that provides complementary information for learning to forecast.
9,296
2107.09031
title_snapshot
ErivP29kYnx
ReSSL: Relational Self-Supervised Learning with Weak Augmentation
https://openreview.net/forum?id=ErivP29kYnx
[ "Mingkai Zheng", "Shan You", "Fei Wang", "Chen Qian", "Changshui Zhang", "Xiaogang Wang", "Chang Xu" ]
Poster
null
Self-supervised Learning (SSL) including the mainstream contrastive learning has achieved great success in learning visual representations without data annotations. However, most of methods mainly focus on the instance level information (\ie, the different augmented images of the same instance should have the same feat...
[ "Self-Supervised Learning" ]
null
9,291
2107.09282
title_snapshot
A9HVNx1J8Pc
Online Facility Location with Multiple Advice
https://openreview.net/forum?id=A9HVNx1J8Pc
[ "Matteo Almanza", "Flavio Chierichetti", "Silvio Lattanzi", "Alessandro Panconesi", "Giuseppe Re" ]
Poster
null
Clustering is a central topic in unsupervised learning and its online formulation has received a lot of attention in recent years. In this paper, we study the classic facility location problem in the presence of multiple machine-learned advice. We design an algorithm with provable performance guarantees such that, if t...
[ "Clustering", "Facility Location", "Online Algorithms", "Machine-Learned Advice", "Online Clustering" ]
null
9,270
null
null
BFYlnDtJSqW
Efficient Combination of Rematerialization and Offloading for Training DNNs
https://openreview.net/forum?id=BFYlnDtJSqW
[ "Olivier Beaumont", "Lionel Eyraud-Dubois", "Alena SHILOVA" ]
Poster
null
Rematerialization and offloading are two well known strategies to save memory during the training phase of deep neural networks, allowing data scientists to consider larger models, batch sizes or higher resolution data. Rematerialization trades memory for computation time, whereas Offloading trades memory for data mo...
[ "combinatorial optimization algorithms", "dynamic programming", "rematerialization", "offloading", "checkpointing", "memory constraint", "training", "deep neural networks", "feed forward backpropagation training" ]
We propose an optimized algorithm to compute a sequence of forward / backward / offload / prefetch operations on activations that optimizes training throughput of linearized DNNs under memory constraints.
9,257
null
null
sBBnfOFtPc
Locality defeats the curse of dimensionality in convolutional teacher-student scenarios
https://openreview.net/forum?id=sBBnfOFtPc
[ "Alessandro Favero", "Francesco Cagnetta", "Matthieu Wyart" ]
Poster
null
Convolutional neural networks perform a local and translationally-invariant treatment of the data: quantifying which of these two aspects is central to their success remains a challenge. We study this problem within a teacher-student framework for kernel regression, using 'convolutional' kernels inspired by the neural ...
[ "Deep learning", "Convolutional Neural Networks", "Kernel Methods", "Learning Curves" ]
We compute the learning curves of convolutional kernels in a teacher-student framework. We find that locality is key to beat the curse of dimensionality, while translational invariance is not.
9,253
2106.08619
title_snapshot
UlSjqPEkI1V
What Makes Multi-Modal Learning Better than Single (Provably)
https://openreview.net/forum?id=UlSjqPEkI1V
[ "Yu Huang", "Chenzhuang Du", "Zihui Xue", "Xuanyao Chen", "Hang Zhao", "Longbo Huang" ]
Poster
null
The world provides us with data of multiple modalities. Intuitively, models fusing data from different modalities outperform their uni-modal counterparts, since more information is aggregated. Recently, joining the success of deep learning, there is an influential line of work on deep multi-modal learning, which has re...
[ "Multi-modal learning theory" ]
We provably show what makes multi-modal outperform unimodal
9,239
2106.04538
title_snapshot
VtlGqVzja48
Identifiability in inverse reinforcement learning
https://openreview.net/forum?id=VtlGqVzja48
[ "Haoyang Cao", "Samuel N Cohen", "Lukasz Szpruch" ]
Poster
null
Inverse reinforcement learning attempts to reconstruct the reward function in a Markov decision problem, using observations of agent actions. As already observed in Russell [1998] the problem is ill-posed, and the reward function is not identifiable, even under the presence of perfect information about optimal behavior...
[ "Inverse reinforcement learning", "identifiability", "entropy regularization", "optimal control", "Markov decision problem" ]
null
9,224
2106.03498
title_snapshot
Ic9vRN3VpZ
Neo-GNNs: Neighborhood Overlap-aware Graph Neural Networks for Link Prediction
https://openreview.net/forum?id=Ic9vRN3VpZ
[ "Seongjun Yun", "Seoyoon Kim", "Junhyun Lee", "Jaewoo Kang", "Hyunwoo J. Kim" ]
Poster
null
Graph Neural Networks (GNNs) have been widely applied to various fields for learning over graph-structured data. They have shown significant improvements over traditional heuristic methods in various tasks such as node classification and graph classification. However, since GNNs heavily rely on smoothed node features r...
[ "Graph Neural Networks", "Link Prediction", "Graph Structure", "Common Neighbors" ]
null
9,219
2206.04216
title_snapshot
6wuE1-G4pu6
Optimal Underdamped Langevin MCMC Method
https://openreview.net/forum?id=6wuE1-G4pu6
[ "Zhengmian Hu", "Feihu Huang", "Heng Huang" ]
Poster
null
In the paper, we study the underdamped Langevin diffusion (ULD) with strongly-convex potential consisting of finite summation of $N$ smooth components, and propose an efficient discretization method, which requires $O(N+d^\frac{1}{3}N^\frac{2}{3}/\varepsilon^\frac{2}{3})$ gradient evaluations to achieve $\varepsilon$-e...
[ "sampling", "underdamped Langevin diffusion", "Bayesian methods" ]
We propose an optimal method for estimating ULD with sum decomposible potential function and prove that it is indeed optimal.
9,214
null
null
FM8auLVlRMo
A Faster Maximum Cardinality Matching Algorithm with Applications in Machine Learning
https://openreview.net/forum?id=FM8auLVlRMo
[ "Nathaniel Lahn", "Sharath Raghvendra", "Jiacheng Ye" ]
Poster
null
Maximum cardinality bipartite matching is an important graph optimization problem with several applications. For instance, maximum cardinality matching in a $\delta$-disc graph can be used in the computation of the bottleneck matching as well as the $\infty$-Wasserstein and the Lévy-Prokhorov distances between probabil...
[ "bipartite matching", "Wasserstein distance", "Levy-Prokhorov distance", "bottleneck matching", "unit disc graph matching" ]
null
9,198
null
null
QkljT4mrfs
Partial success in closing the gap between human and machine vision
https://openreview.net/forum?id=QkljT4mrfs
[ "Robert Geirhos", "Kantharaju Narayanappa", "Benjamin Mitzkus", "Tizian Thieringer", "Matthias Bethge", "Felix A. Wichmann", "Wieland Brendel" ]
Oral
null
A few years ago, the first CNN surpassed human performance on ImageNet. However, it soon became clear that machines lack robustness on more challenging test cases, a major obstacle towards deploying machines "in the wild" and towards obtaining better computational models of human visual perception. Here we ask: Are we ...
[ "psychophysics", "machine vision", "human vision", "out-of-distribution generalisation", "visual perception", "robustness", "human behaviour" ]
Data-rich models are closing the gap to human OOD distortion robustness and improve image-level consistency with human psychophysical data.
9,196
2106.07411
title_snapshot
2E4AT-qj3Dg
Discovering Dynamic Salient Regions for Spatio-Temporal Graph Neural Networks
https://openreview.net/forum?id=2E4AT-qj3Dg
[ "Iulia Duta", "Andrei Liviu Nicolicioiu", "Marius Leordeanu" ]
Poster
null
Graph Neural Networks are perfectly suited to capture latent interactions between various entities in the spatio-temporal domain (e.g. videos). However, when an explicit structure is not available, it is not obvious what atomic elements should be represented as nodes. Current works generally use pre-trained object dete...
[ "object-centric", "graph neural networks", "spatio-temporal" ]
Discover salient regions, correlated with objects, that are useful for visual relational processing
9,180
2009.08427
title_snapshot
a-Lbgfy9RqV
Spot the Difference: Detection of Topological Changes via Geometric Alignment
https://openreview.net/forum?id=a-Lbgfy9RqV
[ "Steffen Czolbe", "Aasa Feragen", "Oswin Krause" ]
Poster
null
Geometric alignment appears in a variety of applications, ranging from domain adaptation, optimal transport, and normalizing flows in machine learning; optical flow and learned augmentation in computer vision and deformable registration within biomedical imaging. A recurring challenge is the alignment of domains whose ...
