Irina Rish

Biographie

Irina Rish est professeure titulaire à l'Université de Montréal (UdeM), où elle dirige le Laboratoire d'IA autonome. Membre du corps professoral de Mila – Institut québécois d’intelligence artificielle, elle est titulaire d'une chaire d'excellence en recherche du Canada (CERC) et d'une chaire en IA Canada-CIFAR. Irina dirige le projet INCITE du ministère américain de l'Environnement au sujet des modèles de fondation évolutifs sur les superordinateurs Summit et Frontier à l'Oak Ridge Leadership Computing Facility (OLCF). Elle est cofondatrice et directrice scientifique de Nolano.ai.

Ses recherches actuelles portent sur les lois de mise à l'échelle neuronale et les comportements émergents (capacités et alignement) dans les modèles de fondation, ainsi que sur l'apprentissage continu, la généralisation hors distribution et la robustesse. Avant de se joindre à l'UdeM en 2019, Irina était chercheuse au Centre de recherche IBM Thomas J. Watson, où elle a travaillé sur divers projets à l'intersection des neurosciences et de l'IA, et dirigé le défi NeuroAI. Elle a reçu plusieurs prix IBM : ceux de l’excellence et de l’innovation exceptionnelle (2018), celui de la réalisation technique exceptionnelle (2017), et celui de l’accomplissement en recherche (2009). Elle détient 64 brevets et a écrit plus de 120 articles de recherche, plusieurs chapitres de livres, trois livres publiés et une monographie sur la modélisation éparse.

Étudiants actuels

Germán Abrevaya

Visiteur de recherche indépendant - UdeM

Co-superviseur⋅e :

George Adamopoulos

Stagiaire de recherche

Ivan Anokhin

Doctorat - UdeM

Co-superviseur⋅e :

Samira Ebrahimi Kahou

Doctorat - UdeM

Arjun Ashok

Doctorat - UdeM

Co-superviseur⋅e :

Reza Bayat

Maîtrise recherche - UdeM

Co-superviseur⋅e :

Pouya Bashivan

Rishika Bhagwatkar

Maîtrise recherche - UdeM

Collaborateur·rice de recherche - UdeM

Ethan Caballero

Doctorat - McGill

Superviseur⋅e principal⋅e :

Blake Richards

Mohammad-javad Darvishi-bayazi Darvishi Bayazi

Amin Darabi

Doctorat - UdeM

Doctorat - UdeM

Doctorat - UdeM

Co-superviseur⋅e :

Doctorat - UdeM

Collaborateur·rice alumni - UdeM

Co-superviseur⋅e :

Sarath Chandar

Nadhir Hassen

Stagiaire de recherche - UdeM

Maîtrise recherche

Doctorat - UdeM

Superviseur⋅e principal⋅e :

Ioannis Mitliagkas

Nizar Islah

Doctorat - UdeM

Superviseur⋅e principal⋅e :

Eilif Benjamin Muller

Doctorat - UdeM

Collaborateur·rice de recherche

Zafir Khalid

Maîtrise recherche - Concordia

Superviseur⋅e principal⋅e :

Maîtrise recherche - UdeM

Neeraj Kumar

Maîtrise professionnelle - UdeM

Gwen Legate

Doctorat - Concordia

Superviseur⋅e principal⋅e :

Eugene Belilovsky

David Lemay

Maîtrise recherche - UdeM

amin.mansouri@mila.quebec

Ekaterina Lobacheva

Postdoctorat - UdeM

Superviseur⋅e principal⋅e :

Maîtrise recherche - UdeM

Collaborateur·rice alumni

Doctorat - UdeM

Maîtrise recherche - UdeM

Diganta Misra

Maîtrise recherche - UdeM

Remus Mocanu

Maîtrise recherche - UdeM

Timothy Nest

Doctorat - UdeM

Co-superviseur⋅e :

Eilif Benjamin Muller

Mohammad Pezeshki

Collaborateur·rice de recherche

Motahareh Pourrahimi

Doctorat - McGill

Superviseur⋅e principal⋅e :

Pouya Bashivan

Mahta Ramezanian

Maîtrise recherche - UdeM

Co-superviseur⋅e :

Roland Riachi

Collaborateur·rice de recherche - UdeM

Matthew Riemer

Doctorat - UdeM

Alexis Roger

Doctorat - McGill

Superviseur⋅e principal⋅e :

Blake Richards

Zahra Sheikhbahaee

Postdoctorat - UdeM

Superviseur⋅e principal⋅e :

Vaibhav Singh

Doctorat - Concordia

Superviseur⋅e principal⋅e :

Doctorat - UdeM

Doctorat - UdeM

Co-superviseur⋅e :

