Alex Lamb

Xu Ji

Pascal Junior Tikeng Notsawo

Michael Mozer

Kenji Kawaguchi

Vector Quantization (VQ) is a method for discretizing latent representations and has become a major part of the deep learning toolkit. It ha… (see more)s been theoretically and empirically shown that discretization of representations leads to improved generalization, including in reinforcement learning where discretization can be used to bottleneck multi-agent communication to promote agent specialization and robustness. The discretization tightness of most VQ-based methods is defined by the number of discrete codes in the representation vector and the codebook size, which are fixed as hyperparameters. In this work, we propose learning to dynamically select discretization tightness conditioned on inputs, based on the hypothesis that data naturally contains variations in complexity that call for different levels of representational coarseness which is observed in many heterogeneous data sets. We show that dynamically varying tightness in communication bottlenecks can improve model performance on visual reasoning and reinforcement learning tasks with heterogeneity in representations.

2023-06-25

Proceedings of the AAAI Conference on Artificial Intelligence (published)

arxiv.org

Leveraging the Third Dimension in Contrastive Learning

Sumukh K Aithal

Michael Curtis Mozer

Self-Supervised Learning (SSL) methods operate on unlabeled data to learn robust representations useful for downstream tasks. Most SSL metho… (see more)ds rely on augmentations obtained by transforming the 2D image pixel map. These augmentations ignore the fact that biological vision takes place in an immersive three-dimensional, temporally contiguous environment, and that low-level biological vision relies heavily on depth cues. Using a signal provided by a pretrained state-of-the-art monocular RGB-to-depth model (the \emph{Depth Prediction Transformer}, Ranftl et al., 2021), we explore two distinct approaches to incorporating depth signals into the SSL framework. First, we evaluate contrastive learning using an RGB+depth input representation. Second, we use the depth signal to generate novel views from slightly different camera positions, thereby producing a 3D augmentation for contrastive learning. We evaluate these two approaches on three different SSL methods -- BYOL, SimSiam, and SwAV -- using ImageNette (10 class subset of ImageNet), ImageNet-100 and ImageNet-1k datasets. We find that both approaches to incorporating depth signals improve the robustness and generalization of the baseline SSL methods, though the first approach (with depth-channel concatenation) is superior. For instance, BYOL with the additional depth channel leads to an increase in downstream classification accuracy from 85.3\% to 88.0\% on ImageNette and 84.1\% to 87.0\% on ImageNet-C.

2023-01-26

ArXiv (preprint)

Discrete Factorial Representations as an Abstraction for Goal Conditioned Reinforcement Learning

Hongyu Zang

Xin Li

Romain Laroche

Remi Tachet des Combes

2022-10-31

ArXiv (preprint)

arxiv.org

Interpolated Adversarial Training: Achieving Robust Neural Networks Without Sacrificing Too Much Accuracy

Vikas Verma

Kenji Kawaguchi

Juho Kannala

Alexander Matyasko

Savya Khosla

Adversarial robustness has become a central goal in deep learning, both in theory and in practice. However, successful methods to improve th… (see more)e adversarial robustness (such as adversarial training) greatly hurt generalization performance on the unperturbed data. This could have a major impact on how achieving adversarial robustness affects real world systems (i.e. many may opt to forego robustness if it can improve accuracy on the unperturbed data). We propose Interpolated Adversarial Training, which employs recently proposed interpolation based training methods in the framework of adversarial training. On CIFAR-10, adversarial training increases the standard test error (when there is no adversary) from 4.43% to 12.32%, whereas with our Interpolated adversarial training we retain adversarial robustness while achieving a standard test error of only 6.45%. With our technique, the relative increase in the standard error for the robust model is reduced from 178.1% to just 45.5%.