[ "Image registration", "Geometric transformations", "Variational Autoencoder", "VAE", "Topology" ]
We use a conditional VAE to detect topological changes in images.
9,160
2106.08233
title_snapshot
lDzLzhUIwBq
Fine-grained Generalization Analysis of Inductive Matrix Completion
https://openreview.net/forum?id=lDzLzhUIwBq
[ "Antoine Ledent", "Rodrigo Alves", "Yunwen Lei", "Marius Kloft" ]
Poster
null
In this paper, we bridge the gap between the state-of-the-art theoretical results for matrix completion with the nuclear norm and their equivalent in \textit{inductive matrix completion}: (1) In the distribution-free setting, we prove bounds improving the previously best scaling of $O(rd^2)$ to $\widetilde{O}(d^{3/2}\s...
[ "Inductive Matrix Completion", "Statistical Learning Theory", "Nuclear Norm Regularisation" ]
We prove distribution-free bounds for inductive matrix completion with rate $\widetilde{O}(d^{3/2}\sqrt{r})$ and provide an inductive analogue of the weighted trace norm which brings the rate down to $\widetilde{O}(rd)$.
9,157
null
null
nzqoh6FN6sF
Estimating the Long-Term Effects of Novel Treatments
https://openreview.net/forum?id=nzqoh6FN6sF
[ "Keith Battocchi", "Eleanor Dillon", "Maggie Hei", "Greg Lewis", "Miruna Oprescu", "Vasilis Syrgkanis" ]
Poster
null
Policy makers often need to estimate the long-term effects of novel treatments, while only having historical data of older treatment options. We propose a surrogate-based approach using a long-term dataset where only past treatments were administered and a short-term dataset where novel treatments have been administere...
[ "long-term effects", "dynamic treatment effects", "surrogates", "high-dimensional", "double machine learning" ]
Estimating long-term causal effects in high-dimensions with short-term surrogates from historical data with dynamic historical treatment policies
9,156
2103.08390
title_snapshot
O8Ffv3aRJr
Unbalanced Optimal Transport through Non-negative Penalized Linear Regression
https://openreview.net/forum?id=O8Ffv3aRJr
[ "Laetitia Chapel", "Rémi Flamary", "Haoran Wu", "Cédric Févotte", "Gilles Gasso" ]
Poster
null
This paper addresses the problem of Unbalanced Optimal Transport (UOT) in which the marginal conditions are relaxed (using weighted penalties in lieu of equality) and no additional regularization is enforced on the OT plan. In this context, we show that the corresponding optimization problem can be reformulated as a no...
[ "Optimal Transport", "Unbalanced Optimal Transport", "Penalized Linear Regression", "Lasso", "MM algorithms" ]
We propose a reformulation of the Exact Unbalanced Optimal Transport as a Non-negative Linear Regression problem which allows us to devise new algorithms such as the first regularization path for OT and multiplicative algorithms.
9,155
2106.04145
title_snapshot
sl_0rQmHxQk
Sparse Quadratic Optimisation over the Stiefel Manifold with Application to Permutation Synchronisation
https://openreview.net/forum?id=sl_0rQmHxQk
[ "Florian Bernard", "Daniel Cremers", "Anders Johan Thunberg" ]
Poster
null
We address the non-convex optimisation problem of finding a sparse matrix on the Stiefel manifold (matrices with mutually orthogonal columns of unit length) that maximises (or minimises) a quadratic objective function. Optimisation problems on the Stiefel manifold occur for example in spectral relaxations of various co...
[ "Stiefel manifold", "quadratic optimisation", "permutation synchronisation", "sparsity", "multi-matching", "correspondence problems", "manifold optimisation", "QR decomposition", "orthogonal iteration algorithm" ]
A method for finding a globally optimal solution of a quadratic objective function over the Stiefel manifold that is sparse.
9,144
2110.00053
title_snapshot
XHHxE-KOK7
Reinforcement Learning in Reward-Mixing MDPs
https://openreview.net/forum?id=XHHxE-KOK7
[ "Jeongyeol Kwon", "Yonathan Efroni", "Constantine Caramanis", "Shie Mannor" ]
Poster
null
Learning a near optimal policy in a partially observable system remains an elusive challenge in contemporary reinforcement learning. In this work, we consider episodic reinforcement learning in a reward-mixing Markov decision process (MDP). There, a reward function is drawn from one of $M$ possible reward models at the...
[ "Reinforcement Learning", "Partially Observable", "Sample-Complexity", "Method-of-Moments", "Latent Confounders", "Mixture Models" ]
Efficient exploration in Reward-Mixing MDPs without assumptions when $M=2$
9,139
2110.03743
title_snapshot
Oeb2LbHAfJ4
SketchGen: Generating Constrained CAD Sketches
https://openreview.net/forum?id=Oeb2LbHAfJ4
[ "Wamiq Reyaz Para", "Shariq Farooq Bhat", "Paul Guerrero", "Tom Kelly", "Niloy Mitra", "Leonidas Guibas", "Peter Wonka" ]
Poster
null
Computer-aided design (CAD) is the most widely used modeling approach for technical design. The typical starting point in these designs is 2D sketches which can later be extruded and combined to obtain complex three-dimensional assemblies. Such sketches are typically composed of parametric primitives, such as points, l...
[ "CAD", "transformers", "generative modelling", "layouts" ]
A generative model to generate constrained CAD models ready for editing.
9,136
2106.02711
title_snapshot
StKuQ0-dltN
Double/Debiased Machine Learning for Dynamic Treatment Effects
https://openreview.net/forum?id=StKuQ0-dltN
[ "Greg Lewis", "Vasilis Syrgkanis" ]
Poster
null
We consider the estimation of treatment effects in settings when multiple treatments are assigned over time and treatments can have a causal effect on future outcomes. We propose an extension of the double/debiased machine learning framework to estimate the dynamic effects of treatments and apply it to a concrete linea...
[ "dynamic treatment regime", "high-dimensional", "treatment effects", "double machine learning" ]
High dimensional causal inference in the dynamic treatment regime via Neyman orthogonality
9,135
null
null
uTqvj8i3xv
Functional Regularization for Reinforcement Learning via Learned Fourier Features
https://openreview.net/forum?id=uTqvj8i3xv
[ "Alexander Cong Li", "Deepak Pathak" ]
Poster
null
We propose a simple architecture for deep reinforcement learning by embedding inputs into a learned Fourier basis and show that it improves the sample efficiency of both state-based and image-based RL. We perform infinite-width analysis of our architecture using the Neural Tangent Kernel and theoretically show that tun...
[ "reinforcement learning", "deep learning", "neural tangent kernel", "regularization", "architecture" ]
We propose a new architecture that can directly control how much it fits high and low target frequencies, and use this to improve off-policy reinforcement learning.
9,134
2112.03257
title_snapshot
4orlVaC95Bo
Task-Agnostic Undesirable Feature Deactivation Using Out-of-Distribution Data
https://openreview.net/forum?id=4orlVaC95Bo
[ "Dongmin Park", "Hwanjun Song", "Minseok Kim", "Jae-Gil Lee" ]
Poster
null
A deep neural network (DNN) has achieved great success in many machine learning tasks by virtue of its high expressive power. However, its prediction can be easily biased to undesirable features, which are not essential for solving the target task and are even imperceptible to a human, thereby resulting in poor general...
[ "Out-of-Distribution", "Regularization", "Task-Agnostic" ]
We propose a novel regularizer that deactivates all undesirable features using OOD examples in the feature extraction layer.
9,132
null
null
JbqW3KmmE6
Multi-Facet Clustering Variational Autoencoders
https://openreview.net/forum?id=JbqW3KmmE6
[ "Fabian Falck", "Haoting Zhang", "Matthew Willetts", "George Nicholson", "Christopher Yau", "Christopher C. Holmes" ]
Poster
null
Work in deep clustering focuses on finding a single partition of data. However, high-dimensional data, such as images, typically feature multiple interesting characteristics one could cluster over. For example, images of objects against a background could be clustered over the shape of the object and separately by the ...
[ "multi-facet clustering", "deep clustering", "variational autoencoders", "deep generative models" ]
We present theoretical and empirical results for a novel class of variational autoencoders for multi-facet clustering.
9,110
2106.05241
title_snapshot
CtaDl9L0bIQ
Group Equivariant Subsampling
https://openreview.net/forum?id=CtaDl9L0bIQ
[ "Jin Xu", "Hyunjik Kim", "Tom Rainforth", "Yee Whye Teh" ]
Poster
null
Subsampling is used in convolutional neural networks (CNNs) in the form of pooling or strided convolutions, to reduce the spatial dimensions of feature maps and to allow the receptive fields to grow exponentially with depth. However, it is known that such subsampling operations are not translation equivariant, unlike c...
