Publications

Hypernetworks for Zero-shot Transfer in Reinforcement Learning
Sahand Rezaei-Shoshtari
Charlotte Morissette
Francois R. Hogan
Gregory Dudek
David Meger
In this paper, hypernetworks are trained to generate behaviors across a range of unseen task conditions, via a novel TD-based training objec… (voir plus)tive and data from a set of near-optimal RL solutions for training tasks. This work relates to meta RL, contextual RL, and transfer learning, with a particular focus on zero-shot performance at test time, enabled by knowledge of the task parameters (also known as context). Our technical approach is based upon viewing each RL algorithm as a mapping from the MDP specifics to the near-optimal value function and policy and seek to approximate it with a hypernetwork that can generate near-optimal value functions and policies, given the parameters of the MDP. We show that, under certain conditions, this mapping can be considered as a supervised learning problem. We empirically evaluate the effectiveness of our method for zero-shot transfer to new reward and transition dynamics on a series of continuous control tasks from DeepMind Control Suite. Our method demonstrates significant improvements over baselines from multitask and meta RL approaches.
2023-06-25
Proceedings of the AAAI Conference on Artificial Intelligence (publié)
doi.org
arxiv.org
Latent Space Evolution under Incremental Learning with Concept Drift (Student Abstract)
Charles Bourbeau
Audrey Durand
This work investigates the evolution of latent space when deep learning models are trained incrementally in non-stationary environments that… (voir plus) stem from concept drift. We propose a methodology for visualizing the incurred change in latent representations. We further show that classes not targeted by concept drift can be negatively affected, suggesting that the observation of all classes during learning may regularize the latent space.
2023-06-25
Proceedings of the AAAI Conference on Artificial Intelligence (publié)
doi.org
Signed Laplacian Graph Neural Networks
Yu Li
Meng Qu
Jian Tang
Yi Chang
This paper studies learning meaningful node representations for signed graphs, where both positive and negative links exist. This problem ha… (voir plus)s been widely studied by meticulously designing expressive signed graph neural networks, as well as capturing the structural information of the signed graph through traditional structure decomposition methods, e.g., spectral graph theory. In this paper, we propose a novel signed graph representation learning framework, called Signed Laplacian Graph Neural Network (SLGNN), which combines the advantages of both. Specifically, based on spectral graph theory and graph signal processing, we first design different low-pass and high-pass graph convolution filters to extract low-frequency and high-frequency information on positive and negative links, respectively, and then combine them into a unified message passing framework. To effectively model signed graphs, we further propose a self-gating mechanism to estimate the impacts of low-frequency and high-frequency information during message passing. We mathematically establish the relationship between the aggregation process in SLGNN and signed Laplacian regularization in signed graphs, and theoretically analyze the expressiveness of SLGNN. Experimental results demonstrate that SLGNN outperforms various competitive baselines and achieves state-of-the-art performance.
2023-06-25
Proceedings of the AAAI Conference on Artificial Intelligence (publié)
doi.org
The Effect of Diversity in Meta-Learning
Ramnath Kumar
Tristan Deleu
Yoshua Bengio
Recent studies show that task distribution plays a vital role in the meta-learner's performance. Conventional wisdom is that task diversity … (voir plus)should improve the performance of meta-learning. In this work, we find evidence to the contrary; (i) our experiments draw into question the efficacy of our learned models: similar manifolds can be learned with a subset of the data (lower task diversity). This finding questions the advantage of providing more data to the model, and (ii) adding diversity to the task distribution (higher task diversity) sometimes hinders the model and does not lead to a significant improvement in performance as previously believed. To strengthen our findings, we provide both empirical and theoretical evidence.
2023-06-25
Proceedings of the AAAI Conference on Artificial Intelligence (publié)
doi.org
openreview.net
Partial Ordered Statistics Decoding with Enhanced Error Patterns
Marwan Jalaleddine
Huayi Zhou
Jiajie Li
Warren J. Gross
Guessing Random Additive Noise Decoding (GRAND) excels at decoding high-rate codes but struggles to decode low-rate codes with reasonable co… (voir plus)mplexity. Ordered Statistics Decoding (OSD) specifically excels in decoding short codes irrespective of rates; however, OSD necessitates the use of Gaussian elimination which introduces additional time, space and computational complexity. Partial Ordered Statistics Decoding (POSD) was proposed to reduce the time, space, and computational complexity of OSD; however, the current partition-based POSD has poor decoding performance since it does not generate test error patterns across partitions. In this paper, we propose to improve the decoding performance of POSD by incorporating test error patterns inspired by GRAND methods. This work offers a trade-off between performance and complexity compared to existing decoders such as GRAND and OSD. We enhance POSD by optimizing the scheduling of Test Error Patterns (TEPs) and show that our technique can be applied to any code in a standard form. At a target BER 10−4 with eBCH (128,64) the enhanced error patterns achieve more than 0.6 dB gain in performance compared to the POSD with partition-based error patterns. Moreover, at a target frame error rate of 10−5, POSD uses 10× less binary operations compared to GRAND when decoding eBCH (128,64) and RLC(128,64) codes. With BCH (127,29) and RLC(128,32), at a target frame error rate of 10−2, POSD with enhanced error patterns with a maximum number of queries (MQ) of 104 achieves up to a 2 dB gain to its GRAND equivalent which is using 107 maximum number of queries.
