Portrait of Guillaume Lajoie

Guillaume Lajoie

Core Academic Member
Canada CIFAR AI Chair
Associate Professor, Université de Montréal
Visiting Researcher, Google
Research Topics
AI for Science
AI in Health
Cognition
Computational Neuroscience
Deep Learning
Dynamical Systems
Optimization
Reasoning
Recurrent Neural Networks
Representation Learning

Biography

Guillaume Lajoie is an Associate professor in the Department of Mathematics and Statistics at Université de Montréal and a Core Academic Member of Mila – Quebec Artificial Intelligence Institute. He holds a Canada-CIFAR AI Research Chair, and a Canada Research Chair (CRC) in Neural Computation and Interfacing.

His research is positioned at the intersection of AI and Neuroscience where he develops tools to better understand mechanisms of intelligence common to both biological and artificial systems. His research group's contributions range from advances in multi-scale learning paradigms for large artificial systems, to applications in neurotechnology. Dr. Lajoie is actively involved in responsible AI development efforts, seeking to identify guidelines and best practices for use of AI in research and beyond.

Current Students

Postdoctorate - Université de Montréal
Collaborating Alumni - Polytechnique Montréal
PhD - Université de Montréal
Co-supervisor :
Independent visiting researcher - Université de Montréal
Co-supervisor :
PhD - Université de Montréal
PhD - Université de Montréal
Principal supervisor :
PhD - Université de Montréal
Postdoctorate - McGill University
Principal supervisor :
Independent visiting researcher - McGill University
PhD - McGill University
Principal supervisor :
Collaborating researcher - Université de Montréal
Co-supervisor :
PhD - McGill University
Principal supervisor :
Research Intern - Concordia University
Co-supervisor :
PhD - Université de Montréal
Co-supervisor :
PhD - Université de Montréal
Co-supervisor :
Independent visiting researcher - Université de Montréal
Master's Research - Université de Montréal
Master's Research - Université de Montréal
Principal supervisor :
PhD - Université de Montréal
Principal supervisor :
PhD - Université de Montréal
Co-supervisor :
Postdoctorate - Université de Montréal