[ "group equivariance", "group invariance", "convolutional architecture", "autoencoders", "representation learning" ]
We propose group equivariant subsampling with applications to equivariant representation learning of objects.
9,072
2106.05886
title_snapshot
we8d1FjibAc
Robust and Fully-Dynamic Coreset for Continuous-and-Bounded Learning (With Outliers) Problems
https://openreview.net/forum?id=we8d1FjibAc
[ "Zixiu Wang", "Yiwen Guo", "Hu Ding" ]
Spotlight
null
In many machine learning tasks, a common approach for dealing with large-scale data is to build a small summary, {\em e.g.,} coreset, that can efficiently represent the original input. However, real-world datasets usually contain outliers and most existing coreset construction methods are not resilient against outlier...
[ "coreset", "Continuous-and-Bounded learning", "outliers", "dynamic setting" ]
We provide a robust coreset construction method for continuous-and-bounded optimization problems
9,070
2107.00068
title_snapshot
1_gaHBaRYt
Fast Federated Learning in the Presence of Arbitrary Device Unavailability
https://openreview.net/forum?id=1_gaHBaRYt
[ "Xinran Gu", "Kaixuan Huang", "Jingzhao Zhang", "Longbo Huang" ]
Poster
null
Federated learning (FL) coordinates with numerous heterogeneous devices to collaboratively train a shared model while preserving user privacy. Despite its multiple advantages, FL faces new challenges. One challenge arises when devices drop out of the training process. In this case, the convergence of popular FL algorit...
[ "Federated Learning", "Distributed Optimization" ]
We study federated learning algorithms under arbitrary device unavailability and show our proposed MIFA avoids excessive latency induced by inactive devices and achieves minimax optimal convergence rates.
9,059
2106.04159
title_snapshot
FFtcBBVIg1T
Overlapping Spaces for Compact Graph Representations
https://openreview.net/forum?id=FFtcBBVIg1T
[ "Kirill Sergeevich Shevkunov", "Liudmila Prokhorenkova" ]
Poster
null
Various non-trivial spaces are becoming popular for embedding structured data such as graphs, texts, or images. Following spherical and hyperbolic spaces, more general product spaces have been proposed. However, searching for the best configuration of a product space is a resource-intensive procedure, which reduces the...
[ "graph embedding", "hyperbolic space", "product space", "overlapping space" ]
A novel approach to combine metric distances and similarity measures to achieve better vector representations of structured data.
9,055
2007.02445
title_snapshot
t5-Mszu1UkO
Information Directed Reward Learning for Reinforcement Learning
https://openreview.net/forum?id=t5-Mszu1UkO
[ "David Lindner", "Matteo Turchetta", "Sebastian Tschiatschek", "Kamil Ciosek", "Andreas Krause" ]
Poster
null
For many reinforcement learning (RL) applications, specifying a reward is difficult. In this paper, we consider an RL setting where the agent can obtain information about the reward only by querying an expert that can, for example, evaluate individual states or provide binary preferences over trajectories. From such ex...
[ "reward learning", "reinforcement learning", "active learning", "preference learning", "human feedback" ]
Improving sample efficiency of active reward learning by focusing on learning a good policy.
9,052
2102.12466
title_snapshot
KmPSe18DGLs
Statistical Undecidability in Linear, Non-Gaussian Causal Models in the Presence of Latent Confounders
https://openreview.net/forum?id=KmPSe18DGLs
[ "Konstantin Genin" ]
Poster
null
If causal relationships are linear and acyclic and noise terms are independent and Gaussian, causal orientation is not identified from observational data --- even if faithfulness is satisfied (Spirtes et al., 2002). Shimizu et al. (2006) showed that acyclic, linear, {\bf non}-Gaussian (LiNGAM) causal models {\em are} i...
[ "causal discovery", "linear non-gaussian models", "faithfulness", "confounding", "topology" ]
We refine causal identifiability results in the linear non-Gaussian setting, allowing for the presence of latent confounders.
9,041
null
null
dsevgAwUH4m
Linear Convergence of Gradient Methods for Estimating Structured Transition Matrices in High-dimensional Vector Autoregressive Models
https://openreview.net/forum?id=dsevgAwUH4m
[ "Xiao Lv", "Wei Cui", "Yulong Liu" ]
Poster
null
In this paper, we present non-asymptotic optimization guarantees of gradient descent methods for estimating structured transition matrices in high-dimensional vector autoregressive (VAR) models. We adopt the projected gradient descent (PGD) for single-structured transition matrices and the alternating projected gradien...
[ "Vector autoregressive models", "dirty statistical models", "linear convergence rate", "non-asymptotic analysis" ]
null
9,040
null
null
NFurmj-rIWe
Dynamic population-based meta-learning for multi-agent communication with natural language
https://openreview.net/forum?id=NFurmj-rIWe
[ "Abhinav Gupta", "Marc Lanctot", "Angeliki Lazaridou" ]
Poster
null
In this work, our goal is to train agents that can coordinate with seen, unseen as well as human partners in a multi-agent communication environment involving natural language. Previous work using a single set of agents has shown great progress in generalizing to known partners, however it struggles when coordinating w...
[ "emergent communication", "meta-learning", "population-based methods", "multi-agent reinforcement learning", "few-shot generalization", "human-AI coordination" ]
We propose a dynamic population-based meta-learning method to train agents in a cooperative multi-agent communication environment.
9,027
2110.14241
title_snapshot
zImiB39pyUL
BooVAE: Boosting Approach for Continual Learning of VAE
https://openreview.net/forum?id=zImiB39pyUL
[ "Evgenii Egorov", "Anna Kuzina", "Evgeny Burnaev" ]
Poster
null
Variational autoencoder (VAE) is a deep generative model for unsupervised learning, allowing to encode observations into the meaningful latent space. VAE is prone to catastrophic forgetting when tasks arrive sequentially, and only the data for the current one is available. We address this problem of continual learning ...
[ "Continual Learning", "Variational Autoencoder", "VAE", "Catastrophic Forgetting", "Entropy regularisation", "Boosting", "Functional regularisation" ]
We propose to used learnable prior to alleviate catastrophic forgetting in continual learning of VAE
9,024
1908.11853
title_snapshot
k9iBo3RmCFd
Can we have it all? On the Trade-off between Spatial and Adversarial Robustness of Neural Networks
https://openreview.net/forum?id=k9iBo3RmCFd
[ "Sandesh Kamath", "Amit Deshpande", "K Venkata Subrahmanyam", "Vineeth N. Balasubramanian" ]
Poster
null
(Non-)robustness of neural networks to small, adversarial pixel-wise perturbations, and as more recently shown, to even random spatial transformations (e.g., translations, rotations) entreats both theoretical and empirical understanding. Spatial robustness to random translations and rotations is commonly attained via e...
[ "Adversarial robustness", "spatial robustness", "invariance", "curriculum learning" ]
We show a trade-off between spatial and adversarial robustness theoretically and empirically. And propose a curriculum based training to achieve both of them simultaneously.
9,022
2002.11318
title_snapshot
ags1UxpXAl
Powerpropagation: A sparsity inducing weight reparameterisation
https://openreview.net/forum?id=ags1UxpXAl
[ "Jonathan Schwarz", "Siddhant Jayakumar", "Razvan Pascanu", "Peter E. Latham", "Yee Whye Teh" ]
Poster
null
The training of sparse neural networks is becoming an increasingly important tool for reducing the computational footprint of models at training and evaluation, as well enabling the effective scaling up of models. Whereas much work over the years has been dedicated to specialised pruning techniques, little attention ha...
[ "Sparse Neural Networks", "Continual Learning" ]
null
9,020
2110.00296
title_snapshot
EocGDCLaw-d
Active clustering for labeling training data
https://openreview.net/forum?id=EocGDCLaw-d
[ "Quentin Lutz", "Elie De Panafieu", "Maya Stein", "Alex Scott" ]
Poster
null
Gathering training data is a key step of any supervised learning task, and it is both critical and expensive. Critical, because the quantity and quality of the training data has a high impact on the performance of the learned function. Expensive, because most practical cases rely on humans-in-the-loop to label the data...
[ "clustering", "active learning", "annotation", "training data" ]
Organise human experts to build a training data set for classification using Boolean pairwise queries.
