Can a Bayesian Oracle Prevent Harm from an Agent?
Yoshua Bengio
Michael K. Cohen
Nikolay Malkin
Matt MacDermott
Damiano Fornasiere
Pietro Greiner
Younesse Kaddar
Is there a way to design powerful AI systems based on machine learning methods that would satisfy probabilistic safety guarantees? With the … (see more)long-term goal of obtaining a probabilistic guarantee that would apply in every context, we consider estimating a context-dependent bound on the probability of violating a given safety specification. Such a risk evaluation would need to be performed at run-time to provide a guardrail against dangerous actions of an AI. Noting that different plausible hypotheses about the world could produce very different outcomes, and because we do not know which one is right, we derive bounds on the safety violation probability predicted under the true but unknown hypothesis. Such bounds could be used to reject potentially dangerous actions. Our main results involve searching for cautious but plausible hypotheses, obtained by a maximization that involves Bayesian posteriors over hypotheses. We consider two forms of this result, in the iid case and in the non-iid case, and conclude with open problems towards turning such theoretical results into practical AI guardrails.
2024-08-09
ArXiv (preprint)
doi.org
openreview.net
Cell Morphology-Guided Small Molecule Generation with GFlowNets
Stephen Zhewen Lu
Ziqing Lu
Ehsan Hajiramezanali
Tommaso Biancalani
Yoshua Bengio
Gabriele Scalia
Michał Koziarski
High-content phenotypic screening, including high-content imaging (HCI), has gained popularity in the last few years for its ability to char… (see more)acterize novel therapeutics without prior knowledge of the protein target. When combined with deep learning techniques to predict and represent molecular-phenotype interactions, these advancements hold the potential to significantly accelerate and enhance drug discovery applications. This work focuses on the novel task of HCI-guided molecular design. Generative models for molecule design could be guided by HCI data, for example with a supervised model that links molecules to phenotypes of interest as a reward function. However, limited labeled data, combined with the high-dimensional readouts, can make training these methods challenging and impractical. We consider an alternative approach in which we leverage an unsupervised multimodal joint embedding to define a latent similarity as a reward for GFlowNets. The proposed model learns to generate new molecules that could produce phenotypic effects similar to those of the given image target, without relying on pre-annotated phenotypic labels. We demonstrate that the proposed method generates molecules with high morphological and structural similarity to the target, increasing the likelihood of similar biological activity, as confirmed by an independent oracle model.
2024-08-09
ArXiv (preprint)
doi.org
arxiv.org
Revisiting Feature Prediction for Learning Visual Representations from Video
Adrien Bardes
Quentin Garrido
Jean Ponce
Xinlei Chen
Michael Rabbat
Yann LeCun
Mahmoud Assran
Nicolas Ballas
2024-08-09
TMLR (accepted)
doi.org
openreview.net
Cardinality Minimization, Constraints, and Regularization: A Survey
Andreas M. Tillmann
Daniel Bienstock
Andrea Lodi
Alexandra Schwartz
We survey optimization problems that involve the cardinality of variable vectors in constraints or the objective function. We provide a unif… (see more)ied viewpoint on the general problem classes and models, and give concrete examples from diverse application fields such as signal and image processing, portfolio selection, or machine learning. The paper discusses general-purpose modeling techniques and broadly applicable as well as problem-specific exact and heuristic solution approaches. While our perspective is that of mathematical optimization, a main goal of this work is to reach out to and build bridges between the different communities in which cardinality optimization problems are frequently encountered. In particular, we highlight that modern mixed-integer programming, which is often regarded as impractical due to commonly unsatisfactory behavior of black-box solvers applied to generic problem formulations, can in fact produce provably high-quality or even optimal solutions for cardinality optimization problems, even in large-scale real-world settings. Achieving such performance typically draws on the merits of problem-specific knowledge that may stem from different fields of application and, e.g., shed light on structural properties of a model or its solutions, or lead to the development of efficient heuristics; we also provide some illustrative examples.
2024-08-08
SIAM Review (published)
doi.org
arxiv.org
Stochastic Wiring of Cell Types Enhances Fitness by Generating Phenotypic Variability
Divyansha Lachi
Ann Huang
Augustine N. Mavor-Parker
Arna Ghosh
Blake Richards
Anthony Zador
The development of neural connectivity is a crucial biological process that gives rise to diverse brain circuits and behaviors. Neural devel… (see more)opment is a stochastic process, but this stochasticity is often treated as a nuisance to overcome rather than as a functional advantage. Here we use a computational model, in which connection probabilities between discrete cell types are genetically specified, to investigate the benefits of stochasticity in the development of neural wiring. We show that this model can be viewed as a generalization of a powerful class of artificial neural networks—Bayesian neural networks—where each network parameter is a sample from a distribution. Our results reveal that stochasticity confers a greater benefit in large networks and variable environments, which may explain its role in organisms with larger brains. Surprisingly, we find that the average fitness over a population of agents is higher than a single agent defined by the average connection probability. Our model reveals how developmental stochasticity, by inducing a form of non-heritable phenotypic variability, can increase the probability that at least some individuals will survive in rapidly changing, unpredictable environments. Our results suggest how stochasticity may be an important feature rather than a bug in neural development.
