Publications

Electric Power Fuse Identification With Deep Learning
Simon Giard-Leroux
Guillaume Cléroux
Shreyas Sunil Kulkarni
François Bouffard
Martin Vallières
As part of arc flash studies, survey pictures of electrical installations need to be manually analyzed. A challenging task is to identify fu… (voir plus)se types, which can be determined from physical characteristics, such as shape, color, and size. To automate this process using deep learning techniques, a new dataset of fuse pictures from past arc flash projects and data from the web was created. Multiple experiments were performed to train a final model, reaching an average precision of 91.06% on the holdout set, which confirms its potential for identification of fuse types in new photos. By identifying fuse types using physical characteristics only, the need to take clear pictures of the label text is eliminated, allowing pictures to be taken away from danger, thereby improving the safety of workers. All the resources needed to repeat the experiments are openly accessible, including the code and datasets.
2023-10-31
IEEE Transactions on Industrial Informatics (publié)
doi.org
Integrating Equity, Diversity, and Inclusion Throughout the Lifecycle of Artificial Intelligence for Better Health and Oral Health Care: A Workshop Summary.
Elham Emami
S. A. Rahimi
Milka Nyariro
Professors Elham Emami and Samira Rahimi organized and co-led an international interdisciplinary workshop in June 2023 at McGill University,… (voir plus) built upon an intersectoral approach addressing equity, diversity and inclusion within the field of AI.
2023-10-31
Journal (publié)
pubmed.ncbi.nlm.nih.gov
Invasive Brain Computer Interface for Motor Restoration in Spinal Cord Injury: A Systematic Review.
Jordan J. Levett
Lior M. Elkaim
Farbod Niazi
Michael H. Weber
Christian Iorio-Morin
Marco Bonizzato
Alexander G. Weil
2023-10-31
Neuromodulation: Technology at the Neural Interface (publié)
doi.org
Investigation of the Dosimetry Characteristics of the GAFCHROMIC® EBT3 Film Response to Alpha Particle Irradiation
Mélodie Cyr
Victor D. Martinez
S. Devic
Nada Tomic
David F. Lewis
S. Enger
2023-10-31
Journal of Physics: Conference Series (publié)
doi.org
Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting
Kashif Rasul
Arjun Ashok
Andrew Robert Williams
Hena Ghonia
Rishika Bhagwatkar
Arian Khorasani
Mohammad Javad Darvishi Bayazi
George Adamopoulos
Roland Riachi
Nadhir Hassen
Marin Biloš
Sahil Garg
Anderson Schneider
Nicolas Chapados
Alexandre Drouin
Valentina Zantedeschi
Yuriy Nevmyvaka
Irina Rish
Over the past years, foundation models have caused a paradigm shift in machine learning due to their unprecedented capabilities for zero-sho… (voir plus)t and few-shot generalization. However, despite the success of foundation models in modalities such as natural language processing and computer vision, the development of foundation models for time series forecasting has lagged behind. We present Lag-Llama, a general-purpose foundation model for univariate probabilistic time series forecasting based on a decoder-only transformer architecture that uses lags as covariates. Lag-Llama is pretrained on a large corpus of diverse time series data from several domains, and demonstrates strong zero-shot generalization capabilities compared to a wide range of forecasting models on downstream datasets across domains. Moreover, when fine-tuned on relatively small fractions of such previously unseen datasets, Lag-Llama achieves state-of-the-art performance, outperforming prior deep learning approaches, emerging as the best general-purpose model on average. Lag-Llama serves as a strong contender to the current state-of-art in time series forecasting and paves the way for future advancements in foundation models tailored to time series data.