8,993
2110.14521
title_snapshot
-OrwaD3bG91
What Matters for Adversarial Imitation Learning?
https://openreview.net/forum?id=-OrwaD3bG91
[ "Manu Orsini", "Anton Raichuk", "Leonard Hussenot", "Damien Vincent", "Robert Dadashi", "Sertan Girgin", "Matthieu Geist", "Olivier Bachem", "Olivier Pietquin", "Marcin Andrychowicz" ]
Poster
null
Adversarial imitation learning has become a popular framework for imitation in continuous control. Over the years, several variations of its components were proposed to enhance the performance of the learned policies as well as the sample complexity of the algorithm. In practice, these choices are rarely tested all tog...
[ "imitation learning", "GAIL", "continuous control" ]
a large-scale study of adversarial imitation learning algorithms
8,981
2106.00672
title_snapshot
DyE8hmj2dse
A Theoretical Analysis of Fine-tuning with Linear Teachers
https://openreview.net/forum?id=DyE8hmj2dse
[ "Gal Shachaf", "Alon Brutzkus", "Amir Globerson" ]
Poster
null
Fine-tuning is a common practice in deep learning, achieving excellent generalization results on downstream tasks using relatively little training data. Although widely used in practice, it is not well understood theoretically. Here we analyze the sample complexity of this scheme for regression with linear teachers in ...
[ "deep learning theory", "learning theory", "transfer learning", "fine-tuning" ]
null
8,956
2107.01641
title_snapshot
ephWA7KaWmD
GRIN: Generative Relation and Intention Network for Multi-agent Trajectory Prediction
https://openreview.net/forum?id=ephWA7KaWmD
[ "Longyuan Li", "Jian Yao", "Li Kevin Wenliang", "Tong He", "Tianjun Xiao", "Junchi Yan", "David Wipf", "Zheng Zhang" ]
Poster
null
Learning the distribution of future trajectories conditioned on the past is a crucial problem for understanding multi-agent systems. This is challenging because humans make decisions based on complex social relations and personal intents, resulting in highly complex uncertainties over trajectories. To address this prob...
[ "Trajectory prediction", "variational Autoencoder", "deep learning", "probabilistic" ]
We build a model that disentangles social relations and human intents to generate multimodal future trajectories of multi-agent systems.
8,952
null
null
hyJKKIhfxxT
VAST: Value Function Factorization with Variable Agent Sub-Teams
https://openreview.net/forum?id=hyJKKIhfxxT
[ "Thomy Phan", "Fabian Ritz", "Lenz Belzner", "Philipp Altmann", "Thomas Gabor", "Claudia Linnhoff-Popien" ]
Poster
null
Value function factorization (VFF) is a popular approach to cooperative multi-agent reinforcement learning in order to learn local value functions from global rewards. However, state-of-the-art VFF is limited to a handful of agents in most domains. We hypothesize that this is due to the flat factorization scheme, where...
[ "Multi-Agent Learning", "Reinforcement Learning", "Value Function Factorization" ]
We propose a hierarchical value function factorization approach based on variable agent sub-teams.
8,944
null
null
ZYJ1r6sStU
On the interplay between data structure and loss function in classification problems
https://openreview.net/forum?id=ZYJ1r6sStU
[ "Stéphane d'Ascoli", "Marylou Gabrié", "Levent Sagun", "Giulio Biroli" ]
Poster
null
One of the central features of modern machine learning models, including deep neural networks, is their generalization ability on structured data in the over-parametrized regime. In this work, we consider an analytically solvable setup to investigate how properties of data impact learning in classification problems, a...
[ "loss function", "generalization", "overparametrization", "double descent", "random features" ]
We analyze the interplay between data structure and loss function in random feature classification problems
8,929
2103.05524
title_snapshot
Aa5oPXc_1IV
Gradient Descent on Two-layer Nets: Margin Maximization and Simplicity Bias
https://openreview.net/forum?id=Aa5oPXc_1IV
[ "Kaifeng Lyu", "Zhiyuan Li", "Runzhe Wang", "Sanjeev Arora" ]
Poster
null
The generalization mystery of overparametrized deep nets has motivated efforts to understand how gradient descent (GD) converges to low-loss solutions that generalize well. Real-life neural networks are initialized from small random values and trained with cross-entropy loss for classification (unlike the "lazy" or "NT...
[ "margin maximization", "gradient flow", "gradient descent", "implicit bias", "linearly separable", "two-layer neural networks" ]
We prove that gradient flow biases two-layer nets to max-margin classifiers on linearly separable data with logistic loss and small initialization, although the margin can be only locally optimal by making simple changes to data assumptions.
8,917
2110.13905
title_snapshot
7m6qvNqFjr
Fast rates for prediction with limited expert advice
https://openreview.net/forum?id=7m6qvNqFjr
[ "El Mehdi Saad", "Gilles Blanchard" ]
Poster
null
We investigate the problem of minimizing the excess generalization error with respect to the best expert prediction in a finite family in the stochastic setting, under limited access to information. We consider that the learner has only access to a limited number of expert advices per training round, as well as for pr...
[ "Online Learning", "prediction with expert advice", "high probability bounds", "budgeted learning" ]
We study the impact of restricted access to information on the generalization error in the setting of prediction with expert advice.
8,916
2110.14485
title_snapshot
eQ7Kh-QeWnO
DualNet: Continual Learning, Fast and Slow
https://openreview.net/forum?id=eQ7Kh-QeWnO
[ "Quang Pham", "Chenghao Liu", "Steven HOI" ]
Poster
null
According to Complementary Learning Systems (CLS) theory~\cite{mcclelland1995there} in neuroscience, humans do effective \emph{continual learning} through two complementary systems: a fast learning system centered on the hippocampus for rapid learning of the specifics and individual experiences, and a slow learning sys...
[ "Continual learning", "fast and slow learning" ]
A novel continual learning paradigm of fast and slow learning.
8,910
2110.00175
title_snapshot
F1D8buayXQT
Self-Supervised Representation Learning on Neural Network Weights for Model Characteristic Prediction
https://openreview.net/forum?id=F1D8buayXQT
[ "Konstantin Schürholt", "Dimche Kostadinov", "Damian Borth" ]
Poster
null
Self-Supervised Learning (SSL) has been shown to learn useful and information-preserving representations. Neural Networks (NNs) are widely applied, yet their weight space is still not fully understood. Therefore, we propose to use SSL to learn hyper-representations of the weights of populations of NNs. To that end, we ...
[ "Representation Learning", "Self-Supervised Learning", "Weight Space", "Parameter Space", "Augmentation", "Model Zoos" ]
This paper proposes to learn self-supervised representations of the weights of populations of NN models using novel data augmentations and an adapted transformer architecture.
8,896
2110.15288
title_judge
7U7JxTiL8gz
Modular Gaussian Processes for Transfer Learning
https://openreview.net/forum?id=7U7JxTiL8gz
[ "Pablo Moreno-Muñoz", "Antonio Artés", "Mauricio A Álvarez" ]
Poster
null
We present a framework for transfer learning based on modular variational Gaussian processes (GP). We develop a module-based method that having a dictionary of well fitted GPs, each model being characterised by its hyperparameters, pseudo-inputs and their corresponding posterior densities, one could build ensemble GP m...
[ "Gaussian Processes" ]
Framework for transfer learning based on modular variational Gaussian processes
8,888
2110.13515
title_snapshot
ykN3tbJ0qmX
Collapsed Variational Bounds for Bayesian Neural Networks
https://openreview.net/forum?id=ykN3tbJ0qmX
[ "Marcin B. Tomczak", "Siddharth Swaroop", "Andrew Y. K. Foong", "Richard E Turner" ]
Poster
null
Recent interest in learning large variational Bayesian Neural Networks (BNNs) has been partly hampered by poor predictive performance caused by underfitting, and their performance is known to be very sensitive to the prior over weights. Current practice often fixes the prior parameters to standard values or tunes them ...
[ "Bayesian Neural Networks", "BNNs", "Bayesian Deep Learning", "variational inference", "VI", "underfitting", "collapsed ELBO", "overpruning" ]
We derive a family of collapsed, tighter ELBOs to learn variational posteriors over weights in Bayesian Neural Networks.
8,882
null
null
F9HNBbytcqT
Distributed Machine Learning with Sparse Heterogeneous Data
https://openreview.net/forum?id=F9HNBbytcqT
[ "Dominic Richards", "Sahand Negahban", "Patrick Rebeschini" ]
Poster
null
Motivated by distributed machine learning settings such as Federated Learning, we consider the problem of fitting a statistical model across a distributed collection of heterogeneous data sets whose similarity structure is encoded by a graph topology. Precisely, we analyse the case where each node is associated with fi...
