top of page

Research Projects

Screenshot 2023-01-12 at 4.56.45 PM.png

Social Inference from Relational Visual Information

Advisors: Dr. Leyla Isik  [Link]

We hypothesize that humans rely on relational visual information in particular, which is lacking from standard neural networks, and develop a new relational, graph neural network model, SocialGNN. We find that SocialGNN accurately predicts human interaction judgments across both animated and natural videos, suggesting that humans can make complex social interaction judgments without explicit simulation or inference about agents’ mental states, and that structured, relational visual representations are key to this behavior.


BCQ4DCA: Budget Constrained Deep Q-Network for Dynamic Campaign Allocation in Computational Advertising

TCS Research and Innovation   [Link]

We developed a deep reinforcement learning model for dynamic optimization of budget-constrained campaign allocation.


Network Analysis of Neuro-Cognitive Processes: Studying McGurk Effect using EEG Data

Advisors: Dr. Arpan Banerjee, Dr. Ganesh Bagler, Dr. Dipanjan Roy   [Link]

The research work involved understanding multi-sensory perception involving auditory and visual cues using the McGurk effect. My focus in this research was to understand the network properties of the brain using EEG data obtained from multiple subjects.

Selected Presentations
  • Talk: Manasi Malik, Leyla Isik. Social Inference from Relational Visual Information, Vision Sciences Society (VSS ’22), Florida, USA

  • Poster: Manasi Malik, Leyla Isik, Human Social Interaction Judgements are Uniquely Explained by both
    Bottom-up Graph Neural Networks and Generative Inverse Planning Models, Johns Hopkins AI-X
    Foundry Fall 2023 Symposium, Baltimore, USA

  • Poster: Manasi Malik, Leyla Isik Both Purely Visual and Simulation-based Models Uniquely Explain Human
    Social Interaction Judgements, Journal of Vision 2023;23(9):5111.

bottom of page