Publications

Transfer Entropy Bottleneck: Learning Sequence to Sequence Information Transfer
Pascal Fortier-Poisson
Blake A. Richards
When presented with a data stream of two statistically dependent variables, predicting the future of one of the variables (the target stream… (see more)) can benefit from information about both its history and the history of the other variable (the source stream). For example, fluctuations in temperature at a weather station can be predicted using both temperatures and barometric readings. However, a challenge when modelling such data is that it is easy for a neural network to rely on the greatest joint correlations within the target stream, which may ignore a crucial but small information transfer from the source to the target stream. As well, there are often situations where the target stream may have previously been modelled independently and it would be useful to use that model to inform a new joint model. Here, we develop an information bottleneck approach for conditional learning on two dependent streams of data. Our method, which we call Transfer Entropy Bottleneck (TEB), allows one to learn a model that bottlenecks the directed information transferred from the source variable to the target variable, while quantifying this information transfer within the model. As such, TEB provides a useful new information bottleneck approach for modelling two statistically dependent streams of data in order to make predictions about one of them.
Steerable Equivariant Representation Learning
Willie McClinton
Tongzhou Wang
Chen Sun
Phillip Isola
Dilip Krishnan
Pre-trained deep image representations are useful for post-training tasks such as classification through transfer learning, image retrieval,… (see more) and object detection. Data augmentations are a crucial aspect of pre-training robust representations in both supervised and self-supervised settings. Data augmentations explicitly or implicitly promote invariance in the embedding space to the input image transformations. This invariance reduces generalization to those downstream tasks which rely on sensitivity to these particular data augmentations. In this paper, we propose a method of learning representations that are instead equivariant to data augmentations. We achieve this equivariance through the use of steerable representations. Our representations can be manipulated directly in embedding space via learned linear maps. We demonstrate that our resulting steerable and equivariant representations lead to better performance on transfer learning and robustness: e.g. we improve linear probe top-1 accuracy by between 1% to 3% for transfer; and ImageNet-C accuracy by upto 3.4%. We further show that the steerability of our representations provides significant speedup (nearly 50x) for test-time augmentations; by applying a large number of augmentations for out-of-distribution detection, we significantly improve OOD AUC on the ImageNet-C dataset over an invariant representation.
« Que notre cerveau soit constitué de neurones n’est pas un accident »
Roman Ikonicoff
Multi-view manifold learning of human brain-state trajectories.
Erica L. Busch
Andrew Benz
Tom Wallenstein
Nicholas B. Turk-Browne
The complexity of the human brain gives the illusion that brain activity is intrinsically high-dimensional. Nonlinear dimensionality-reducti… (see more)on methods such as uniform manifold approximation and t-distributed stochastic neighbor embedding have been used for high-throughput biomedical data. However, they have not been used extensively for brain activity data such as those from functional magnetic resonance imaging (fMRI), primarily due to their inability to maintain dynamic structure. Here we introduce a nonlinear manifold learning method for time-series data—including those from fMRI—called temporal potential of heat-diffusion for affinity-based transition embedding (T-PHATE). In addition to recovering a low-dimensional intrinsic manifold geometry from time-series data, T-PHATE exploits the data’s autocorrelative structure to faithfully denoise and unveil dynamic trajectories. We empirically validate T-PHATE on three fMRI datasets, showing that it greatly improves data visualization, classification, and segmentation of the data relative to several other state-of-the-art dimensionality-reduction benchmarks. These improvements suggest many potential applications of T-PHATE to other high-dimensional datasets of temporally diffuse processes.
NEURAL MANIFOLDS AND GRADIENT-BASED ADAPTATION IN NEURAL-INTERFACE TASKS
. Neural activity tends to reside on manifolds whose dimension is much lower than the dimension of the whole neural state space. Experiments… (see more) using brain-computer interfaces with microelectrode arrays implanted in the motor cortex of nonhuman primates tested the hypothesis that external perturbations should produce different adaptation strategies depending on how “aligned” the perturbation is with respect to a pre-existing intrinsic manifold. On the one hand, perturbations within the manifold (WM) evoked fast reassociations of existing patterns for rapid adaptation. On the other hand, perturbations outside the manifold (OM) triggered the slow emergence of new neural patterns underlying a much slower—and, without adequate training protocols, inconsistent or virtually impossible—adaptation. This suggests that the time scale and the overall difficulty of the brain to adapt depend fundamentally on the structure of neural activity. Here, we used a simplified static Gaussian model to show that gradient-descent learning could explain the differences between adaptation to WM and OM perturbations. For small learning rates, we found that the adaptation speeds were different but the model eventually adapted to both perturbations. Moreover, sufficiently large learning rates could entirely prohibit adaptation to OM perturbations while preserving adaptation to WM perturbations, in agreement with experiments. Adopting an incremental training protocol, as has been done in experiments, permitted a swift recovery of a full adaptation in the cases where OM perturbations were previously impossible to relearn. Finally, we also found that gradient descent was compatible with the reassociation mechanism on short adaptation time scales. Since gradient descent has many biologically plausible variants, our findings thus establish gradient-based learning as a plausible mechanism for adaptation under network-level constraints, with a central role for the learning rate.
Lazy vs hasty: linearization in deep networks impacts learning schedule based on example difficulty
Among attempts at giving a theoretical account of the success of deep neural networks, a recent line of work has identified a so-called `laz… (see more)y' training regime in which the network can be well approximated by its linearization around initialization. Here we investigate the comparative effect of the lazy (linear) and feature learning (non-linear) regimes on subgroups of examples based on their difficulty. Specifically, we show that easier examples are given more weight in feature learning mode, resulting in faster training compared to more difficult ones. In other words, the non-linear dynamics tends to sequentialize the learning of examples of increasing difficulty. We illustrate this phenomenon across different ways to quantify example difficulty, including c-score, label noise, and in the presence of easy-to-learn spurious correlations. Our results reveal a new understanding of how deep networks prioritize resources across example difficulty.
Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules
Yuhan Helena Liu
Blake A. Richards
Eric Shea-Brown
To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurre… (see more)nt neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize.
Is a Modular Architecture Enough?