[ "Distributed Machine Learning", "Heterogeneous Data", "Federated Machine Learning" ]
null
8,875
1912.01417
title_snapshot
anxHcl9_sE
On the Convergence of Prior-Guided Zeroth-Order Optimization Algorithms
https://openreview.net/forum?id=anxHcl9_sE
[ "Shuyu Cheng", "Guoqiang Wu", "Jun Zhu" ]
Poster
null
Zeroth-order (ZO) optimization is widely used to handle challenging tasks, such as query-based black-box adversarial attacks and reinforcement learning. Various attempts have been made to integrate prior information into the gradient estimation procedure based on finite differences, with promising empirical results. Ho...
[ "zeroth-order optimization", "gradient-free optimization", "evolutionary strategies", "convex optimization", "randomized methods" ]
We conduct convergence analysis on frameworks of ZO algorithms including existing prior-guided random gradient-free methods and a new prior-guided accelerated random search method.
8,873
2107.10110
title_snapshot
YN4TMf3sv52
Exploring Forensic Dental Identification with Deep Learning
https://openreview.net/forum?id=YN4TMf3sv52
[ "Yuan Liang", "Weikun Han", "Liang Qiu", "Chen Wu", "Yiting Shao", "Kun Wang", "Lei He" ]
Poster
null
Dental forensic identification targets to identify persons with dental traces. The task is vital for the investigation of criminal scenes and mass disasters because of the resistance of dental structures and the wide-existence of dental imaging. However, no widely accepted automated solution is available for this labo...
[ "person identification", "medical imaging", "forensics" ]
The first in-depth exploration of deep learning for the forensic dental identification.
8,867
null
null
JOOsoL_J6Fc
Stability & Generalisation of Gradient Descent for Shallow Neural Networks without the Neural Tangent Kernel
https://openreview.net/forum?id=JOOsoL_J6Fc
[ "Dominic Richards", "Ilja Kuzborskij" ]
Poster
null
We revisit on-average algorithmic stability of Gradient Descent (GD) for training overparameterised shallow neural networks and prove new generalisation and excess risk bounds without the Neural Tangent Kernel (NTK) or Polyak-Łojasiewicz (PL) assumptions. In particular, we show oracle type bounds whic...
[ "Gradient Descent", "Neural Networks", "Stability", "Generalisation", "Implicit Regularisation", "Neural Tangent Kernel" ]
null
8,866
2107.12723
title_snapshot
ypj3xKoRfmr
Credal Self-Supervised Learning
https://openreview.net/forum?id=ypj3xKoRfmr
[ "Julian Lienen", "Eyke Hüllermeier" ]
Poster
null
Self-training is an effective approach to semi-supervised learning. The key idea is to let the learner itself iteratively generate "pseudo-supervision" for unlabeled instances based on its current hypothesis. In combination with consistency regularization, pseudo-labeling has shown promising performance in various doma...
[ "semi-supervised learning", "self-training", "superset learning", "credal sets", "label relaxation", "pseudo-labeling" ]
We propose a novel semi-supervised learning method that uses credal sets as pseudo-labels for self-supervision.
8,864
2106.11853
title_snapshot
YIyYkoJX2eA
Learning to Draw: Emergent Communication through Sketching
https://openreview.net/forum?id=YIyYkoJX2eA
[ "Daniela Mihai", "Jonathon Hare" ]
Oral
null
Evidence that visual communication preceded written language and provided a basis for it goes back to prehistory, in forms such as cave and rock paintings depicting traces of our distant ancestors. Emergent communication research has sought to explore how agents can learn to communicate in order to collaboratively solv...
[ "Emergent Communication", "Sketching", "Drawing", "Perceptual Loss", "Interpretability", "Visual Communication" ]
We use self-supervised play to train artificial agents to communicate by drawing and then show that with the appropriate inductive bias a human can successfully play the same games with the pretrained drawing agent.
8,863
2106.02067
title_snapshot
90M-91IZ0JC
Distilling Robust and Non-Robust Features in Adversarial Examples by Information Bottleneck
https://openreview.net/forum?id=90M-91IZ0JC
[ "Junho Kim", "Byung-Kwan Lee", "Yong Man Ro" ]
Poster
null
Adversarial examples, generated by carefully crafted perturbation, have attracted considerable attention in research fields. Recent works have argued that the existence of the robust and non-robust features is a primary cause of the adversarial examples, and investigated their internal interactions in the feature space...
[ "Adversarial Examples", "Information Bottleneck", "Feature Disentanglement", "Feature Interpretation" ]
We present a way of distilling the robust and non-robust features in adversarial examples, using Information Bottleneck.
8,856
2204.02735
title_snapshot
3-GCM92yaB3
The Unbalanced Gromov Wasserstein Distance: Conic Formulation and Relaxation
https://openreview.net/forum?id=3-GCM92yaB3
[ "Thibault Sejourne", "François-Xavier Vialard", "Gabriel Peyré" ]
Poster
null
Comparing metric measure spaces (i.e. a metric space endowed with a probability distribution) is at the heart of many machine learning problems. The most popular distance between such metric measure spaces is the Gromov-Wasserstein (GW) distance, which is the solution of a quadratic assignment problem. The GW distance ...
[ "Optimal transport", "Quadratic assignment problem", "Gromov-Wasserstein" ]
We propose a generalization of the Gromov-Wasserstein distance to unbalanced input data with a GPU-friendly algorithm.
8,855
2009.04266
title_snapshot
_RSgXL8gNnx
Batch Normalization Orthogonalizes Representations in Deep Random Networks
https://openreview.net/forum?id=_RSgXL8gNnx
[ "Hadi Daneshmand", "Amir Joudaki", "Francis Bach" ]
Spotlight
null
This paper underlines an elegant property of batch-normalization (BN): Successive batch normalizations with random linear updates make samples increasingly orthogonal. We establish a non-asymptotic characterization of the interplay between depth, width, and the orthogonality of deep representations. More precisely, we ...
[ "Batch normalization", "Theory of deep neural networks", "Markov chains", "Random Neural Networks", "Optimization for neural networks" ]
We prove that successive batch normalizations, together with random linear layers, incrementally orthogonalize representations of samples.
8,850
2106.03970
title_snapshot
0NXUSlb6oEu
Improving Robustness using Generated Data
https://openreview.net/forum?id=0NXUSlb6oEu
[ "Sven Gowal", "Sylvestre-Alvise Rebuffi", "Olivia Wiles", "Florian Stimberg", "Dan Andrei Calian", "Timothy Mann" ]
Poster
null
Recent work argues that robust training requires substantially larger datasets than those required for standard classification. On CIFAR-10 and CIFAR-100, this translates into a sizable robust-accuracy gap between models trained solely on data from the original training set and those trained with additional data extrac...
[ "adversarial", "robustness", "generative" ]
We identity conditions under which incorporating generated data can improve robustness to adversarial examples, and demonstrate that it is possible to significantly improve upon the state-of-the-art without the need for external data.
8,839
2110.09468
title_snapshot
kAm9By0R5ME
A Causal Lens for Controllable Text Generation
https://openreview.net/forum?id=kAm9By0R5ME
[ "Zhiting Hu", "Li Erran Li" ]
Poster
null
Controllable text generation concerns two fundamental tasks of wide applications, namely generating text of given attributes (i.e., attribute-conditional generation), and minimally editing existing text to possess desired attributes (i.e., text attribute transfer). Extensive prior work has largely studied the two probl...
[ "causal inference", "controllable text generation", "text style transfer", "natural language processing" ]
The first unified causal framework for controllable text generation that unifies two fundamental tasks and mitigates generation biases
8,838
2201.09119
title_snapshot
i_Q1yrOegLY
Revisiting Deep Learning Models for Tabular Data
https://openreview.net/forum?id=i_Q1yrOegLY
[ "Yury Gorishniy", "Ivan Rubachev", "Valentin Khrulkov", "Artem Babenko" ]
Poster
null
The existing literature on deep learning for tabular data proposes a wide range of novel architectures and reports competitive results on various datasets. However, the proposed models are usually not properly compared to each other and existing works often use different benchmarks and experiment protocols. As a result...
[ "tabular data", "architecture", "DNN" ]
Compared many DL models for tabular data, identified a strong baseline (ResNet) and proposed a powerful Transformer-based model.
8,834
2106.11959
title_snapshot
pX7gwTNljqa
Non-Gaussian Gaussian Processes for Few-Shot Regression
https://openreview.net/forum?id=pX7gwTNljqa
[ "Marcin Sendera", "Jacek Tabor", "Aleksandra Nowak", "Andrzej Bedychaj", "Massimiliano Patacchiola", "Tomasz Trzcinski", "Przemysław Spurek", "Maciej Zieba" ]
Poster
null
Gaussian Processes (GPs) have been widely used in machine learning to model distributions over functions, with applications including multi-modal regression, time-series prediction, and few-shot learning. GPs are particularly useful in the last application since they rely on Normal distributions and enable closed-form ...
