Publications

SemEval-2023 Task 12: Sentiment Analysis for African Languages (AfriSenti-SemEval)
Shamsuddeen Hassan Muhammad
Idris Abdulmumin
Seid Muhie Yimam
David Ifeoluwa Adelani
Ibrahim Ahmad
Nedjma OUSIDHOUM
Abinew Ayele
Saif Mohammad
Meriem Beloucif
2023-04-12
ArXiv (prépublication)
doi.org
arxiv.org
Adaptive patch foraging in deep reinforcement learning agents
Nathan Wispinski
Andrew Butcher
Kory Mathewson
Craig S Chapman
Matthew Botvinick
Patrick M. Pilarski
Patch foraging is one of the most heavily studied behavioral optimization challenges in biology. However, despite its importance to biologic… (voir plus)al intelligence, this behavioral optimization problem is understudied in artificial intelligence research. Patch foraging is especially amenable to study given that it has a known optimal solution, which may be difficult to discover given current techniques in deep reinforcement learning. Here, we investigate deep reinforcement learning agents in an ecological patch foraging task. For the first time, we show that machine learning agents can learn to patch forage adaptively in patterns similar to biological foragers, and approach optimal patch foraging behavior when accounting for temporal discounting. Finally, we show emergent internal dynamics in these agents that resemble single-cell recordings from foraging non-human primates, which complements experimental and theoretical work on the neural mechanisms of biological foraging. This work suggests that agents interacting in complex environments with ecologically valid pressures arrive at common solutions, suggesting the emergence of foundational computations behind adaptive, intelligent behavior in both biological and artificial agents.
2023-04-11
TMLR (accepté)
doi.org
openreview.net
Finite time analysis of temporal difference learning with linear function approximation: Tail averaging and regularisation
Gandharv Patil
Prashanth L.A.
Dheeraj M. Nagaraj
Doina Precup
We study the finite-time behaviour of the popular temporal difference (TD) learning algorithm, when combined with tail-averaging. We derive … (voir plus)finite time bounds on the parameter error of the tail-averaged TD iterate under a step-size choice that does not require information about the eigenvalues of the matrix underlying the projected TD fixed point. Our analysis shows that tail-averaged TD converges at the optimal O (1/t) rate, both in expectation and with high probability. In addition, our bounds exhibit a sharper rate of decay for the initial error (bias), which is an improvement over averaging all iterates. We also propose and analyse a variant of TD that incorporates regularisation, and show that this variant fares favourably in problems with ill-conditioned features.
2023-04-10
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics (publié)
doi.org
proceedings.mlr.press
A Novel Stochastic Gradient Descent Algorithm for LearningPrincipal Subspaces
Charline Le Lan
Joshua Greaves
Jesse Farebrother
Mark Rowland
Fabian Pedregosa
Rishabh Agarwal
Bellemare Marc-Emmanuel
In this paper, we derive an algorithm that learns a principal subspace from sample entries, can be applied when the approximate subspace i… (voir plus)s represented by a neural network, and hence can bescaled to datasets with an effectively infinite number of rows and columns. Our method consistsin defining a loss function whose minimizer is the desired principal subspace, and constructing agradient estimate of this loss whose bias can be controlled.
2023-04-10
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics (publié)
doi.org
proceedings.mlr.press
A surprisingly simple technique to control the pretraining bias for better transfer: Expand or Narrow your representation
Florian Bordes
Samuel Lavoie
Randall Balestriero
Nicolas Ballas
P Vincent
2023-04-10
ArXiv (prépublication)
doi.org
arxiv.org
Conservative objective models are a special kind of contrastive divergence-based energy model
Christopher Beckham
Christopher Pal
In this work we theoretically show that conservative objective models (COMs) for offline model-based optimisation (MBO) are a special kind o… (voir plus)f contrastive divergence-based energy model, one where the energy function represents both the unconditional probability of the input and the conditional probability of the reward variable. While the initial formulation only samples modes from its learned distribution, we propose a simple fix that replaces its gradient ascent sampler with a Langevin MCMC sampler. This gives rise to a special probabilistic model where the probability of sampling an input is proportional to its predicted reward. Lastly, we show that better samples can be obtained if the model is decoupled so that the unconditional and conditional probabilities are modelled separately.
