A5069876930- Advanced Bandit Algorithms Research
- Cell Image Analysis Techniques
- Image Processing Techniques and Applications
- Reinforcement Learning in Robotics
- Machine Learning and Algorithms
- Data Stream Mining Techniques
- Advanced Multi-Objective Optimization Algorithms
- Advanced Fluorescence Microscopy Techniques
- Machine Learning and Data Classification
- Pharmaceutical Practices and Patient Outcomes
- Adversarial Robustness in Machine Learning
- Bone health and osteoporosis research
- Anomaly Detection Techniques and Applications
- Advanced Electron Microscopy Techniques and Applications
- Traumatic Brain Injury Research
- Metaheuristic Optimization Algorithms Research
- Diabetes Management and Research
- Machine Learning in Healthcare
- Topic Modeling
- Genetic Associations and Epidemiology
- Diabetes Management and Education
- Diabetes Treatment and Management
- Data Mining Algorithms and Applications
- Machine Learning in Materials Science
- Pharmacy and Medical Practices
Université Laval
2015-2025
Mila - Quebec Artificial Intelligence Institute
2021-2024
Canadian Institute for Advanced Research
2021-2024
Centre Universitaire de Mila
2023
McGill University
2018-2021
Hôpital de l'Enfant-Jésus
1991
Traditional approaches for finding well-performing parameterizations of complex imaging systems, such as super-resolution microscopes rely on an extensive exploration phase over the illumination and acquisition settings, prior to task. This strategy suffers from several issues: it requires a large amount parameter configurations be evaluated, leads discrepancies between parameters in task, results waste time resources given that optimization final tasks are conducted separately. Here we show...
Since screening programs identify only a small proportion of the population as eligible for an intervention, genomic prediction heritable risk factors could decrease number needing to be screened by removing individuals at low genetic risk. We therefore tested whether polygenic score heel quantitative ultrasound speed sound (SOS)-a factor osteoporotic fracture-can low-risk who can safely excluded from fracture program.
In all, 80% of antenatal karyotypes are generated by Down's syndrome screening programmes (DSSP). After a positive screening, women offered prenatal foetus karyotyping, the gold standard. Reliable molecular methods for rapid aneuploidy diagnosis (RAD: fluorescence in situ hybridization (FISH) and quantitative PCR (QF-PCR)) can detect common aneuploidies, faster less expensive than karyotyping. UK, RAD is recommended as standalone approach DSSP, whereas US guidelines recommend that be...
Generative Adversarial Networks (GANs) can successfully approximate a probability distribution and produce realistic samples. However, open questions such as sufficient convergence conditions mode collapse still persist. In this paper, we build on existing work in the area by proposing novel framework for training generator against an ensemble of discriminator networks, which be seen one-student/multiple-teachers setting. We formalize problem within full-information adversarial bandit...
Aiming for more effective experiment design, such as in video content advertising where different options compete user engagement, these scenarios can be modeled multi-arm bandit problems. In cases limited interactions are available due to external factors, the cost of conducting experiments, recommenders often face constraints small number interactions. addition, there is a trade-off between selecting best treatment and ability personalize contextualize based on individual factors. A...
Multi-messenger observations of gravitational waves and electromagnetic emission from compact object mergers offer unique insights into the structure neutron stars, formation heavy elements, expansion rate Universe. With LIGO-Virgo-KAGRA (LVK) gravitational-wave detectors currently in their fourth observing run (O4), it is an exciting time for detecting these mergers. However, assessing whether to follow up a candidate event given limited telescope resources challenging; can be false alert...
The purpose of this study was to assess the feasibility deep learning (DL) methods enhance prediction visual acuity (VA) improvement after macular hole (MH) surgery from a combined model using DL on high-definition optical coherence tomography (HD-OCT) B-scans and clinical features.We trained convolutional neural network (CNN) pre-operative HD-OCT macula with logistic regression features predict VA increase ≥15 Early Treatment Diabetic Retinopathy Study (ETDRS) letters at 6 months...