[ "Meta-Learning", "Few-Shot Learning", "Few-Shot Regression", "Normalizing Flows", "Gaussian Processes" ]
In this work we address the limitations of Gaussian Processes in Few-Shot Regression by leveraging the flexibility of Normalizing Flows to modulate the posterior predictive distribution.
8,826
2110.13561
title_snapshot
vERYhbX_6Y
Constrained Optimization to Train Neural Networks on Critical and Under-Represented Classes
https://openreview.net/forum?id=vERYhbX_6Y
[ "Sara Sangalli", "Ertunc Erdil", "Andeas Hötker", "Olivio F Donati", "Ender Konukoglu" ]
Poster
null
Deep neural networks (DNNs) are notorious for making more mistakes for the classes that have substantially fewer samples than the others during training. Such class imbalance is ubiquitous in clinical applications and very crucial to handle because the classes with fewer samples most often correspond to critical cases ...
[ "Constrained optimization", "class imbalance", "augmented Lagrangian method" ]
null
8,820
2102.12894
title_snapshot
F-H4oe3MXXI
Denoising Normalizing Flow
https://openreview.net/forum?id=F-H4oe3MXXI
[ "Christian Horvat", "Jean-Pascal Pfister" ]
Poster
null
Normalizing flows (NF) are expressive as well as tractable density estimation methods whenever the support of the density is diffeomorphic to the entire data-space. However, real-world data sets typically live on (or very close to) low-dimensional manifolds thereby challenging the applicability of standard NF on real-w...
[ "Normalizing Flow", "Manifold Learning", "Density Estimation" ]
We propose an NF for simultaneous manifold and density-on-manifold learning.
8,819
null
null
_nRSyha2SP
Latent Execution for Neural Program Synthesis Beyond Domain-Specific Languages
https://openreview.net/forum?id=_nRSyha2SP
[ "Xinyun Chen", "Dawn Song", "Yuandong Tian" ]
Poster
null
Program synthesis from input-output (IO) examples has been a long-standing challenge. While recent works demonstrated limited success on domain-specific languages (DSL), it remains highly challenging to apply them to real-world programming languages, such as C. Due to complicated syntax and token variation, there are t...
[ "neural program synthesis", "program execution" ]
null
8,812
2107.00101
title_judge
ot2ORiBqTa1
Going Beyond Linear Transformers with Recurrent Fast Weight Programmers
https://openreview.net/forum?id=ot2ORiBqTa1
[ "Kazuki Irie", "Imanol Schlag", "Róbert Csordás", "Jürgen Schmidhuber" ]
Poster
null
Transformers with linearised attention (''linear Transformers'') have demonstrated the practical scalability and effectiveness of outer product-based Fast Weight Programmers (FWPs) from the '90s. However, the original FWP formulation is more general than the one of linear Transformers: a slow neural network (NN) contin...
[ "Transformers", "memory augmented recurrent neural networks", "fast weight programmers" ]
We augment linear Transformers with recurrent connections.
8,811
2106.06295
title_snapshot
vMWHOumNj5
A Unified Approach to Fair Online Learning via Blackwell Approachability
https://openreview.net/forum?id=vMWHOumNj5
[ "Evgenii E Chzhen", "Christophe Giraud", "Gilles Stoltz" ]
Spotlight
null
We provide a setting and a general approach to fair online learning with stochastic sensitive and non-sensitive contexts. The setting is a repeated game between the Player and Nature, where at each stage both pick actions based on the contexts. Inspired by the notion of unawareness, we assume that the Player can only a...
[ "online learning", "fairness", "Blackwell approachability", "calibration" ]
We provide a general approachability-theorem based tool to tacke faire online adversarial learning (with stochastic contexts) and illustrate its application by working out several examples.
8,801
2106.12242
title_snapshot
Esd7tGH3Spl
Evaluating Efficient Performance Estimators of Neural Architectures
https://openreview.net/forum?id=Esd7tGH3Spl
[ "Xuefei Ning", "Changcheng Tang", "Wenshuo Li", "Zixuan Zhou", "Shuang Liang", "Huazhong Yang", "Yu Wang" ]
Poster
null
Conducting efficient performance estimations of neural architectures is a major challenge in neural architecture search (NAS). To reduce the architecture training costs in NAS, one-shot estimators (OSEs) amortize the architecture training costs by sharing the parameters of one supernet between all architectures. Recent...
[ "Neural architecture search (NAS)", "Parameter sharing", "One-shot NAS", "Zero-shot NAS" ]
We conduct a comprehensive empirical study on how and why the one-shot / zero-shot estimators in NAS have biases & variances.
8,797
2008.03064
title_snapshot
zlhpIYub2d0
Variational Automatic Curriculum Learning for Sparse-Reward Cooperative Multi-Agent Problems
https://openreview.net/forum?id=zlhpIYub2d0
[ "Jiayu Chen", "Yuanxin Zhang", "Yuanfan Xu", "Huimin Ma", "Huazhong Yang", "Jiaming Song", "Yu Wang", "Yi Wu" ]
Poster
null
We introduce an automatic curriculum algorithm, Variational Automatic Curriculum Learning (VACL), for solving challenging goal-conditioned cooperative multi-agent reinforcement learning problems. We motivate our curriculum learning paradigm through a variational perspective, where the learning objective can be decompos...
[ "multi-agent reinforcement learning", "curriculum learning", "variational inference" ]
We introduce a curriculum learning algorithm for solving goal-conditioned cooperative multi-agent reinforcement learning problems.
8,794
2111.04613
title_snapshot
-_D-ss8su3
Novel Upper Bounds for the Constrained Most Probable Explanation Task
https://openreview.net/forum?id=-_D-ss8su3
[ "Tahrima Rahman", "Sara Rouhani", "Vibhav Giridhar Gogate" ]
Poster
null
We propose several schemes for upper bounding the optimal value of the constrained most probable explanation (CMPE) problem. Given a set of discrete random variables, two probabilistic graphical models defined over them and a real number $q$, this problem involves finding an assignment of values to all the variables s...
[ "Discrete Optimization", "Probabilistic Graphical Models", "Constrained Most Probable Explanation", "Explainable AI" ]
A novel method that integrates fast knapsack algorithms, mini buckets and Lagrange relaxations and decompositions to yield upper bounds on the optimal value of a hard discrete constrained optimization problem.
8,793
null
null
cAw860ncLRW
Anti-Backdoor Learning: Training Clean Models on Poisoned Data
https://openreview.net/forum?id=cAw860ncLRW
[ "Yige Li", "Xixiang Lyu", "Nodens Koren", "Lingjuan Lyu", "Bo Li", "Xingjun Ma" ]
Poster
null
Backdoor attack has emerged as a major security threat to deep neural networks (DNNs). While existing defense methods have demonstrated promising results on detecting or erasing backdoors, it is still not clear whether robust training methods can be devised to prevent the backdoor triggers being injected into the train...
[ "Backdoor Defense", "Deep Neural Networks", "Anti-backdoor learning" ]
We present the first anti-backdoor learning method that allows one to train clean models out of backdoor-poisoned data.
8,783
2110.11571
title_snapshot
_CmrI7UrmCl
G-PATE: Scalable Differentially Private Data Generator via Private Aggregation of Teacher Discriminators
https://openreview.net/forum?id=_CmrI7UrmCl
[ "Yunhui Long", "Boxin Wang", "Zhuolin Yang", "Bhavya Kailkhura", "Aston Zhang", "Carl A. Gunter", "Bo Li" ]
Poster
null
Recent advances in machine learning have largely benefited from the massive accessible training data. However, large-scale data sharing has raised great privacy concerns. In this work, we propose a novel privacy-preserving data Generative model based on the PATE framework (G-PATE), aiming to train a scalable differenti...
[ "Differential Privacy", "Generative Models" ]
We propose a novel privacy-preserving data Generative model based on the PATE framework (G-PATE) to generate high-dimensional differentially private data while preserving high data utility.
8,782
1906.09338
title_snapshot
m4k66oJFK9P
INDIGO: GNN-Based Inductive Knowledge Graph Completion Using Pair-Wise Encoding
https://openreview.net/forum?id=m4k66oJFK9P
[ "Shuwen Liu", "Bernardo Cuenca Grau", "Ian Horrocks", "Egor V. Kostylev" ]
Poster
null
The aim of knowledge graph (KG) completion is to extend an incomplete KG with missing triples. Popular approaches based on graph embeddings typically work by first representing the KG in a vector space, and then applying a predefined scoring function to the resulting vectors to complete the KG. These approaches work we...