Inspired from human cognition, machine learning systems are gradually revealing advantages of sparser and more modular architectures. Recent… (see more) work demonstrates that not only do some modular architectures generalize well, but they also lead to better out-of-distribution generalization, scaling properties, learning speed, and interpretability. A key intuition behind the success of such systems is that the data generating system for most real-world settings is considered to consist of sparsely interacting parts, and endowing models with similar inductive biases will be helpful. However, the field has been lacking in a rigorous quantitative assessment of such systems because these real-world data distributions are complex and unknown. In this work, we provide a thorough assessment of common modular architectures, through the lens of simple and known modular data distributions. We highlight the benefits of modularity and sparsity and reveal insights on the challenges faced while optimizing modular systems. In doing so, we propose evaluation metrics that highlight the benefits of modularity, the regimes in which these benefits are substantial, as well as the sub-optimality of current end-to-end learned modular systems as opposed to their claimed potential.
A connectomics-based taxonomy of mammals
Laura E Suarez
Yossi Yovel
Martijn P van den Heuvel
Olaf Sporns
Yaniv Assaf
Bratislav Misic
Mammalian taxonomies are conventionally defined by morphological traits and genetics. How species differ in terms of neural circuits and whe… (see more)ther inter-species differences in neural circuit organization conform to these taxonomies is unknown. The main obstacle to the comparison of neural architectures has been differences in network reconstruction techniques, yielding species-specific connectomes that are not directly comparable to one another. Here, we comprehensively chart connectome organization across the mammalian phylogenetic spectrum using a common reconstruction protocol. We analyse the mammalian MRI (MaMI) data set, a database that encompasses high-resolution ex vivo structural and diffusion MRI scans of 124 species across 12 taxonomic orders and 5 superorders, collected using a unified MRI protocol. We assess similarity between species connectomes using two methods: similarity of Laplacian eigenspectra and similarity of multiscale topological features. We find greater inter-species similarities among species within the same taxonomic order, suggesting that connectome organization reflects established taxonomic relationships defined by morphology and genetics. While all connectomes retain hallmark global features and relative proportions of connection classes, inter-species variation is driven by local regional connectivity profiles. By encoding connectomes into a common frame of reference, these findings establish a foundation for investigating how neural circuits change over phylogeny, forging a link from genes to circuits to behaviour.
Learning Shared Neural Manifolds from Multi-Subject fMRI Data
Erica Busch
Tom Wallenstein
Michal Gerasimiuk
Andrew Benz
Nicholas Turk-Browne
Functional magnetic resonance imaging (fMRI) is a notoriously noisy measurement of brain activity because of the large variations between in… (see more)dividuals, signals marred by environmental differences during collection, and spatiotemporal averaging required by the measurement resolution. In addition, the data is extremely high dimensional, with the space of the activity typically having much lower intrinsic dimension. In order to understand the connection between stimuli of interest and brain activity, and analyze differences and commonalities between subjects, it becomes important to learn a meaningful embedding of the data that denoises, and reveals its intrinsic structure. Specifically, we assume that while noise varies significantly between individuals, true responses to stimuli will share common, low-dimensional features between subjects which are jointly discoverable. Similar approaches have been exploited previously but they have mainly used linear methods such as PCA and shared response modeling (SRM). In contrast, we propose a neural network called MRMD-AE (manifold-regularized multiple decoder, autoencoder), that learns a common embedding from multiple subjects in an experiment while retaining the ability to decode to individual raw fMRI signals. We show that our learned common space represents an extensible manifold (where new points not seen during training can be mapped), improves the classification accuracy of stimulus features of unseen timepoints, as well as improves cross-subject translation of fMRI signals. We believe this framework can be used for many downstream applications such as guided brain-computer interface (BCI) training in the future.
Multi-scale Feature Learning Dynamics: Insights for Double Descent
A key challenge in building theoretical foundations for deep learning is the complex optimization dynamics of neural networks, resulting fro… (see more)m the high-dimensional interactions between the large number of network parameters. Such non-trivial dynamics lead to intriguing behaviors such as the phenomenon of "double descent" of the generalization error. The more commonly studied aspect of this phenomenon corresponds to model-wise double descent where the test error exhibits a second descent with increasing model complexity, beyond the classical U-shaped error curve. In this work, we investigate the origins of the less studied epoch-wise double descent in which the test error undergoes two non-monotonous transitions, or descents as the training time increases. By leveraging tools from statistical physics, we study a linear teacher-student setup exhibiting epoch-wise double descent similar to that in deep neural networks. In this setting, we derive closed-form analytical expressions for the evolution of generalization error over training. We find that double descent can be attributed to distinct features being learned at different scales: as fast-learning features overfit, slower-learning features start to fit, resulting in a second descent in test error. We validate our findings through numerical experiments where our theory accurately predicts empirical findings and remains consistent with observations in deep neural networks.
On Neural Architecture Inductive Biases for Relational Tasks
Current deep learning approaches have shown good in-distribution generalization performance, but struggle with out-of-distribution generaliz… (see more)ation. This is especially true in the case of tasks involving abstract relations like recognizing rules in sequences, as we find in many intelligence tests. Recent work has explored how forcing relational representations to remain distinct from sensory representations, as it seems to be the case in the brain, can help artificial systems. Building on this work, we further explore and formalize the advantages afforded by 'partitioned' representations of relations and sensory details, and how this inductive bias can help recompose learned relational structure in newly encountered settings. We introduce a simple architecture based on similarity scores which we name Compositional Relational Network (CoRelNet). Using this model, we investigate a series of inductive biases that ensure abstract relations are learned and represented distinctly from sensory data, and explore their effects on out-of-distribution generalization for a series of relational psychophysics tasks. We find that simple architectural choices can outperform existing models in out-of-distribution generalization. Together, these results show that partitioning relational representations from other information streams may be a simple way to augment existing network architectures' robustness when performing out-of-distribution relational computations.