Publications

Cross-ethnicity/race generalization failure of behavioral prediction from resting-state functional connectivity
Jingwei Li
Jianzhong Chen
Angela Tam
Leon Qi Rong Ooi
Leon Qi Rong Ooi
Avram J. Holmes
Tian Ge
Kaustubh R. Patil
Mbemba Jabbi
Simon B. Eickhoff
B. T. Thomas Yeo
Sarah Genon
Algorithmic biases that favor majority populations pose a key challenge to the application of machine learning for precision medicine. Here,… (voir plus) we assessed such bias in prediction models of behavioral phenotypes from brain functional magnetic resonance imaging. We examined the prediction bias using two independent datasets (preadolescent versus adult) of mixed ethnic/racial composition. When predictive models were trained on data dominated by white Americans (WA), out-of-sample prediction errors were generally higher for African Americans (AA) than for WA. This bias toward WA corresponds to more WA-like brain-behavior association patterns learned by the models. When models were trained on AA only, compared to training only on WA or an equal number of AA and WA participants, AA prediction accuracy improved but stayed below that for WA. Overall, the results point to the need for caution and further research regarding the application of current brain-behavior prediction models in minority populations.
Enjeux éthiques de l’IA en santé - Fiche 4
Joé T. Martineau
Frédérique Romy Godin
Janine Badr
Alexandre Castonguay
Martin Cousineau
Philippe Després
Aude Motulsky
Jean Noel Nikiema
Cecile Petitgand
La présente fiche propose une revue des différents enjeux éthiques liés au développement et à l’utilisation des technologies d’int… (voir plus)elligence artificielle dans le milieu de la santé, en trois parties. D’abord, nous aborderons les enjeux éthiques liés à l’exploitation de données massives nécessaires à l’entrainement des algorithmes de l’IA. Ensuite, nous présenterons les principaux enjeux éthiques liés au développement et à l’utilisation des SIA en santé, en abordant la façon dont ces systèmes impactent nos vies ainsi que l’environnement physique et social dans lequel nous vivons. Nous présenterons finalement les principales initiatives nationales et internationales en matière d’éthique de l’IA et de la gestion des données, fruits et reflets d’une réflexion globale sur ces sujets. Ces initiatives ont notamment proposé des lignes directrices et principes normatifs servant de guides pour le développement de technologies de l’IA éthiques et responsables Il s'agit de la quatrième fiche d'une série de 4 développée dans le cadre d'un mandat réalisé pour le Ministère de la Santé et des Services sociaux du Québec (MSSS).
Guidelines for the Computational Testing of Machine Learning approaches to Vehicle Routing Problems
Luca Accorsi
Andrea Lodi
Daniele Vigo
Interindividual Differences in Cortical Thickness and Their Genomic Underpinnings in Autism Spectrum Disorder.
Christine Ecker
Charlotte M. Pretzsch
Anke Bletsch
Caroline Mann
Tim Schaefer
Sara Ambrosino
Julian Tillmann
Afsheen Yousaf
Andreas Chiocchetti
Michael V. Lombardo
Varun Warrier
Nico Bast
Carolin Moessnang
Sarah Baumeister
Flavio Dell’Acqua
Dorothea L. Floris
Mariam Zabihi
Andre Marquand
Freddy Cliquet
Claire Leblond … (voir 19 de plus)
Clara A. Moreau
Nick Puts
Tobias Banaschewski
Emily J. H. Jones
Luke Mason
Sven Bölte
Andreas Meyer-Lindenberg
Antonio Persico
Sarah Durston
Simon Baron-Cohen
Will Spooren
Eva Loth
Christine M. Freitag
Tony Charman
Thomas Bourgeron
Christian Beckmann
Jan K. Buitelaar
Declan Murphy
JANOS: An Integrated Predictive and Prescriptive Modeling Framework
David Bergman
Teng Huang
Philip Brooks
Andrea Lodi
Arvind U. Raghunathan
Business research practice is witnessing a surge in the integration of predictive modeling and prescriptive analysis. We describe a modeling… (voir plus) framework JANOS that seamlessly integrates the two streams of analytics, allowing researchers and practitioners to embed machine learning models in an end-to-end optimization framework. JANOS allows for specifying a prescriptive model using standard optimization modeling elements such as constraints and variables. The key novelty lies in providing modeling constructs that enable the specification of commonly used predictive models within an optimization model, have the features of the predictive model as variables in the optimization model, and incorporate the output of the predictive models as part of the objective. The framework considers two sets of decision variables: regular and predicted. The relationship between the regular and the predicted variables is specified by the user as pretrained predictive models. JANOS currently supports linear regression, logistic regression, and neural network with rectified linear activation functions. In this paper, we demonstrate the flexibility of the framework through an example on scholarship allocation in a student enrollment problem and provide a numeric performance evaluation. Summary of Contribution. This paper describes a new software tool, JANOS, that integrates predictive modeling and discrete optimization to assist decision making. Specifically, the proposed solver takes as input user-specified pretrained predictive models and formulates optimization models directly over those predictive models by embedding them within an optimization model through linear transformations.