2023-10-31
R0-FoMo @ Neural Information Processing Systems (poster)
doi.org
openreview.net
MasakhaNEWS: News Topic Classification for African languages
David Ifeoluwa Adelani
Marek Masiak
Israel Abebe Azime
Jesujoba Alabi
Atnafu Lambebo Tonja
Christine Mwase
Odunayo Ogundepo
Bonaventure F. P. Dossou
Akintunde Oladipo
Doreen Nixdorf
Chris Chinenye Emezue
sana al-azzawi
Blessing Sibanda
Davis David
Lolwethu Ndolela
Jonathan Mukiibi
Tunde Ajayi
Tatiana Moteu
Brian Odhiambo
Abraham Owodunni … (voir 45 de plus)
Nnaemeka Obiefuna
Muhidin Mohamed
Shamsuddeen Hassan Muhammad
Teshome Mulugeta Ababu
Saheed Abdullahi Salahudeen
Mesay Gemeda Yigezu
Tajuddeen Gwadabe
Idris Abdulmumin
Mahlet Taye
Oluwabusayo Awoyomi
Iyanuoluwa Shode
Tolulope Adelani
Habiba Abdulganiyu
Abdul-Hakeem Omotayo
Adetola Adeeko
Abeeb Afolabi
Anuoluwapo Aremu
Olanrewaju Samuel
Clemencia Siro
Wangari Kimotho
Onyekachi Ogbu
Chinedu Mbonu
Chiamaka Chukwuneke
Samuel Fanijo
Jessica Ojo
Oyinkansola Awosan
Tadesse Kebede
Toadoum Sari Sakayo
Pamela Nyatsine
Freedmore Sidume
Oreen Yousuf
Mardiyyah Oduwole
Tshinu Tshinu
Ussen Kimanuka
Thina Diko
Siyanda Nxakama
Sinodos Nigusse
Abdulmejid Johar
Shafie Mohamed
Fuad Mire Hassan
Moges Ahmed Mehamed
Evrard Ngabire
Jules Jules
Ivan Ssenkungu
Pontus Stenetorp
African languages are severely under-represented in NLP research due to lack of datasets covering several NLP tasks. While there are individ… (voir plus)ual language specific datasets that are being expanded to different tasks, only a handful of NLP tasks (e.g. named entity recognition and machine translation) have standardized benchmark datasets covering several geographical and typologically-diverse African languages. In this paper, we develop MasakhaNEWS -- a new benchmark dataset for news topic classification covering 16 languages widely spoken in Africa. We provide an evaluation of baseline models by training classical machine learning models and fine-tuning several language models. Furthermore, we explore several alternatives to full fine-tuning of language models that are better suited for zero-shot and few-shot learning such as cross-lingual parameter-efficient fine-tuning (like MAD-X), pattern exploiting training (PET), prompting language models (like ChatGPT), and prompt-free sentence transformer fine-tuning (SetFit and Cohere Embedding API). Our evaluation in zero-shot setting shows the potential of prompting ChatGPT for news topic classification in low-resource African languages, achieving an average performance of 70 F1 points without leveraging additional supervision like MAD-X. In few-shot setting, we show that with as little as 10 examples per label, we achieved more than 90\% (i.e. 86.0 F1 points) of the performance of full supervised training (92.6 F1 points) leveraging the PET approach.
2023-10-31
International Joint Conference on Natural Language Processing and the Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (publié)
doi.org
openreview.net
Mining Mass Spectra for Peptide Facts
Jeremie Zumer
S. Lemieux
The current mainstream software for peptide-centric tandem mass spectrometry data analysis can be categorized as either database-driven, whi… (voir plus)ch rely on a library of mass spectra to identify the peptide associated with novel query spectra, or de novo sequencing-based, which aim to find the entire peptide sequence by relying only on the query mass spectrum. While the first paradigm currently produces state-of-the-art results in peptide identification tasks, it does not inherently make use of information present in the query mass spectrum itself to refine identifications. Meanwhile, de novo approaches attempt to solve a complex problem in one go, without any search space constraints in the general case, leading to comparatively poor results. In this paper, we decompose the de novo problem into putatively easier subproblems, and we show that peptide identification rates of database-driven methods may be improved in terms of peptide identification rate by solving one such subsproblem without requiring a solution for the complete de novo task. We demonstrate this using a de novo peptide length prediction task as the chosen subproblem. As a first prototype, we show that a deep learning-based length prediction model increases peptide identification rates in the ProteomeTools dataset as part of an Pepid-based identification pipeline. Using the predicted information to better rank the candidates, we show that combining ideas from the two paradigms produces clear benefits in this setting. We propose that the next generation of peptide-centric tandem mass spectrometry identification methods should combine elements of these paradigms by mining facts “de novo; about the peptide represented in a spectrum, while simultaneously limiting the search space with a peptide candidates database.
