Publications

Hierarchical Classification of Narwhal Subpopulations Using Social Distance

Bertrand Charry

Marianne Marcoux

Jeffrey A. Cardille

Xavier Giroux‐Bougard

Murray M. Humphries

Effective wildlife management and conservation require knowledge of distribution, sex composition, and age structure of a population. We exp… (see more)lored the distribution of the Baffin Bay narwhal (Monodon monoceros) population in August 2013 by documenting sex and age distribution across the Canadian Arctic Archipelago covering 2,317,152 km2. For 6,314 narwhals identified in 3,393 aerial images taken across the Eastern Canadian Arctic, we calculated a matrix of swimming distances between all individuals. We then used a quantitative clustering approach to partition our dataset (partitioning around the medoids). The clusters obtained from the analysis supported the delimitation of the 5 narwhal management stocks currently used by the Department of Fisheries and Oceans but did not support the hypothesized division of Jones Sound and Smith Sound stocks. Across the 5 clusters, male:female ratios varied between 0.72 and 1.44 and the proportion of newborns relative to the number of females varied between 0.07 and 0.18. As a highly detailed snapshot of narwhal distribution across a very large region, our study is a step toward better documentation of the basic population information required for stock assessment, sustainable harvest, and habitat protection of narwhals in an era of rapid Arctic change. © 2019 The Wildlife Society.

2023-12-31

International Conference on Machine Learning (published)

doi.org

How Should We Extract Discrete Audio Tokens from Self-Supervised Models?

Jarod Duret

Yusuf Cem Sübakan

Mirco Ravanaelli

2023-12-31

INTERSPEECH (published)

doi.org

arxiv.org

ImageFlowNet: Forecasting Multiscale Image-Level Trajectories of Disease Progression with Irregularly-Sampled Longitudinal Medical Images

Chen Liu

Ke Xu

Liangbo L. Shen

Jay Stewart

Jay C. Wang

Lucian V. Del Priore

Smita Krishnaswamy

Advances in medical imaging technologies have enabled the collection of longitudinal images, which involve repeated scanning of the same pat… (see more)ients over time, to monitor disease progression. However, predictive modeling of such data remains challenging due to high dimensionality, irregular sampling, and data sparsity. To address these issues, we propose ImageFlowNet, a novel model designed to forecast disease trajectories from initial images while preserving spatial details. ImageFlowNet first learns multiscale joint representation spaces across patients and time points, then optimizes deterministic or stochastic flow fields within these spaces using a position-parameterized neural ODE/SDE framework. The model leverages a UNet architecture to create robust multiscale representations and mitigates data scarcity by combining knowledge from all patients. We provide theoretical insights that support our formulation of ODEs, and motivate our regularizations involving high-level visual features, latent space organization, and trajectory smoothness. We validate ImageFlowNet on three longitudinal medical image datasets depicting progression in geographic atrophy, multiple sclerosis, and glioblastoma, demonstrating its ability to effectively forecast disease progression and outperform existing methods. Our contributions include the development of ImageFlowNet, its theoretical underpinnings, and empirical validation on real-world datasets. The official implementation is available at https://github.com/KrishnaswamyLab/ImageFlowNet.

2023-12-31

arXiv (preprint)

doi.org

Improving the Generalizability and Robustness of Large-Scale Traffic Signal Control

Tianyu Shi

FranÃ§ois-Xavier Devailly

Denis Larocque

Laurent Charlin

A number of deep reinforcement-learning (RL) approaches propose to control traffic signals. Compared to traditional approaches, RL approache… (see more)s can learn from higher-dimensionality input road and vehicle sensors and better adapt to varying traffic conditions resulting in reduced travel times (in simulation). However, these RL methods require training from massive traffic sensor data. To offset this relative inefficiency, some recent RL methods have the ability to first learn from small-scale networks and then generalize to unseen city-scale networks without additional retraining (zero-shot transfer). In this work, we study the robustness of such methods along two axes. First, sensor failures and GPS occlusions create missing-data challenges and we show that recent methods remain brittle in the face of these missing data. Second, we provide a more systematic study of the generalization ability of RL methods to new networks with different traffic regimes. Again, we identify the limitations of recent approaches. We then propose using a combination of distributional and vanilla reinforcement learning through a policy ensemble. Building upon the state-of-the-art previous model which uses a decentralized approach for large-scale traffic signal control with graph convolutional networks (GCNs), we first learn models using a distributional reinforcement learning (DisRL) approach. In particular, we use implicit quantile networks (IQN) to model the state-action return distribution with quantile regression. For traffic signal control problems, an ensemble of standard RL and DisRL yields superior performance across different scenarios, including different levels of missing sensor data and traffic flow patterns. Furthermore, the learning scheme of the resulting model can improve zero-shot transferability to different road network structures, including both synthetic networks and real-world networks (e.g., Luxembourg, Manhattan). We conduct extensive experiments to compare our approach to multi-agent reinforcement learning and traditional transportation approaches. Results show that the proposed method improves robustness and generalizability in the face of missing data, varying road networks, and traffic flows.