Multistep networks for roll force prediction in hot strip rolling mill
Shuhong Shen
Denzel Guye
Xiaoping Ma
Stephen Yue
Software-Engineering Design Patterns for Machine Learning Applications
Hironori Washizaki
Yann‐Gaël Guéhéneuc
Hironori Takeuchi
Naotake Natori
Takuo Doi
Satoshi Okuda
In this study, a multivocal literature review identified 15 software-engineering design patterns for machine learning applications. Findings… (voir plus) suggest that there are opportunities to increase the patterns’ adoption in practice by raising awareness of such patterns within the community.
A time-space formulation for the locomotive routing problem at the Canadian National Railways
Pedro L. Miranda
Jean-François Cordeau
Combining Modular Skills in Multitask Learning
Edoardo M. Ponti
A modular design encourages neural models to disentangle and recombine different facets of knowledge to generalise more systematically to ne… (voir plus)w tasks. In this work, we assume that each task is associated with a subset of latent discrete skills from a (potentially small) inventory. In turn, skills correspond to parameter-efficient (sparse / low-rank) model parameterisations. By jointly learning these and a task-skill allocation matrix, the network for each task is instantiated as the average of the parameters of active skills. To favour non-trivial soft partitions of skills across tasks, we experiment with a series of inductive biases, such as an Indian Buffet Process prior and a two-speed learning rate. We evaluate our latent-skill model on two main settings: 1) multitask reinforcement learning for grounded instruction following on 8 levels of the BabyAI platform; and 2) few-shot adaptation of pre-trained text-to-text generative models on CrossFit, a benchmark comprising 160 NLP tasks. We find that the modular design of a network significantly increases sample efficiency in reinforcement learning and few-shot generalisation in supervised learning, compared to baselines with fully shared, task-specific, or conditionally generated parameters where knowledge is entangled across tasks. In addition, we show how discrete skills help interpretability, as they yield an explicit hierarchy of tasks.
More Than Meets the Eye: Art Engages the Social Brain
Janneke E. P. van Leeuwen
Jeroen Boomgaard
Sebastian J. Crutch
Jason D. Warren
Quantitative electrophysiological assessments as predictive markers of lower limb motor recovery after spinal cord injury: a pilot study with an adaptive trial design
Yin Nan Huang
El-Mehdi Meftah
Charlotte H. Pion
Jean-Marc Mac-Thiong
Dorothy Barthélemy
Observational, cohort study. (1) Determine the feasibility and relevance of assessing corticospinal, sensory, and spinal pathways early aft… (voir plus)er traumatic spinal cord injury (SCI) in a rehabilitation setting. (2) Validate whether electrophysiological and magnetic resonance imaging (MRI) measures taken early after SCI could identify preserved neural pathways, which could then guide therapy. Intensive functional rehabilitation hospital (IFR). Five individuals with traumatic SCI and eight controls were recruited. The lower extremity motor score (LEMS), electrical perceptual threshold (EPT) at the S2 dermatome, soleus (SOL) H-reflex, and motor evoked potentials (MEPs) in the tibialis anterior (TA) muscle were assessed during the stay in IFR and in the chronic stage (>6 months post-SCI). Control participants were only assessed once. Feasibility criteria included the absence of adverse events, adequate experimental session duration, and complete dataset gathering. The relationship between electrophysiological data collected in IFR and LEMS in the chronic phase was studied. The admission MRI was used to calculate the maximal spinal cord compression (MSCC). No adverse events occurred, but a complete dataset could not be collected for all subjects due to set-up configuration limitations and time constraints. EPT measured at IFR correlated with LEMS in the chronic phases (r = −0.67), whereas SOL H/M ratio, H latency, MEPs and MSCC did not. Adjustments are necessary to implement electrophysiological assessments in an IFR setting. Combining MRI and electrophysiological measures may lead to better assessment of neuronal deficits early after SCI.
Stringency of containment and closures on the growth of SARS-CoV-2 in Canada prior to accelerated vaccine roll-out
David Vickers
Stefan Baral
Sharmistha Mishra
Jeffrey C. Kwong
Maria Sundaram
Alan Katz
Andrew Calzavara
Mathieu Maheu-Giroux
David L Buckeridge
Tyler Williamson