2024-08-08
bioRxiv (preprint)
doi.org
Critical dynamics in spontaneous EEG predict anesthetic-induced loss of consciousness and perturbational complexity
Charlotte Maschke
Jordan O’Byrne
Michele Angelo Colombo
Melanie Boly
Olivia Gosseries
Steven Laureys
Mario Rosanova
Karim Jerbi
Stefanie Blain-Moraes
2024-08-05
Communications Biology (published)
doi.org
Critical dynamics in spontaneous EEG predict anesthetic-induced loss of consciousness and perturbational complexity
Charlotte Maschke
Jordan O’Byrne
Michele Angelo Colombo
Melanie Boly
Olivia Gosseries
Steven Laureys
Mario Rosanova
Karim Jerbi
Stefanie Blain-Moraes
2024-08-05
Communications Biology (published)
doi.org
Critical dynamics in spontaneous EEG predict anesthetic-induced loss of consciousness and perturbational complexity
Charlotte Maschke
Jordan O’Byrne
Michele Angelo Colombo
Melanie Boly
Olivia Gosseries
Steven Laureys
Mario Rosanova
Karim Jerbi
Stefanie Blain-Moraes
2024-08-05
Communications Biology (published)
doi.org
Learning Hybrid Interpretable Models: Theory, Taxonomy, and Methods
Julien Ferry
gabriel laberge
Ulrich Aivodji
A hybrid model involves the cooperation of an interpretable model and a complex black box. At inference, any input of the hybrid model is as… (see more)signed to either its interpretable or complex component based on a gating mechanism. The advantages of such models over classical ones are two-fold: 1) They grant users precise control over the level of transparency of the system and 2) They can potentially perform better than a standalone black box since redirecting some of the inputs to an interpretable model implicitly acts as regularization. Still, despite their high potential, hybrid models remain under-studied in the interpretability/explainability literature. In this paper, we remedy this fact by presenting a thorough investigation of such models from three perspectives: Theory, Taxonomy, and Methods. First, we explore the theory behind the generalization of hybrid models from the Probably-Approximately-Correct (PAC) perspective. A consequence of our PAC guarantee is the existence of a sweet spot for the optimal transparency of the system. When such a sweet spot is attained, a hybrid model can potentially perform better than a standalone black box. Secondly, we provide a general taxonomy for the different ways of training hybrid models: the Post-Black-Box and Pre-Black-Box paradigms. These approaches differ in the order in which the interpretable and complex components are trained. We show where the state-of-the-art hybrid models Hybrid-Rule-Set and Companion-Rule-List fall in this taxonomy. Thirdly, we implement the two paradigms in a single method: HybridCORELS, which extends the CORELS algorithm to hybrid modeling. By leveraging CORELS, HybridCORELS provides a certificate of optimality of its interpretable component and precise control over transparency. We finally show empirically that HybridCORELS is competitive with existing hybrid models, and performs just as well as a standalone black box (or even better) while being partly transparent.
2024-08-05
TMLR (accepted)
doi.org
openreview.net
The effect of gestational age on short- and long-term complications following primary esophageal atresia repair
Mathias Johansen
Samuel Wasserman
Dan Poenaru
Jean Martin Laberge
Sam J. Daniel
Thomas Engelhardt
2024-08-03
Brazilian Journal of Anesthesiology (published)
doi.org
Are self-explanations from Large Language Models faithful?
Andreas Madsen
Sarath Chandar
Siva Reddy
2024-08-01
Findings of the Association for Computational Linguistics ACL 2024 (published)
doi.org
arxiv.org
Enhancing Changepoint Detection: Penalty Learning through Deep Learning Techniques
Tung L. Nguyen
Toby Dylan Hocking
Changepoint detection, a technique for identifying significant shifts within data sequences, is crucial in various fields such as finance, g… (see more)enomics, medicine, etc. Dynamic programming changepoint detection algorithms are employed to identify the locations of changepoints within a sequence, which rely on a penalty parameter to regulate the number of changepoints. To estimate this penalty parameter, previous work uses simple models such as linear or tree-based models. This study introduces a novel deep learning method for predicting penalty parameters, leading to demonstrably improved changepoint detection accuracy on large benchmark supervised labeled datasets compared to previous methods.
2024-08-01
ArXiv (preprint)
arxiv.org
arxiv.org