2023-06-24
2023 IEEE International Symposium on Information Theory (ISIT) (publié)
doi.org
CeBed: A Benchmark for Deep Data-Driven OFDM Channel Estimation
Amal Feriani
Di Wu
Steve Liu
Gregory Dudek
2023-06-22
ArXiv (prépublication)
doi.org
arxiv.org
Goal-conditioned GFlowNets for Controllable Multi-Objective Molecular Design
Julien Roy
Pierre-Luc Bacon
Christopher Pal
Emmanuel Bengio
In recent years, in-silico molecular design has received much attention from the machine learning community. When designing a new compound f… (voir plus)or pharmaceutical applications, there are usually multiple properties of such molecules that need to be optimised: binding energy to the target, synthesizability, toxicity, EC50, and so on. While previous approaches have employed a scalarization scheme to turn the multi-objective problem into a preference-conditioned single objective, it has been established that this kind of reduction may produce solutions that tend to slide towards the extreme points of the objective space when presented with a problem that exhibits a concave Pareto front. In this work we experiment with an alternative formulation of goal-conditioned molecular generation to obtain a more controllable conditional model that can uniformly explore solutions along the entire Pareto front.
2023-06-22
ICML.cc/2023/Workshop/DeployableGenerativeAI (publié)
doi.org
openreview.net
Using modular connectome-based predictive modeling to reveal brain-behavior relationships of individual differences in working memory
Huayi Yang
Junjun Zhang
Zhenlan Jin
Pouya Bashivan
Ling Li
2023-06-21
Brain Structure and Function (publié)
doi.org
Accelerating exploration and representation learning with offline pre-training
Bogdan Mazoure
Jacob Bruce
Doina Precup
Rob Fergus
Ankit Anand
Sequential decision-making agents struggle with long horizon tasks, since solving them requires multi-step reasoning. Most reinforcement lea… (voir plus)rning (RL) algorithms address this challenge by improved credit assignment, introducing memory capability, altering the agent's intrinsic motivation (i.e. exploration) or its worldview (i.e. knowledge representation). Many of these components could be learned from offline data. In this work, we follow the hypothesis that exploration and representation learning can be improved by separately learning two different models from a single offline dataset. We show that learning a state representation using noise-contrastive estimation and a model of auxiliary reward separately from a single collection of human demonstrations can significantly improve the sample efficiency on the challenging NetHack benchmark. We also ablate various components of our experimental setting and highlight crucial insights.
2023-06-19
ICML.cc/2023/Workshop/ILHF (accepté)
doi.org
openreview.net
Cognitive Models as Simulators: Using Cognitive Models to Tap into Implicit Human Feedback
Ardavan S. Nobandegani
Thomas Shultz
Irina Rish
2023-06-19
ICML.cc/2023/Workshop/ILHF (publié)
openreview.net
openreview.net
Constant Memory Attention Block
Leo Feng
Frederick Tung
Hossein Hajimirsadeghi
Yoshua Bengio
Mohamed Osama Ahmed
2023-06-19
ICML.cc/2023/Workshop/ES-FoMO (poster)
doi.org
openreview.net
Continual Pre-Training of Large Language Models: How to (re)warm your model?
Kshitij Gupta
Benjamin Therien
Adam Ibrahim
Mats Leon Richter
Quentin Gregory Anthony
Eugene Belilovsky
Irina Rish
Timothee LESORT
Large language models (LLMs) are routinely pre-trained on billions of tokens, only to restart the process over again once new data becomes a… (voir plus)vailable. A much cheaper and more efficient solution would be to enable the continual pre-training of these models, i.e. updating pre-trained models with new data instead of re-training them from scratch. However, the distribution shift induced by novel data typically results in degraded performance on past data. Taking a step towards efficient continual pre-training, in this work, we examine the effect of different warm-up strategies. Our hypothesis is that the learning rate must be re-increased to improve compute efficiency when training on a new dataset. We study the warmup phase of models pre-trained on the Pile (upstream data, 300B tokens) as we continue to pre-train on SlimPajama (downstream data, 297B tokens), following a linear warmup and cosine decay schedule. We conduct all experiments on the Pythia 410M language model architecture and evaluate performance through validation perplexity. We experiment with different pre-training checkpoints, various maximum learning rates, and various warmup lengths. Our results show that while rewarming models first increases the loss on upstream and downstream data, in the longer run it improves the downstream performance, outperforming models trained from scratch
2023-06-19
ICML.cc/2023/Workshop/ES-FoMO (poster)
doi.org
openreview.net