Maîtrise recherche - UdeM

Dave Whipps

Maîtrise recherche - UdeM

Doctorat - UdeM

Maîtrise recherche - UdeM

Omar Younis

Collaborateur·rice de recherche

Superviseur⋅e principal⋅e :

Yoshua Bengio

Publications

Realtime Reinforcement Learning: Towards Rapid Asynchronous Deployment of Large Models

Matthew D Riemer

Gopeshh Subbaraj

Glen Berseth

Realtime environments change even as agents perform action inference and learning, thus requiring high interaction frequencies to effectivel… (voir plus)y minimize long-term regret. However, recent advances in machine learning involve larger neural networks with longer inference times, raising questions about their applicability in realtime systems where reaction time is crucial. We present an analysis of lower bounds on regret in realtime environments to show that minimizing long-term regret is generally impossible within the typical sequential interaction and learning paradigm, but often becomes possible when sufficient asynchronous compute is available. We propose novel algorithms for staggering asynchronous inference processes to ensure that actions are taken at consistent time intervals, and demonstrate that use of models with high action inference times is only constrained by the environment's effective stochasticity over the inference horizon, and not by action frequency. Our analysis shows that the number of inference processes needed scales linearly with increasing inference times while enabling use of models that are multiple orders of magnitude larger than existing approaches when learning from a realtime simulation of Game Boy games such as Pokemon and Tetris.

2024-06-19

ICML.cc/2024/Workshop/ARLET (poster)

Scalable Approaches for a Theory of Many Minds

Maximilian Puelma Touzel

Amin Memarian

Matthew D Riemer

Andrei Mircea

Andrew Robert Williams

Elin Ahlstrand

Lucas Lehnert

Rupali Bhati

A major challenge as we move towards building agents for real-world problems, which could involve a massive number of human and/or machine a… (voir plus)gents, is that we must learn to reason about the behavior of these many other agents. In this paper, we consider the problem of scaling a predictive Theory of Mind (ToM) model to a very large number of interacting agents with a fixed computational budget. Motivated by the limited diversity of agent types, existing approaches to scalable TOM learn versatile single-agent representations for quickly adapting to new agents encountered sequentially. We consider the more general setting that many agents are observed in parallel and formulate the corresponding Theory of Many Minds (ToMM) problem of estimating the joint policy. We frame the scaling behavior of solutions in terms of parameter sharing schemes and in particular propose two parameter-free architectural features that endow models with the ability to exploit action correlations: encoding a multi-agent context, and decoding through an abstracted joint action space. The increased predictive capabilities that have come with foundation models have made it easier to imagine the possibility of using these models to make simulations that imitate the behavior of many agents within complex real-world systems. Being able to perform these simulations in a general-purpose way would not only help make more capable agents, it also would be a very useful capability for applications in social science, political science, and economics.

2024-06-18

ICML.cc/2024/Workshop/Agentic_Markets (poster)

Is a Good Description Worth a Thousand Pictures? Reducing Multimodal Alignment to Text-Based, Unimodal Alignment

Amin Memarian

Touraj Laleh

Ardavan S. Nobandegani

Generative AI systems (ChatGPT, Llama, etc.) are increasingly adopted across a range of high-stake domains, including healthcare and crimina… (voir plus)l justice system. This rapid adoption indeed raises moral and ethical concerns. The emerging field of AI alignment aims to make AI systems that respect human values. In this work, we focus on evaluating the ethics of multimodal AI systems involving both text and images --- a relatively under-explored area, as most alignment work is currently focused on language models. Specifically, here we investigate whether the multimodal alignment problem (i.e., the problem of aligning a multimodal system) could be effectively reduced to the (text-based) unimodal alignment problem, wherein a language model would make a moral judgment purely based on a description of an image. Focusing on GPT-4 and LLaVA as two prominent examples of multimodal systems, here we demonstrate, rather surprisingly, that this reduction can be achieved with a relatively small loss in moral judgment performance in the case of LLaVa, and virtually no loss in the case of GPT-4.

2024-06-17

ICML.cc/2024/Workshop/MFHAIA (poster)

Lost in Translation: The Algorithmic Gap Between LMs and the Brain

Tosato Tommaso

Tikeng Notsawo Pascal Junior

Helbling Saskia

Language Models (LMs) have achieved impressive performance on various linguistic tasks, but their relationship to human language processing … (voir plus)in the brain remains unclear. This paper examines the gaps and overlaps between LMs and the brain at different levels of analysis, emphasizing the importance of looking beyond input-output behavior to examine and compare the internal processes of these systems. We discuss how insights from neuroscience, such as sparsity, modularity, internal states, and interactive learning, can inform the development of more biologically plausible language models. Furthermore, we explore the role of scaling laws in bridging the gap between LMs and human cognition, highlighting the need for efficiency constraints analogous to those in biological systems. By developing LMs that more closely mimic brain function, we aim to advance both artificial intelligence and our understanding of human cognition.