[ "Knowledge Graph Completion", "Graph Neural Networks" ]
We propose a novel GNN-based approach with pair-wise encoding for inductive knowledge graph completion problem.
8,745
null
null
bqGK5PyI6-N
Compacter: Efficient Low-Rank Hypercomplex Adapter Layers
https://openreview.net/forum?id=bqGK5PyI6-N
[ "Rabeeh Karimi mahabadi", "James Henderson", "Sebastian Ruder" ]
Poster
null
Adapting large-scale pretrained language models to downstream tasks via fine-tuning is the standard method for achieving state-of-the-art performance on NLP benchmarks. However, fine-tuning all weights of models with millions or billions of parameters is sample-inefficient, unstable in low-resource settings, and wastef...
[ "parameter-compact fine-tuning methods", "adapters", "low-rank methods", "hypercomplex multiplication layers", "adapting large-scale language models" ]
We have proposed Compacter, a light-weight fine-tuning method for large-scale language models and demonstrated that despite learning 2127.66x fewer parameters than standard fine-tuning, it obtains comparable performance.
8,743
2106.04647
title_snapshot
MckiHYXsBT
Learning to Select Exogenous Events for Marked Temporal Point Process
https://openreview.net/forum?id=MckiHYXsBT
[ "Ping Zhang", "Rishabh K Iyer", "Ashish V. Tendulkar", "Gaurav Aggarwal", "Abir De" ]
Poster
null
Marked temporal point processes (MTPPs) have emerged as a powerful modeling tool for a wide variety of applications which are characterized using discrete events localized in continuous time. In this context, the events are of two types endogenous events which occur due to the influence of the previous events and exoge...
[ "Temporal point process", "time series", "event selection" ]
It learns to select exogenous events from a set of events in the context of marked temporal point processes
8,721
null
null
Bix4uw5GcbE
Learning rule influences recurrent network representations but not attractor structure in decision-making tasks
https://openreview.net/forum?id=Bix4uw5GcbE
[ "Brandon J McMahan", "Michael Kleinman", "Jonathan Kao" ]
Poster
null
Recurrent neural networks (RNNs) are popular tools for studying computational dynamics in neurobiological circuits. However, due to the dizzying array of design choices, it is unclear if computational dynamics unearthed from RNNs provide reliable neurobiological inferences. Understanding the effects of design choices o...
[ "RNN", "learning rules", "hyperparameters", "RNN dynamics", "decision-making", "computational neuroscience", "interpretability", "biological models" ]
RNNs adopt similar dynamical mechanisms when trained with different learning rules on sufficiently complex tasks
8,704
null
null
1XwPDFrJObw
Towards Instance-Optimal Offline Reinforcement Learning with Pessimism
https://openreview.net/forum?id=1XwPDFrJObw
[ "Ming Yin", "Yu-Xiang Wang" ]
Poster
null
We study the \emph{offline reinforcement learning} (offline RL) problem, where the goal is to learn a reward-maximizing policy in an unknown \emph{Markov Decision Process} (MDP) using the data coming from a policy $\mu$. In particular, we consider the sample complexity problems of offline RL for the finite horizon MDP...
[ "Theory", "Reinforcement Learning Theory", "Markov Decision Process Theory" ]
null
8,703
2110.08695
title_snapshot
lEf52hTHq0Q
Escape saddle points by a simple gradient-descent based algorithm
https://openreview.net/forum?id=lEf52hTHq0Q
[ "Chenyi Zhang", "Tongyang Li" ]
Poster
null
Escaping saddle points is a central research topic in nonconvex optimization. In this paper, we propose a simple gradient-based algorithm such that for a smooth function $f\colon\mathbb{R}^n\to\mathbb{R}$, it outputs an $\epsilon$-approximate second-order stationary point in $\tilde{O}(\log n/\epsilon^{1.75})$ iteratio...
[ "saddle points", "gradient descent", "stochastic optimization", "nonconvex optimization", "negative curvature finding" ]
We propose a simple gradient-based algorithm to find an eps-approx. second-order stationary point of an n-dim function in ~O(log n/eps^1.75) iterations, achieving poly-speedup in log n. It's also applicable to stochastic optimization.
8,698
2111.14069
title_snapshot
xV6ZDMwRspN
Unlabeled Principal Component Analysis
https://openreview.net/forum?id=xV6ZDMwRspN
[ "Yunzhen Yao", "Liangzu Peng", "Manolis C. Tsakiris" ]
Poster
null
We introduce robust principal component analysis from a data matrix in which the entries of its columns have been corrupted by permutations, termed Unlabeled Principal Component Analysis (UPCA). Using algebraic geometry, we establish that UPCA is a well-defined algebraic problem in the sense that the only matrices of m...
[ "unlabeled sensing", "linear regression without correspondences", "robust principal component analysis", "algebraic geometry" ]
null
8,690
null
null
c1p817YZAx6
Brick-by-Brick: Combinatorial Construction with Deep Reinforcement Learning
https://openreview.net/forum?id=c1p817YZAx6
[ "Hyunsoo Chung", "Jungtaek Kim", "Boris Knyazev", "Jinhwi Lee", "Graham W. Taylor", "Jaesik Park", "Minsu Cho" ]
Poster
null
Discovering a solution in a combinatorial space is prevalent in many real-world problems but it is also challenging due to diverse complex constraints and the vast number of possible combinations. To address such a problem, we introduce a novel formulation, combinatorial construction, which requires a building agent to...
[ "Combinatorial construction", "Sequential assembly", "Deep reinforcement learning" ]
We introduce a novel problem formulation, combinatorial construction, which requires a building agent to assemble unit primitives sequentially, given incomplete partial target information.
8,675
2110.15481
title_snapshot
QcwJmp1sTnk
Agent Modelling under Partial Observability for Deep Reinforcement Learning
https://openreview.net/forum?id=QcwJmp1sTnk
[ "Georgios Papoudakis", "Filippos Christianos", "Stefano V Albrecht" ]
Poster
null
Modelling the behaviours of other agents is essential for understanding how agents interact and making effective decisions. Existing methods for agent modelling commonly assume knowledge of the local observations and chosen actions of the modelled agents during execution. To eliminate this assumption, we extract repres...
[ "agent modelling", "deep reinforcement learning", "partial observability", "opponent modelling" ]
null
8,671
2006.09447
title_snapshot
VuzPO_TZHPc
Associative Memories via Predictive Coding
https://openreview.net/forum?id=VuzPO_TZHPc
[ "Tommaso Salvatori", "Yuhang Song", "Yujian Hong", "Lei Sha", "Simon Frieder", "Zhenghua Xu", "Rafal Bogacz", "Thomas Lukasiewicz" ]
Poster
null
Associative memories in the brain receive and store patterns of activity registered by the sensory neurons, and are able to retrieve them when necessary. Due to their importance in human intelligence, computational models of associative memories have been developed for several decades now. In this paper, we present a n...
[ "Deep Learning", "Associative Memory", "Cognitive Science" ]
null
8,669
2109.08063
title_snapshot
GWRkOYr4jxQ
Luna: Linear Unified Nested Attention
https://openreview.net/forum?id=GWRkOYr4jxQ
[ "Xuezhe Ma", "Xiang Kong", "Sinong Wang", "Chunting Zhou", "Jonathan May", "Hao Ma", "Luke Zettlemoyer" ]
Poster
null
The quadratic computational and memory complexities of the Transformer's attention mechanism have limited its scalability for modeling long sequences. In this paper, we propose Luna, a linear unified nested attention mechanism that approximates softmax attention with two nested linear attention functions, yielding onl...
[ "Efficient Attention", "Transformer", "Linear Attention" ]
Linear Unified Nested Attention
8,658
2106.01540
title_snapshot
ERzpLwEDOY
Bandits with many optimal arms
https://openreview.net/forum?id=ERzpLwEDOY
[ "Rianne de Heide", "James Cheshire", "Pierre MENARD", "Alexandra Carpentier" ]
Poster
null
We consider a stochastic bandit problem with a possibly infinite number of arms. We write $p^*$ for the proportion of optimal arms and $\Delta$ for the minimal mean-gap between optimal and sub-optimal arms. We characterize the optimal learning rates both in the cumulative regret setting, and in the best-arm identificat...
[ "Infinitely-Armed Bandit Models", "Cumulative Regret", "Best-Arm Identification" ]
Stochastic bandit problems with possibly infinitely many arms but a fixed proportion of optimal arms.
8,657
2103.12452
title_snapshot
wgeK563QgSw
Offline Reinforcement Learning as One Big Sequence Modeling Problem
https://openreview.net/forum?id=wgeK563QgSw
[ "Michael Janner", "Qiyang Li", "Sergey Levine" ]
Spotlight
null
Reinforcement learning (RL) is typically viewed as the problem of estimating single-step policies (for model-free RL) or single-step models (for model-based RL), leveraging the Markov property to factorize the problem in time. However, we can also view RL as a sequence modeling problem: predict a sequence of actions th...