2023-12-31

IEEE Open Journal of Intelligent Transportation Systems (published)

doi.org

arxiv.org

Inertia-Based Indices to Determine the Number of Clusters in K-Means: An Experimental Evaluation

Andrei Rykov

Renato Cordeiro De Amorim

Vladimir Makarenkov

Boris Mirkin

This paper gives an experimentally supported review and comparison of several indices based on the conventional K-means inertia criterion fo… (see more)r determining the number of clusters,

2023-12-31

IEEE Access (published)

doi.org

Initial Separation of Poles to form a Bipolar Spindle is a Rapid and Irreversible Process

Allen Y. Leary

Elena Nazarova

Shannon Sim

Eileen O'Toole

Paul François

Jackie Vogel

2023-12-31

International Conference on Machine Learning (published)

doi.org

Interoceptive technologies for psychiatric interventions: From diagnosis to clinical applications

Felix Schoeller

Adam Haar Horowitz

Abhinandan Jain

Pattie Maes

Nicco Reggente

Leonardo Christov-Moore

Giovanni Pezzulo

Laura Barca

Micah Allen

Roy Salomon

Mark Miller

Daniele Di Lernia

Giuseppe Riva

Manos Tsakiris

Moussa A. Chalah

Arno Klein

Ben Zhang

Teresa Garcia

Ursula Pollack

Marion Trousselard … (see 4 more)

Charles Verdonk

Guillaume Dumas

Vladimir Adrien

Karl Friston

The perception of body signals play a crucial role in cognition and emotion, which may lead to catastrophic outcomes when it becomes dysf… (see more)unctional. To characterize these mechanisms and intervene on interoception for either diagnostic or treatment purposes, a mounting body of research is concerned with interventions on interoceptive channels such as respiration, cardioception, or thermoception. However, we are still lacking a mechanistic understanding of the underlying psychophysiology. For example, interoceptive signals are often both the cause and consequences of some distress in various mental disorders, and it is still unclear how interoceptive signals bind with exteroceptive cues. In this article, we present existing technologies for manipulating interoception and review their clinical potential in light of the predictive processing framework describing interoception as a process of minimization of prediction errors. We distinguish between three kinds of stimuli: artificial sensations that concern the direct manipulation of interoceptive signals, interoceptive illusions that manipulate contextual cues to induce a predictable drift in body perception, and emotional augmentation technologies that blend artificial sensations with contextual cues of personal significance to generate specific moods or emotions. We discuss how each technology can assess and intervene on the precision-weighting of prediction errors along the cognitive and emotional processing hierarchy and conclude by discussing the clinical relevance of interoceptive technologies in terms of diagnostic stress tests for evaluating interoceptive abilities across clinical conditions and as intervention protocols for conditions such as generalized anxiety disorders, post-traumatic stress disorders, and autism spectrum disorders.

2023-12-31

Neuroscience & Biobehavioral Reviews (published)

doi.org

INTREPPPID—an orthologue-informed quintuplet network for cross-species prediction of protein-protein interaction

Joseph Szymborski

Amin Emad

An overwhelming majority of protein–protein interaction (PPI) studies are conducted in a select few model organisms largely due to constra… (see more)ints in time and cost of the associated ‘wet lab’ experiments. In silico PPI inference methods are ideal tools to overcome these limitations, but often struggle with cross-species predictions. We present INTREPPPID, a method that incorporates orthology data using a new ‘quintuplet’ neural network, which is constructed with five parallel encoders with shared parameters. INTREPPPID incorporates both a PPI classification task and an orthologous locality task. The latter learns embeddings of orthologues that have small Euclidean distances between them and large distances between embeddings of all other proteins. INTREPPPID outperforms all other leading PPI inference methods tested on both the intraspecies and cross-species tasks using strict evaluation datasets. We show that INTREPPPID’s orthologous locality loss increases performance because of the biological relevance of the orthologue data and not due to some other specious aspect of the architecture. Finally, we introduce PPI.bio and PPI Origami, a web server interface for INTREPPPID and a software tool for creating strict evaluation datasets, respectively. Together, these two initiatives aim to make both the use and development of PPI inference tools more accessible to the community.

2023-12-31

Briefings Bioinform. (published)

doi.org

ÌròyìnSpeech: A multi-purpose Yorùbá Speech Corpus

Tolúlope' Ògúnremí

Kọ́lá Túbọ̀sún

Aremu Anuoluwapo

Iroro Orife

David Ifeoluwa Adelani

2023-12-31

LREC/COLING (published)

doi.org

arxiv.org

KD-LoRA: A Hybrid Approach to Efficient Fine-Tuning with LoRA and Knowledge Distillation

Rambod Azimi

Rishav

Marek Teichmann

S Ebrahimi Kahou

2023-12-31

ENLSP (published)

doi.org

proceedings.mlr.press

Large‐scale geographic patterns of diversity and community structure of pelagic crustacean zooplankton in <scp>C</scp>anadian lakes

Bernadette Pinel‐Alloul

Adrien André

Pierre Legendre

Jeffrey A. Cardille

Kasimierz Patalas

Alex Salki

The richness‐energy theory explains diversity patterns of freshwater crustacean zooplankton in Canadian ecoprovinces. Solar radiation is t… (see more)he best predictor explaining regional species richness in ecoprovinces and community structure of pelagic crustaceans in Canadian lakes.

2023-12-31

International Conference on Machine Learning (published)

doi.org

Learnable Filters for Geometric Scattering Modules

Alexander Tong

Frederik Wenkel

Dhananjay Bhaskar

Kincaid MacDonald

Jackson Grady

Michael Perlmutter

Smita Krishnaswamy

Guy Wolf