2022-09-30

Neural Networks (published)

arxiv.org

Coordination Among Neural Modules Through a Shared Global Workspace

Nan Rosemary Ke

Nasim Rahaman

Jonathan Binas

Charles Blundell

Michael Mozer

Deep learning has seen a movement away from representing examples with a monolithic hidden state towards a richly structured state. For exam… (see more)ple, Transformers segment by position, and object-centric architectures decompose images into entities. In all these architectures, interactions between different elements are modeled via pairwise interactions: Transformers make use of self-attention to incorporate information from other positions; object-centric architectures make use of graph neural networks to model interactions among entities. However, pairwise interactions may not achieve global coordination or a coherent, integrated representation that can be used for downstream tasks. In cognitive science, a global workspace architecture has been proposed in which functionally specialized components share information through a common, bandwidth-limited communication channel. We explore the use of such a communication channel in the context of deep learning for modeling the structure of complex environments. The proposed method includes a shared workspace through which communication among different specialist modules takes place but due to limits on the communication bandwidth, specialist modules must compete for access. We show that capacity limitations have a rational basis in that (1) they encourage specialization and compositionality and (2) they facilitate the synchronization of otherwise independent specialists.

2022-01-27

ICLR.cc/2022/Conference (oral)

Biasly: a machine learning based platform for automatic racial discrimination detection in online texts

David Bamman

Chris Dyer

Noah A. Smith. 2014

Steven Bird

Ewan Klein

Edward Loper

Nat-527

Jacob Devlin

Ming-Wei Chang

Kenton Lee

Kristina Toutanova. 2019

Bert

Samuel Gehman

Suchin Gururangan

Maarten Sap

Dan Hendrycks

Kevin Gimpel. 2020

Gaussian

Di He … (see 22 more)

Guolin Ke

Feng-Ju Liao

Mirco Ravanaelli

Zhenzhong Lan

Mingda Chen

Sebastian Goodman

Yann Lecun

Bernhard E. Boser

J. Denker

Don-608 nie Henderson

Robin Howard

Wayne Hubbard

Yinhan Liu

Myle Ott

Naman Goyal

Jingfei Du

Mandar Joshi

Danqi Chen

Omer Levy

Mike Lewis

Warning : this paper contains content that may 001 be offensive or upsetting. 002 Detecting hateful, toxic, and otherwise racist 003 or sexi… (see more)st language in user-generated online con-004 tents has become an increasingly important task 005 in recent years. Indeed, the anonymity, the 006 transience, the size of messages, and the dif-007 ficulty of management, facilitate the diffusion 008 of racist or hateful messages across the Inter-009 net. The critical influence of this cyber-racism 010 is no longer limited to social media, but also 011 has a significant effect on our society : corpo-012 rate business operation, users’ health, crimes, 013 etc. Traditional racist speech reporting chan-014 nels have proven inadequate due to the enor-015 mous explosion of information, so there is an 016 urgent need for a method to automatically and 017 promptly detect texts with racial discrimination. 018 We propose in this work, a machine learning-019 based approach to enable automatic detection 020 of racist text content over the internet. State-of-021 the-art machine learning models that are able 022 to grasp language structures are adapted in this 023 study. Our main contribution include 1) a large 024 scale racial discrimination data set collected 025 from three distinct sources and annotated ac-026 cording to a guideline developed by specialists, 027 2) a set of machine learning models with vari-028 ous architectures for racial discrimination de-029 tection, and 3) a web-browser-based software 030 that assist users to debias their texts when us-031 ing the internet. All these resources are made 032 publicly available.

2021-12-31

(published)

www.semanticscholar.org

Discrete Factorial Representations as an Abstraction for Goal Conditioned RL

Hongyu Zang

Xin Li

Romain Laroche

Remi Tachet des Combes

2021-12-31

Neural Information Processing Systems (published)

Discrete-Valued Neural Communication in Structured Architectures Enhances Generalization