[ "Reinforcement learning", "transformers" ]
null
8,655
2106.02039
title_snapshot
bhdntUKwA1
Parallel and Efficient Hierarchical k-Median Clustering
https://openreview.net/forum?id=bhdntUKwA1
[ "Vincent Cohen-Addad", "Silvio Lattanzi", "Ashkan Norouzi-Fard", "Christian Sohler", "Ola Svensson" ]
Poster
null
As a fundamental unsupervised learning task, hierarchical clustering has been extensively studied in the past decade. In particular, standard metric formulations as hierarchical $k$-center, $k$-means, and $k$-median received a lot of attention and the problems have been studied extensively in different models of compu...
[ "Clustering", "k-Median", "Hierarchical" ]
In this work we propose an efficient algorithm for Hierarchical k-Median Clustering problem in distributed setting.
8,652
null
null
ebQXflQre5a
AutoBalance: Optimized Loss Functions for Imbalanced Data
https://openreview.net/forum?id=ebQXflQre5a
[ "Mingchen Li", "Xuechen Zhang", "Christos Thrampoulidis", "Jiasi Chen", "Samet Oymak" ]
Poster
null
Imbalanced datasets are commonplace in modern machine learning problems. The presence of under-represented classes or groups with sensitive attributes results in concerns about generalization and fairness. Such concerns are further exacerbated by the fact that large capacity deep nets can perfectly fit the training dat...
[ "Label imbalance", "fairness", "bilevel optimization", "personalization" ]
null
8,648
2201.01212
title_snapshot
AklttWFnxS9
Maximum Likelihood Training of Score-Based Diffusion Models
https://openreview.net/forum?id=AklttWFnxS9
[ "Yang Song", "Conor Durkan", "Iain Murray", "Stefano Ermon" ]
Spotlight
null
Score-based diffusion models synthesize samples by reversing a stochastic process that diffuses data to noise, and are trained by minimizing a weighted combination of score matching losses. The log-likelihood of score-based diffusion models can be tractably computed through a connection to continuous normalizing flows,...
[ "generative models", "density estimation", "score matching", "score-based generative models", "diffusion models", "stochastic differential equations", "normalizing flows", "neural ODEs", "likelihood", "continuous normalizing flows" ]
Score-based generative models can achieve state-of-the-art likelihoods when re-weighting the training objective.
8,646
2101.09258
title_snapshot
HhUmPH22Vpn
CoFiNet: Reliable Coarse-to-fine Correspondences for Robust PointCloud Registration
https://openreview.net/forum?id=HhUmPH22Vpn
[ "Hao Yu", "Fu Li", "Mahdi Saleh", "Benjamin Busam", "Slobodan Ilic" ]
Poster
null
We study the problem of extracting correspondences between a pair of point clouds for registration. For correspondence retrieval, existing works benefit from matching sparse keypoints detected from dense points but usually struggle to guarantee their repeatability. To address this issue, we present CoFiNet - Coarse-to-...
[ "Point cloud Matching", "Point Cloud Registration", "Coarse-to-fine Mechanism", "Deep Neural Networks" ]
We present CoFiNet - Coarse-to-Fine Network which extracts hierarchical correspondences from coarse to fine for point cloud registration.
8,643
2110.14076
title_judge
9lwprXiGdR4
Nearly Minimax Optimal Reinforcement Learning for Discounted MDPs
https://openreview.net/forum?id=9lwprXiGdR4
[ "Jiafan He", "Dongruo Zhou", "Quanquan Gu" ]
Poster
null
We study the reinforcement learning problem for discounted Markov Decision Processes (MDPs) under the tabular setting. We propose a model-based algorithm named UCBVI-$\gamma$, which is based on the \emph{optimism in the face of uncertainty principle} and the Bernstein-type bonus. We show that UCBVI-$\gamma$ achieves a...
[ "reinforcement learning", "discounted MDP", "Bernstein inequality" ]
null
8,634
2010.00587
title_snapshot
_hKvtsqItc
The Effect of the Intrinsic Dimension on the Generalization of Quadratic Classifiers
https://openreview.net/forum?id=_hKvtsqItc
[ "Fabian Latorre", "Leello Tadesse Dadi", "Paul Rolland", "Volkan Cevher" ]
Poster
null
It has been recently observed that neural networks, unlike kernel methods, enjoy a reduced sample complexity when the distribution is isotropic (i.e., when the covariance matrix is the identity). We find that this sensitivity to the data distribution is not exclusive to neural networks, and the same phenomenon can be o...
[ "intrinsic dimension", "generalization", "rademacher complexity", "statistical learning theory" ]
We show that the complexity of nuclear-norm regularized quadratic functions adapts to the intrinsic dimension of the data distribution.
8,633
null
null
4bKbEP9b65v
Littlestone Classes are Privately Online Learnable
https://openreview.net/forum?id=4bKbEP9b65v
[ "Noah Golowich", "Roi Livni" ]
Spotlight
null
We consider the problem of online classification under a privacy constraint. In this setting a learner observes sequentially a stream of labelled examples $(x_t, y_t)$, for $1 \leq t \leq T$, and returns at each iteration $t$ a hypothesis $h_t$ which is used to predict the label of each new example $x_t$. The learn...
[ "differential privacy", "online classification", "online learning" ]
We provide a differentially private online learning algorithm for Littlestone classes with finite mistake bound.
8,629
2106.13513
title_snapshot
M0J1c3PqwKZ
Not All Images are Worth 16x16 Words: Dynamic Transformers for Efficient Image Recognition
https://openreview.net/forum?id=M0J1c3PqwKZ
[ "Yulin Wang", "Rui Huang", "Shiji Song", "Zeyi Huang", "Gao Huang" ]
Poster
null
Vision Transformers (ViT) have achieved remarkable success in large-scale image recognition. They split every 2D image into a fixed number of patches, each of which is treated as a token. Generally, representing an image with more tokens would lead to higher prediction accuracy, while it also results in drastically inc...
[ "Vision Transformer", "Efficient Inference", "Image Recognition" ]
We develop a Dynamic Vision Transformer (DVT) to automatically configure a proper number of tokens for each individual image, leading to a significant improvement in computational efficiency, both theoretically and empirically.
8,621
2105.15075
title_snapshot
ui4xChWcA4R
Shape Registration in the Time of Transformers
https://openreview.net/forum?id=ui4xChWcA4R
[ "Giovanni Trappolini", "Luca Cosmo", "Luca Moschella", "Riccardo Marin", "Simone Melzi", "Emanuele Rodolà" ]
Poster
null
In this paper, we propose a transformer-based procedure for the efficient registration of non-rigid 3D point clouds. The proposed approach is data-driven and adopts for the first time the transformers architecture in the registration task. Our method is general and applies to different settings. Given a fixed template...
[ "3D Registration", "Point Clouds", "Deep Learning", "Transformers", "Shape Matching" ]
We propose a transformer based architecture to tackle the problem of non–rigid registration, with a novel surface attention mechanisms better suited to exploit the local geometric priors of the underlying structure.
8,618
2106.13679
title_snapshot
ejmqyWW0MK6
Mixability made efficient: Fast online multiclass logistic regression
https://openreview.net/forum?id=ejmqyWW0MK6
[ "Rémi Jézéquel", "Pierre Gaillard", "Alessandro Rudi" ]
Spotlight
null
Mixability has been shown to be a powerful tool to obtain algorithms with optimal regret. However, the resulting methods often suffer from high computational complexity which has reduced their practical applicability. For example, in the case of multiclass logistic regression, the aggregating forecaster (see Foster et ...
[ "Online learning", "Logistic regression" ]
null
8,596
2110.03960
title_snapshot
admg0sZZm1e
Scalable Inference in SDEs by Direct Matching of the Fokker–Planck–Kolmogorov Equation
https://openreview.net/forum?id=admg0sZZm1e
[ "Arno Solin", "Ella Maija Tamir", "Prakhar Verma" ]
Poster
null
Simulation-based techniques such as variants of stochastic Runge–Kutta are the de facto approach for inference with stochastic differential equations (SDEs) in machine learning. These methods are general-purpose and used with parametric and non-parametric models, and neural SDEs. Stochastic Runge–Kutta relies on the us...
[ "Stochastic differential equation", "approximative inference", "neural SDE", "Gaussian process" ]
Gaussian approximations are a fast and more scalable option to stochastic Runge–Kutta methods for SDEs in ML
8,590
2110.15739
title_snapshot