2023-04-06
ArXiv (prépublication)
doi.org
arxiv.org
Approach Intelligent Writing Assistants Usability with Seven Stages of Action
Avinash Bhat
Disha Shrivastava
Jin L.C. Guo
2023-04-05
ArXiv (prépublication)
doi.org
arxiv.org
Bugs in machine learning-based systems: a faultload benchmark
Mohammad Mehdi Morovati
Amin Nikanjam
Foutse Khomh
Z. Jiang
2023-04-04
Empirical Software Engineering (publié)
doi.org
arxiv.org
Abstract 2987: BamQuery: a new proteogenomic tool to explore the immunopeptidome and prioritize actionable tumor antigens
Maria-Virginia Ruiz Cuevas
Marie-Pierre Hardy
Jean-David Larouche
Anca Apavaloaei
Eralda Kina
Krystel Vincent
Patrick Gendron
Jean-Philippe Laverdure
Chantal Durette
Pierre Thibault
S. Lemieux
Claude Perreault
Gregory Ehx
MHC class I-associated peptides (MAPs), collectively referred to as the immunopeptidome, have a pivotal role in cancer immunosurveillance. W… (voir plus)hile MAPs were long thought to be solely generated by the degradation of canonical proteins, recent advances in the field of proteogenomics (genomically-informed proteomics) evidenced that ∼10% of them originate from allegedly noncoding genomic sequences. Among these sequences, endogenous retroelements (EREs) are under intense scrutiny as a possible source of actionable tumor antigens (TAs). With the increasing number of cancer-oriented immunopeptidomic and proteogenomic studies comes the need to accurately attribute an RNA expression level to each MAP identified by mass-spectrometry. Here, we introduce BamQuery (BQ), a computational tool to attribute an exhaustive RNA expression to MAPs of any genomic origin (exon, intron, UTR, intergenic) from bulk and single-cell RNA-sequencing data. By using BQ on large datasets of published MAPs identified by mass spectrometry, we show that many of them can arise from more than one genomic region. Indeed, 27% of MAPs reported as deriving from protein-coding exons (canonical MAPs) could also arise from non-canonical genomic regions, sometimes with greater probability, and 61% of non-canonical MAPs could arise from more than a single genomic origin (334 possible regions on average per non-canonical MAP; up to 35,343 for EREs). The consideration of all these origins evidenced an unsuspected high RNA expression in normal human tissues of (i) published neoantigens/TAs (mutated or not); (ii) MAPs derived from proteasomal splicing, supposedly not genomically templated, and (iii) MAPs derived from viruses. In particular, the high expression of candidate immunotherapeutic targets such as TAs highlights the relevance of BamQuery and the necessity of using it to validate such antigens before translating their usage in clinical trials. We also demonstrate that BamQuery can be used to directly identify safe and actionable TAs as well as to predict their immunogenicity through our freely accessible web portal (https://bamquery.iric.ca/search). Therefore, BQ could become an essential tool in any TA prioritization pipeline in the near future. Citation Format: Maria-Virginia Ruiz Cuevas, Marie-Pierre Hardy, Jean-David Larouche, Anca Apavaloaei, Eralda Kina, Krystel Vincent, Patrick Gendron, Jean-Philippe Laverdure, Chantal Durette, Pierre Thibault, Sebastien Lemieux, Claude Perreault, Gregory Ehx. BamQuery: a new proteogenomic tool to explore the immunopeptidome and prioritize actionable tumor antigens [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2987.