Hyperparameter optimization is now widely applied to tune the hyperparameters of learning algorithms. The can have structure, resulting in depending on conditions, or values other hyperparameters. We target problem combined algorithm selection and hyperparameter optimization, which includes at least one conditional hyperparameter: choice algorithm. In this work, we show that Bayesian with Gaussian processes be used for spaces injection knowledge concerning conditions kernel. propose examine...
Major depressive disorder (MDD) is the leading cause of disability worldwide. Of individuals with MDD, 30% to 50% are unresponsive common antidepressants, highlighting untapped causal biological mechanisms. Dysfunction in microbiota-gut-brain axis has been implicated MDD pathogenesis. Exposure chronic stress disrupts blood-brain barrier integrity; still, little known about intestinal function these conditions, particularly for small intestine, where absorption most foods and drugs takes place.
Abstract A patient-level Markov decision model was used to simulate a virtual cohort of 500,000 women 40 years old and over, in relation osteoporosis-related hip, clinical vertebral, wrist bone fractures events. Sixteen different screening options three main scenario groups were compared: (1) the status quo (no specific national prevention program); (2) universal primary program; (3) treatment program based on 10-year absolute risk fracture. The outcomes measured total directs costs from...
The study sought to assess the clinical performance of a machine learning model aiming identify unusual medication orders.This prospective was conducted at CHU Sainte-Justine, Canada, from April August 2020. An unsupervised based on GANomaly and 2 baselines were trained learn order patterns 10 years data. Clinical pharmacists dichotomously (typical or atypical) labeled orders pharmacological profiles (patients' lists). Confusion matrices, areas under precision-recall curve (AUPRs), F1 scores...
Clinical guidelines for most adults with diabetes recommend maintaining hemoglobin A
Plants adapt over time to their surrounding conditions. We argue that no current plant simulator addresses responses an environmental stimulus through gamification and sequential decision making. propose the GrowSpace is built within a reinforcement learning framework, providing sandbox environment test out hypotheses. currently focuses on behavioral response of plants light displacement, with objective controlling plant's shape by moving source. The back-end implemented using Space...
In this paper, we propose a novel approach for the multi-objective optimization of classifier ensembles in ROC space. We first evolve pool simple classifiers with NSGA-II using values curves as objectives. These are then combined at decision level Iterative Boolean Combination method (IBC). This produces multiple optimized various operating conditions. perform rigorous series experiments to demonstrate properties and behaviour approach. allows us interesting venues future research on...
We consider the problem of streaming kernel regression, when observations arrive sequentially and goal is to recover underlying mean function, assumed belong an RKHS. The variance noise not be known. In this context, we tackle tuning regularization parameter adaptively at each time step, while maintaining tight confidence bounds estimates on value function point. To end, first generalize existing results for finite-dimensional linear regression with fixed known setup a allowed measurable...
The ability to discover approximately optimal policies in domains with sparse rewards is crucial applying reinforcement learning (RL) many real-world scenarios. Approaches such as neural density models and continuous exploration (e.g., Go-Explore) have been proposed maintain the high rate necessary find performing generalizable policies. Soft actor-critic(SAC) another method for improving that aims combine efficient via off-policy updates while maximizing policy entropy. In this work, we...
Background Genomics-based prediction could be useful since genome-wide genotyping costs less than many clinical tests. We tested whether machine learning methods provide a clinically-relevant genomic of quantitative ultrasound speed sound (SOS)—a risk factor for osteoporotic fracture. Methods used 341,449 individuals from UK Biobank with SOS measures to develop genomically-predicted (gSOS) using algorithms. selected the optimal algorithm in 5,335 independent and then validated it its ability...
We propose $\tt RandUCB$, a bandit strategy that builds on theoretically derived confidence intervals similar to upper bound (UCB) algorithms, but akin Thompson sampling (TS), it uses randomization trade off exploration and exploitation. In the $K$-armed setting, we show there are infinitely many variants of all which achieve minimax-optimal $\widetilde{O}(\sqrt{K T})$ regret after $T$ rounds. Moreover, for specific multi-armed both UCB TS can be recovered as special cases RandUCB$. For...