Connectionists: PhD & Postdoctoral positions in AI/ML

[Garg Vikas]

Hi all,

We’ve several research positions (PhD & Postdoctoral) available for two of our AI/ML based drug design initiatives. The work will augment and consolidate existing well-developed research pipelines of the supervisors.  Selected researchers may also have the opportunity to work with and visit our collaborators at leading pharma companies and academic groups in Europe, USA, and Canada.

Project 1: Physics-inspired geometric deep representation learning for drug design

Supervisors: Prof. Vikas Garg, Prof. Samuel Kaski, PhD Markus Heinonen (Department of Computer Science, Aalto University)

We invite applications for positions in geometric deep learning aimed at advancing state of the art in drug design. Key research directions include new 3D generative models for molecules that reflect their underlying physical-chemical processes and dynamics, spatial constraints, local invariances and equivariances, and energy considerations; domain generalization in both structural and sequence spaces; deep geometric models for inducing diversity in molecular generation; learning with limited training data; and non-autoregressive inference methods.

Facility in implementing deep learning models is expected, and training in one or more of the following would be a plus: Statistical Mechanics, Bayesian Learning, Graph Neural Networks, Generative Models, Ordinary and Partial Differential Equations, and Computational Biochemistry. Some representative publications are given below.


[1] John Ingraham, Vikas Garg, Regina Barzilay, Tommi Jaakkola. Generative Models for Graph-Based Protein Design<https://www.mit.edu/~vgarg/GenerativeModelsForProteinDesign.pdf>. NeurIPS (2019).
[2] Vikas Garg, Stefanie Jegelka, Tommi Jaakkola. Generalization and Representational Limits of Graph Neural Networks<https://www.mit.edu/~vgarg/GNNs_FinalVersion.pdf>. ICML (2020).
[3] Cagatay Yildiz, Markus Heinonen, Harri Lähdesmäki. ODE2VAE: Deep generative second order ODEs with Bayesian neural networks<https://arxiv.org/abs/1905.10994>. NeurIPS (2019)
[4] Pashupati Hegde, Markus Heinonen, Harri Lähdesmäki, Samuel Kaski. Deep learning with differential Gaussian process flows<https://arxiv.org/abs/1810.04066>. AISTATS (2019), notable paper award (top 1%)
[5] Kyle Barlow, Shane O Conchuir, Samuel Thompson, Pooja Suresh, James Lucas, Markus Heinonen, Tanja Kortemme. Flex ddG: Rosetta Ensemble-Based Estimation of Changes in Protein-Protein Binding Affinity Upon Mutation<https://pubs.acs.org/doi/10.1021/acs.jpcb.7b11367>. Journal of Physical Chemistry B, 122(21):5389-5399 (2018)

Project 2: Probabilistic modelling for collaborative human-in-the-loop design

Supervisors: Prof. Samuel Kaski, Prof. Vikas Garg, PhD Markus Heinonen (Department of Computer Science, Aalto University)

We are looking for a doctoral student interested in developing probabilistic modelling and inference methods needed for complex design tasks, with drug design as a case study. The idea is to help experts steer the modelling system towards their design goals, while eliciting their prior knowledge to improve the models of the drugs. This is difficult because the goals may be tacit, uncertain and evolving. Another postdoc will develop the new molecular models for drugs, to which you are welcome to contribute. This will be a transformative project, resulting in a virtual drug design laboratory.


Key methods we will need: probabilistic modelling and Bayesian inference, multi-agent modelling, sequential experimental design, POMDPs, reinforcement learning and inverse reinforcement learning. We expect applicants to master some of these, or be exceptionally eager and quick learners.

[1] Celikok et al. Teaching to Learn: Sequential Teaching of Agents with Inner State<https://arxiv.org/pdf/2009.06227.pdf>. arXiv:2009.06227
[2] Mikkola et al. Projective Preferential Bayesian Optimization<https://arxiv.org/pdf/2002.03113>. ICML 2020
[3] Peltola et al. Machine Teaching of Active Sequential Learners<https://aaltopml.github.io/machine-teaching-of-active-sequential-learners/>. NeurIPS 2019
[4] Kangasrääsiö et al. Parameter inference for computational cognitive models with Approximate Bayesian Computation<https://doi.org/10.1111/cogs.12738>. Cognitive Science 43 (2019): e12738.

Best regards,
Vikas

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20210707/fceabfa5/attachment.html>