2024-06-17

ICML.cc/2024/Workshop/LLMs_and_Cognition (poster)

Revisiting Successor Features for Inverse Reinforcement Learning

Arnav Kumar Jain

Harley Wiltzer

Jesse Farebrother

Glen Berseth

Sanjiban Choudhury

2024-06-17

ICML.cc/2024/Workshop/MFHAIA (poster)

Gradient Dissent in Language Model Training and Saturation

Andrei Mircea

Ekaterina Lobacheva

We seek to shed light on language model (LM) saturation from the perspective of learning dynamics. To this end, we define a decomposition o… (voir plus)f the cross-entropy gradient, which forms a shared low-dimensional basis for analyzing the training dynamics of models across scales. Intuitively, this decomposition consists of attractive and repulsive components that increase the logit of the correct class and decrease the logits of incorrect classes, respectively. Our analysis in this subspace reveals a phenomenon we term \textit{gradient dissent}, characterized by gradient components becoming systematically opposed such that loss cannot be improved along one component without being degraded along the other. Notably, we find that complete opposition, which we term \textit{total dissent}, reliably occurs in tandem with the saturation of smaller LMs. Based on these results, we hypothesize that gradient dissent can provide a useful foundation for better understanding and mitigating saturation.

2024-06-16

ICML.cc/2024/Workshop/HiLD (poster)

$\mu$LO: Compute-Efficient Meta-Generalization of Learned Optimizers

Benjamin Thérien

Charles-Étienne Joseph

Boris Knyazev

Edouard Oyallon

Eugene Belilovsky

2024-05-31

ArXiv (prépublication)

doi.org

arxiv.org

Deep Generative Sampling in the Dual Divergence Space: A Data-efficient&Interpretative Approach for Generative AI

Sahil Garg

Anderson Schneider

Anant Raj

Kashif Rasul

Yuriy Nevmyvaka

S. Gopal

Amit Dhurandhar

Guillermo A. Cecchi

Building on the remarkable achievements in generative sampling of natural images, we propose an innovative challenge, potentially overly amb… (voir plus)itious, which involves generating samples of entire multivariate time series that resemble images. However, the statistical challenge lies in the small sample size, sometimes consisting of a few hundred subjects. This issue is especially problematic for deep generative models that follow the conventional approach of generating samples from a canonical distribution and then decoding or denoising them to match the true data distribution. In contrast, our method is grounded in information theory and aims to implicitly characterize the distribution of images, particularly the (global and local) dependency structure between pixels. We achieve this by empirically estimating its KL-divergence in the dual form with respect to the respective marginal distribution. This enables us to perform generative sampling directly in the optimized 1-D dual divergence space. Specifically, in the dual space, training samples representing the data distribution are embedded in the form of various clusters between two end points. In theory, any sample embedded between those two end points is in-distribution w.r.t. the data distribution. Our key idea for generating novel samples of images is to interpolate between the clusters via a walk as per gradients of the dual function w.r.t. the data dimensions. In addition to the data efficiency gained from direct sampling, we propose an algorithm that offers a significant reduction in sample complexity for estimating the divergence of the data distribution with respect to the marginal distribution. We provide strong theoretical guarantees along with an extensive empirical evaluation using many real-world datasets from diverse domains, establishing the superiority of our approach w.r.t. state-of-the-art deep learning methods.

2024-04-10

ArXiv (prépublication)

doi.org

arxiv.org

Simple and Scalable Strategies to Continually Pre-train Large Language Models

Adam Ibrahim

Benjamin Thérien

Kshitij Gupta

Mats Leon Richter

Quentin Anthony

Timothee LESORT

Eugene Belilovsky

Large language models (LLMs) are routinely pre-trained on billions of tokens, only to start the process over again once new data becomes ava… (voir plus)ilable. A much more efficient solution is to continually pre-train these models, saving significant compute compared to re-training. However, the distribution shift induced by new data typically results in degraded performance on previous data or poor adaptation to the new data. In this work, we show that a simple and scalable combination of learning rate (LR) re-warming, LR re-decaying, and replay of previous data is sufficient to match the performance of fully re-training from scratch on all available data, as measured by the final loss and the average score on several language model (LM) evaluation benchmarks. Specifically, we show this for a weak but realistic distribution shift between two commonly used LLM pre-training datasets (English

2024-03-13

ArXiv (prépublication)

doi.org

arxiv.org

Predicting Grokking Long Before it Happens: A look into the loss landscape of models which grok

Tikeng Notsawo Pascal Junior

Pascal Notsawo

Hattie Zhou

Mohammad Pezeshki