A5054812796- Machine Learning and Algorithms
- Machine Learning and Data Classification
- Domain Adaptation and Few-Shot Learning
- Muscle activation and electromyography studies
- Imbalanced Data Classification Techniques
- Genomics and Phylogenetic Studies
- Advanced Bandit Algorithms Research
- Machine Learning in Bioinformatics
- Topic Modeling
- Neural Networks and Applications
- Statistical Methods and Inference
- Adversarial Robustness in Machine Learning
- Gene expression and cancer classification
- Hand Gesture Recognition Systems
- Advanced Graph Theory Research
- Algorithms and Data Compression
- Bayesian Modeling and Causal Inference
- Machine Learning and ELM
- Formal Methods in Verification
- Reinforcement Learning in Robotics
- Gaussian Processes and Bayesian Inference
- Advanced Sensor and Energy Harvesting Materials
- Computability, Logic, AI Algorithms
- vaccines and immunoinformatics approaches
- Biomedical Text Mining and Ontologies
Université Laval
2014-2023
Mila - Quebec Artificial Intelligence Institute
2021
Max Planck Institute for Intelligent Systems
2011
Max Planck Society
2011
Montanuniversität Leoben
2011
Saarland University
2008
University of Windsor
2008
University of Regina
2008
Concordia University
2008
Concordia University
2008
The process of generating raw genome sequence data continues to become cheaper, faster, and more accurate. However, assembly such into high-quality, finished sequences remains challenging. Many tools are available, but they differ greatly in terms their performance (speed, scalability, hardware requirements, acceptance newer read technologies) final output (composition assembled sequence). More importantly, it largely unclear how best assess the quality sequences. Assemblathon competitions...
In recent years, deep learning algorithms have become increasingly more prominent for their unparalleled ability to automatically learn discriminant features from large amounts of data. However, within the field electromyography-based gesture recognition, are seldom employed as they require an unreasonable amount effort a single person, generate tens thousands examples. This paper's hypothesis is that general, informative can be learned data generated by aggregating signals multiple users,...
Abstracta Voluminous parallel sequencing datasets, especially metagenomic experiments, require distributed computing for de novo assembly and taxonomic profiling. Ray Meta is a massively metagenome assembler that coupled with Communities, which profiles microbiomes based on uniquely-colored k-mers. It can accurately assemble profile three billion read experiment representing 1,000 bacterial genomes of uneven proportions in 15 hours 1,024 processor cores, using only 1.5 GB per core. The...
An accurate genome sequence of a desired species is now pre-requisite for research. important step in obtaining high-quality to correctly assemble short reads into longer sequences accurately representing contiguous genomic regions. Current sequencing technologies continue offer increases throughput, and corresponding reductions cost time. Unfortunately, the benefit large number complicated by errors, with different biases being observed each platform. Although software are available...
We introduce a new representation learning algorithm suited to the context of domain adaptation, in which data at training and test time come from similar but different distributions. Our is directly inspired by theory on adaptation suggesting that, for effective transfer be achieved, predictions must made based that cannot discriminate between (source) (target) domains. propose objective implements this idea neural network, whose hidden layer trained predictive classification task,...
Untargeted metabolomic measurements using mass spectrometry are a powerful tool for uncovering new small molecules with environmental and biological importance. The molecule identification step, however, still remains an enormous challenge due to fragmentation difficulties or unspecific fragment ion information. Current methods address this often dependent on databases require the use of nuclear magnetic resonance (NMR), which have their own difficulties. gas-phase collision cross section...
We present a general PAC-Bayes theorem from which all known risk bounds are obtained as particular cases. also propose different learning algorithms for finding linear classifiers that minimize these bounds. These generally competitive with both AdaBoost and the SVM.
In the realm of surface electromyography (sEMG) gesture recognition, deep learning algorithms are seldom employed. This is due in part to large quantity data required for them train on. Consequently, it would be prohibitively time consuming a single user generate sufficient amount training such algorithms. this paper, two datasets 18 and 17 able-bodied participants respectively recorded using low-cost, low-sampling rate (200Hz), 8-channel, consumer-grade, dry electrode sEMG device named Myo...
We introduce a new representation learning approach for domain adaptation, in which data at training and test time come from similar but different distributions. Our is directly inspired by the theory on adaptation suggesting that, effective transfer to be achieved, predictions must made based features that cannot discriminate between (source) (target) domains. The implements this idea context of neural network architectures are trained labeled source unlabeled target (no target-domain...
The identification of genomic biomarkers is a key step towards improving diagnostic tests and therapies. We present reference-free method for this task that relies on k-mer representation genomes machine learning algorithm produces intelligible models. computationally scalable well-suited whole genome sequencing studies.The was validated by generating models predict the antibiotic resistance C. difficile, M. tuberculosis, P. aeruginosa, S. pneumoniae 17 antibiotics. obtained are accurate,...
Recently, robotics has been seen as a key solution to improve the quality of life amputees. In order create smarter robotic prosthetic devices be used in an everyday context, one must able interface them seamlessly with end-user inexpensive, yet reliable way. this paper, we are looking at guiding device by detecting gestures through measurement electrical activity muscles captured surface electromyography (sEMG). Reliable sEMG-based gesture classifiers for end-users challenging design, they...
Existing research on myoelectric control systems primarily focuses extracting discriminative characteristics of the electromyographic (EMG) signal by designing handcrafted features. Recently,however, deep learning techniques have been applied to challenging task EMG-based gesture recognition. The adoption these slowly shifts focus from feature engineering learning. Nevertheless, black-box nature makes it hard understand type information learned network and how relates Additionally, due high...
Understanding the relationship between genome of a cell and its phenotype is central problem in precision medicine. Nonetheless, genotype-to-phenotype prediction comes with great challenges for machine learning algorithms that limit their use this setting. The high dimensionality data tends to hinder generalization scalability most algorithms. Additionally, produce models are complex difficult interpret. We alleviate these limitations by proposing strong performance guarantees, based on...
We tackle the problem of non robustness simulation and bisimulation when dealing with probabilistic processes. It is important to ignore tiny deviations in probabilities because these often come from experiments or estimations. A few approaches have been proposed treat this issue, for example metrics quantify bisimilarity (or closeness) Relaxing definition another avenue which we follow. define a new semantics known simple logic processes show that it characterises notion epsi-simulation....
Wearable technology can be employed to elevate the abilities of humans perform demanding and complex tasks more efficiently. Armbands capable surface electromyography (sEMG) are attractive noninvasive devices from which human intent derived by leveraging machine learning. However, sEMG acquisition systems currently available tend prohibitively costly for personal use or sacrifice wearability signal quality affordable. This work introduces 3DC Armband designed Biomedical Microsystems...
Despite their superior performance, deep learning models often lack interpretability. In this paper, we explore the modeling of insightful relations between words, in order to understand and enhance predictions. To effect, propose Self-Attention Network (SANet), a flexible interpretable architecture for text classification. Experiments indicate that gains obtained by self-attention is task-dependent. For instance, experiments on sentiment analysis tasks showed an improvement around 2% when...
The discovery of peptides possessing high biological activity is very challenging due to the enormous diversity for which only a minority have desired properties. To lower cost and reduce time obtain promising peptides, machine learning approaches can greatly assist in process even partly replace expensive laboratory experiments by predictor with existing data or smaller amount generation. Unfortunately, once model learned, selecting having greatest predicted bioactivity often requires...