Michael Bronstein Journalist Profile

Articles

Legalizing Natural Marijuana Is The Pathway To Protect Public Health, Not Pretending Synthetic THC Products Are Hemp (Op-Ed)

Oct 27, 2024 | marijuanamoment.net | Michael Bronstein

“Misleading marketing, which frames synthetic THC as a ‘hemp product,’ tricks buyers into thinking they are purchasing something natural and safe.”By Michael Bronstein, American Trade Association for Cannabis and Hemp (ATACH)In a recent op-ed by the Reason Foundation titled “Banning Hemp Products Isn’t The Way to Protect Public Health,” the author correctly makes the case that bans are often ineffective and can push consumers toward more dangerous, unregulated markets.
Clarifying Congress' intent for THC necessary for marijuana reform

Sep 6, 2024 | mjbizdaily.com | Michael Bronstein |Kate Lavin

(This contributed guest column is part of the cover package in the September-October issue of MJBizMagazine. To be considered as an MJBizDaily guest columnist, please submit your request here.)The U.S. Congress is caught up in a political haze over what to do with hemp. In 2018, Congress was told hemp laws were for “rope, not dope.”But the industry’s hemp debate today is entirely about dope.
The Road to Biology 2.0 Will Pass Through Black-Box Data

Mar 18, 2024 | towardsdatascience.com | Michael Bronstein

AI-first BiotechThis year marks perhaps the zenith of expectations for AI-based breakthroughs in biology, transforming it into an engineering discipline that is programmable, predictable, and replicable. Drawing insights from AI breakthroughs in perception, natural language, and protein structure prediction, we endeavour to pinpoint the characteristics of biological problems that are most conducive to being solved by AI techniques.
Combinatorial prediction of therapeutic perturbations using causally-inspired neural networks

Jan 8, 2024 | biorxiv.org | Guadalupe Gonzalez |Guadalupe E González |Isuru Herath |Kirill Veselkov |Michael Bronstein

AbstractAs an alternative to target-driven drug discovery, phenotype-driven approaches identify compounds that counteract the overall disease effects by analyzing phenotypic signatures. Our study introduces a novel approach to this field, aiming to expand the search space for new therapeutic agents. We introduce PDGrapher, a causally-inspired graph neural network model designed to predict arbitrary perturbagens - sets of therapeutic targets - capable of reversing disease effects.
Combinatorial prediction of therapeutic perturbations using causally-inspired neural networks

Jan 3, 2024 | biorxiv.org | Guadalupe Gonzalez |Guadalupe E González |Kirill Veselkov |Michael Bronstein |Isuru Herath

AbstractPhenotype-driven drug discovery, as an emerging alternative to target-driven strategies, identifies compounds that counteract the overall effects of diseases by analyzing phenotypic signatures. Our study introduces a novel approach to this field, aiming to expand the search space for new therapeutic agents. We introduce PDGrapher, a causally-inspired graph neural network model designed to predict arbitrary perturbagens - a set of therapeutic targets - capable of reversing disease effects.
Co-operative Graph Neural Networks

Dec 5, 2023 | towardsdatascience.com | Michael Bronstein

New GNN architectureA large majority of Graph Neural Networks (GNNs) follow the message-passing paradigm, where node states are updated based on aggregated neighbour messages.
Dynamically rewired delayed message passing GNNs

Jun 19, 2023 | towardsdatascience.com | Michael Bronstein

How I Turned My Company’s Docs into a Searchable Database with OpenAIAnd how you can do the same with your docs
Dynamically rewired delayed message passing GNNs

Jun 19, 2023 | towardsdatascience.com | Michael Bronstein

Classical message-passing graph neural networks (MPNNs) operate by aggregating information from 1-hop neighbours of every node. Consequently, learning tasks requiring long-range interactions (i.e., there exists a node v whose representation needs to account for the information contained in some node u at shortest-walk (geodesic) distance d( u, v) = r > 1) require deep MPNNs with multiple message-passing layers.
Direction Improves Graph Learning

Jun 8, 2023 | towardsdatascience.com | Michael Bronstein

GNNs on Directed GraphsDirection Improves Graph LearningHow a wise use of direction when doing message passing on heterophilic graphs can result in very significant gains. Graph Neural Networks (GNNs) are highly effective at modelling relational data. However, current GNN models frequently assume the input graph to be undirected, overlooking the inherent directionality of many real-world graphs, such as social, transportation, transaction, and citation networks. In this blog post, we…
Hyperbolic Deep Reinforcement Learning

Apr 30, 2023 | towardsdatascience.com | Michael Bronstein

RL meets hyperbolic geometryMany RL problems have hierarchical tree-like nature. Hyperbolic geometry offers a powerful prior for such problems. Many problems in Reinforcement Learning manifest a hierarchical tree-like nature. Hyperbolic spaces, which can be conceptualised as continuous analogies of trees, are thus suitable candidates to parameterise the agent’s deep model.