2023-10-31
bioRxiv (prépublication)
doi.org
Open design of a reproducible videogame controller for MRI and MEG
Yann Harel
André Cyr
Julie Boyle
Basile Pinsard
Jeremy Bernard
Marie-France Fourcade
Himanshu Aggarwal
Ana Fernanda Ponce
Bertrand Thirion
Karim Jerbi
Pierre Bellec
Videogames are emerging as a promising experimental paradigm in neuroimaging. Acquiring gameplay in a scanner remains challenging due to the… (voir plus) lack of a scanner-compatible videogame controller that provides a similar experience to standard, commercial devices. In this paper, we introduce a videogame controller designed for use in the functional magnetic resonance imaging as well as magnetoencephalography. The controller is made exclusively of 3D-printed and commercially available parts. We evaluated the quality of our controller by comparing it to a non-MRI compatible controller that was kept outside the scanner. The comparison of response latencies showed reliable button press accuracies of adequate precision. Comparison of the subjects’ motion during fMRI recordings of various tasks showed that the use of our controller did not increase the amount of motion produced compared to a regular MR compatible button press box. Motion levels during an ecological videogame task were of moderate amplitude. In addition, we found that the controller only had marginal effect on temporal SNR in fMRI, as well as on covariance between sensors in MEG, as expected due to the use of non-magnetic building materials. Finally, the reproducibility of the controller was demonstrated by having team members who were not involved in the design build a reproduction using only the documentation. This new videogame controller opens new avenues for ecological tasks in fMRI, including challenging videogames and more generally tasks with complex responses. The detailed controller documentation and build instructions are released under an Open Source Hardware license to increase accessibility, and reproducibility and enable the neuroimaging research community to improve or modify the controller for future experiments.
2023-10-31
PLOS ONE (publié)
doi.org
SAGE: Smart home Agent with Grounded Execution
Dmitriy Rivkin
Francois Hogan
Amal Feriani
Abhisek Konar
Adam Sigal
Steve Liu
Gregory Dudek
2023-10-31
ArXiv (prépublication)
doi.org
arxiv.org
Spatial Distribution Modeling of Pistacia atlantica using Artificial Neural Network in Khohir National Park
Shahab Bakhtiari
Tymour Rostani Shahraji
Reza Akhavan
Reza Ebrahimi Atani
2023-10-31
Ecology of Iranian Forests (publié)
doi.org
Tuning Minimum-Norm regularization parameters for optimal MEG connectivity estimation
Elisabetta Vallarino
Ana Sofia Hincapié
Karim Jerbi
Richard M. Leahy
Annalisa Pascarella
Alberto Sorrentino
Sara Sommariva
The regularization parameter of the Minimum Norm Estimate of neural activity impacts connectivity estimationWe study empirically the optimal… (voir plus) parameter for connectivity estimation using realistic synthetic datasetsWe find the optimal parameter for connectivity estimation is systematically smaller than the optimal parameter for source imaging; different connectivity metrics yield the same resultCode and data are available open source.
2023-10-31
NeuroImage (publié)
doi.org
Adaptive Resolution Residual Networks
Léa Demeule
Mahtab Sandhu
Glen Berseth
We introduce Adaptive Resolution Residual Networks (ARRNs), a form of neural operator that enables the creation of networks for signal-based… (voir plus) tasks that can be rediscretized to suit any signal resolution. ARRNs are composed of a chain of Laplacian residuals that each contain ordinary layers, which do not need to be rediscretizable for the whole network to be rediscretizable. ARRNs have the property of requiring a lower number of Laplacian residuals for exact evaluation on lower-resolution signals, which greatly reduces computational cost. ARRNs also implement Laplacian dropout, which encourages networks to become robust to low-bandwidth signals. ARRNs can thus be trained once at high-resolution and then be rediscretized on the fly at a suitable resolution with great robustness.
2023-10-30
NeurIPS.cc/2023/Workshop/DLDE (spotlight)
openreview.net
openreview.net