Dianbo Liu

Kenji Kawaguchi

Chen Sun

Michael C. Mozer

Deep learning has advanced from fully connected architectures to structured models organized into components, e.g., the transformer composed… (see more) of positional elements, modular architectures divided into slots, and graph neural nets made up of nodes. In structured models, an interesting question is how to conduct dynamic and possibly sparse communication among the separate components. Here, we explore the hypothesis that restricting the transmitted information among components to discrete representations is a beneficial bottleneck. The motivating intuition is human language in which communication occurs through discrete symbols. Even though individuals have different understandings of what a "cat" is based on their specific experiences, the shared discrete token makes it possible for communication among individuals to be unimpeded by individual differences in internal representation. To discretize the values of concepts dynamically communicated among specialist components, we extend the quantization mechanism from the Vector-Quantized Variational Autoencoder to multi-headed discretization with shared codebooks and use it for discrete-valued neural communication (DVNC). Our experiments show that DVNC substantially improves systematic generalization in a variety of architectures -- transformers, modular architectures, and graph neural networks. We also show that the DVNC is robust to the choice of hyperparameters, making the method very useful in practice. Moreover, we establish a theoretical justification of our discretization process, proving that it has the ability to increase noise robustness and reduce the underlying dimensionality of the model.

2021-12-31

(published)

Temporal Latent Bottleneck: Synthesis of Fast and Slow Processing Mechanisms in Sequence Learning

Aniket Didolkar

Kshitij Gupta

Nitesh B. Gundavarapu

Nan Rosemary Ke

Recurrent neural networks have a strong inductive bias towards learning temporally compressed representations, as the entire history of a se… (see more)quence is represented by a single vector. By contrast, Transformers have little inductive bias towards learning temporally compressed representations, as they allow for attention over all previously computed elements in a sequence. Having a more compressed representation of a sequence may be beneficial for generalization, as a high-level representation may be more easily re-used and re-purposed and will contain fewer irrelevant details. At the same time, excessive compression of representations comes at the cost of expressiveness. We propose a solution which divides computation into two streams. A slow stream that is recurrent in nature aims to learn a specialized and compressed representation, by forcing chunks of

2021-12-31

Advances in Neural Information Processing Systems 35 (NeurIPS 2022) (published)

Neural Function Modules with Sparse Arguments: A Dynamic Approach to Integrating Information across Layers

Agnieszka Słowik

Michael Mozer

Philippe Beaudoin

Feed-forward neural networks consist of a sequence of layers, in which each layer performs some processing on the information from the previ… (see more)ous layer. A downside to this approach is that each layer (or module, as multiple modules can operate in parallel) is tasked with processing the entire hidden state, rather than a particular part of the state which is most relevant for that module. Methods which only operate on a small number of input variables are an essential part of most programming languages, and they allow for improved modularity and code re-usability. Our proposed method, Neural Function Modules (NFM), aims to introduce the same structural capability into deep learning. Most of the work in the context of feed-forward networks combining top-down and bottom-up feedback is limited to classification problems. The key contribution of our work is to combine attention, sparsity, top-down and bottom-up feedback, in a flexible algorithm which, as we show, improves the results in standard classification, out-of-domain generalization, generative modeling, and learning representations in the context of reinforcement learning.

2021-03-17

Proceedings of The 24th International Conference on Artificial Intelligence and Statistics (published)

proceedings.mlr.press

Transformers with Competitive Ensembles of Independent Mechanisms

Di He

Guolin Ke

Chien-Feng Liao

Mirco Ravanelli

An important development in deep learning from the earliest MLPs has been a move towards architectures with structural inductive biases whic… (see more)h enable the model to keep distinct sources of information and routes of processing well-separated. This structure is linked to the notion of independent mechanisms from the causality literature, in which a mechanism is able to retain the same processing as irrelevant aspects of the world are changed. For example, convnets enable separation over positions, while attention-based architectures (especially Transformers) learn which combination of positions to process dynamically. In this work we explore a way in which the Transformer architecture is deficient: it represents each position with a large monolithic hidden representation and a single set of parameters which are applied over the entire hidden representation. This potentially throws unrelated sources of information together, and limits the Transformer's ability to capture independent mechanisms. To address this, we propose Transformers with Independent Mechanisms (TIM), a new Transformer layer which divides the hidden representation and parameters into multiple mechanisms, which only exchange information through attention. Additionally, we propose a competition mechanism which encourages these mechanisms to specialize over time steps, and thus be more independent. We study TIM on a large-scale BERT model, on the Image Transformer, and on speech enhancement and find evidence for semantically meaningful specialization as well as improved performance.

2021-02-26

ArXiv (preprint)