2023-04-03
Cancer Research (publié)
doi.org
Abstract 2993: Unmutated tumor antigens are abundant and contribute to tumor control in melanoma
Anca Apavaloaei
Qingchuan Zhao
Leslie Hesnard
Krystel Vincent
Marie-Pierre Hardy
Chantal Durette
Joël Lanoix
Jean-Philippe Laverdure
Jean-David Larouche
Maria-Virginia Ruiz Cuevas
Gregory Ehx
S. Lemieux
Pierre Thibault
Claude Perreault
Recognition of MHC-I-associated tumor antigens (TAs) by CD8+ T cells is central to antitumor immunity. Owing to the elevated tumor mutationa… (voir plus)l burden (TMB) in melanoma, the marked efficacy of immune checkpoint blockade (ICB) has been attributed to the recognition of mutated TAs. However, recent reports showed that response to ICB in melanomas with low TMB is associated with CD8+ T-cell reactivity against melanocyte lineage-associated antigens (LSAs). Here, we systematically evaluated the contribution of all TA classes, i.e., mutated and unmutated, canonical and non-canonical, to the antigenic landscape of melanoma. We characterized the TAs from melanoma biopsies and patient-derived cell lines using proteogenomics. Out of 79450 MHC-I-associated peptides (MAPs) identified from 19 samples, we found 557 unmutated TAs classified as tumor-specific (TSA), tumor-associated (TAA), or LSAs. These TAs most often derived from annotated open-reading frames, followed by ncRNAs and intergenic regions. By contrast, only 6 MAPs were mutated and tumor-specific, which could be partially explained by a decreased expression of mutations within MAP-generating genomic regions. While the number of unmutated TAs with predicted presentation (TApres) in melanoma patients was similar between responders and non-responders pre-ICB, non-responders showed marks of inefficient antigen presentation. In consequence, only responders lost TApres upon treatment, in tandem with an expansion in tumor-infiltrating lymphocytes. These results reveal a previously underappreciated contribution of unmutated TAs to tumor control in melanoma and suggest that enhancing their recognition could improve the ICB efficacy in non-responders. Citation Format: Anca Apavaloaei, Qingchuan Zhao, Leslie Hesnard, Krystel Vincent, Marie-Pierre Hardy, Chantal Durette, Joël Lanoix, Jean-Philippe Laverdure, Jean-David Larouche, Maria Virginia Ruiz Cuevas, Grégory Ehx, Sébastien Lemieux, Pierre Thibault, Claude Perreault. Unmutated tumor antigens are abundant and contribute to tumor control in melanoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2993.
2023-04-03
Cancer Research (publié)
doi.org
Evaluating the Fairness of Deep Learning Uncertainty Estimates in Medical Image Analysis
Raghav Mehta
Changjian Shui
Tal Arbel
2023-04-03
MIDL.io/2023/Conference (poster)
doi.org
proceedings.mlr.press
Intent-aware Multi-source Contrastive Alignment for Tag-enhanced Recommendation
Haolun Wu
Yingxue Zhang
Chen Ma
Wei Guo
Ruiming Tang
Xue Liu
Mark J. Coates
To offer accurate and diverse recommendation services, recent methods use auxiliary information to foster the learning process of user and i… (voir plus)tem representations. Many state-of-the-art (SOTA) methods fuse different sources of information (user, item, knowledge graph, tags, etc.) into a graph and use Graph Neural Networks (GNNs) to introduce the auxiliary information through the message passing paradigm. In this work, we seek an alternative framework that is light and effective through self-supervised learning across different sources of information, particularly for the commonly accessible item tag information. We use a self-supervision signal to pair users with the auxiliary information (tags) associated with the items they have interacted with before. To achieve the pairing, we create a proxy training task. For a given item, the model predicts which is the correct pairing between the representations obtained from the users that have interacted with this item and the tags assigned to it. This design provides an efficient solution, using the auxiliary information directly to enhance the quality of user and item embeddings. User behavior in recommendation systems is driven by the complex interactions of many factors behind the users’ decision-making processes. To make the pairing process more fine-grained and avoid embedding collapse, we propose a user intent-aware self-supervised pairing process where we split the user embeddings into multiple sub-embedding vectors. Each sub-embedding vector captures a specific user intent via self-supervised alignment with a particular cluster of tags. We integrate our designed framework with various recommendation models, demonstrating its flexibility and compatibility. Through comparison with numerous SOTA methods on seven real-world datasets, we show that our method can achieve better performance while requiring less training time. This indicates the potential of applying our approach on web-scale datasets.
2023-04-02
2023 IEEE 39th International Conference on Data Engineering (ICDE) (publié)
doi.org
arxiv.org