Vaisakh Shaj

Postdoctoral Researcher at the University Of Edinburgh, UK

avatar.JPG

Room 234 , Informatics Forum

Edinburgh, Scotland

United Kingdom EH89AB

Welcome to my personal website. I am a Postdoctoral Researcher in AI and Machine Learning at the University Of Edinburgh working with Prof Amos Storkey. I am affiliated with the Institute Of Adaptive and Neural Computation, School Of Informatics.

I did my PhD in Machine Learning and Robotics under Prof Gerhard Neumann at the ALR Lab, Karlsruhe Institute Of Technology. Before that I was a Data Scientist at the cybersecurity firm McAfee (Intel Security). Previously I worked with Intel for 2 years. I hold a post graduate degree in Machine Learning and Computing from the Indian Institute of Space Science and Technology.

My doctoral thesis focussed on building “World models With Hierarchical Temporal Abstractions” based on probabilistic and Bayesian principles. My primary interests lie in both understanding human intelligence and replicating its capabilities within AI agents, using mathematical and computational tools. In this regard I am primarily intersted in sub-areas on Machine Learning like Meta/Continual Leanring (Learning Under Non-Stationarity), Hierarchical Modelling, Model Based Reinforcement Learning and Planning. I employ neural network architectures grounded in probabilistic principles to develop these models.

🌈Diversity and Inclusion Statement: I care deeply about making work places more diverse and the inclusion of women, LGBTQ+ and under-represented minorities in AI research.

news

Oct 11, 2024 Our new paper on Adaptive World Models and Non-Stationary RL accepted at NeurIPS 2024 Adaptive Foundation Models Workshop.
Dec 9, 2023 Attending NeurIPS 2023 in New Orleans, US.
Sep 21, 2023 PhD work “Multi Time Scale World Models” accepted in NeurIPS 2023 as a Spotlight (Top 3% of all submitted papers).
Jan 21, 2022 PhD work “Hidden Parameter Recurrent State Space Models For Changing Dynamics Scenarios” accepted in ICLR 2022.
Oct 1, 2020 PhD work “Action Conditional Recurrent Kalman Networks For Forward and Inverse Dynamics Learning” accepted in Conference on Robot Learning(CoRL) 2020. Video

latest posts

selected publications

2024

  1. multiFig.png
    Learning World Models With Hierarchical Temporal Abstractions: A Probabilistic Perspective
    Vaisakh Shaj
    PhD Thesis preprint arXiv:2404.16078, 2024

2023

  1. mts3.png
    Multi Time Scale World Models
    Vaisakh Shaj, Saleh Gholam Zadeh, Ozan Demir, and 2 more authors
    2023

2021

  1. hiprssm4.png
    Hidden Parameter Recurrent State Space Models For Changing Dynamics Scenarios
    Vaisakh Shaj, Dieter Büchler, Rohit Sonker, and 2 more authors
    In International Conference on Learning Representations, 2021

2020

  1. acrkn1.png
    Action-Conditional Recurrent Kalman Networks For Forward and Inverse Dynamics Learning
    Vaisakh Shaj, Philipp Becker, Dieter Büchler, and 5 more authors
    In Conference on Robot Learning, 2020

2019

  1. icml2019.png
    Zero-shot knowledge distillation in deep networks (ICML 2019)
    Vaisakh Shaj*, Gaurav Kumar Nayak*, Konda Reddy Mopuri*, and 2 more authors
    In International Conference on Machine Learning, 2019