My current research interests are primarily in Adversarial Machine Learning applied to different domains such as Images, Text and most recently Graphs. Specifically, I’m interested in all things related to GAN’s, adversarial attacks and adversarial disentanglement. More broadly, I’m interested in understanding how to build better machine learning models that can operate in the world around us and serve humans better. This entails understanding the data generative process, the biases that are reflected within the data and most importantly understanding how these models fail.

Invited talks

  • Adversarial Attacks on Face Detectors using Neural Net based Constrained Optimization
    PDF (Best Paper Nominee IEEE MMSP)

  • Adversarial Contrastive Estimation.
    (Association of Computational Linguistics 2018) ORAL
    PDF

Publications and Preprints

  • Generalizable Adversarial Attacks using Generative Models
    (Submitted to NeurIPS 2019)
    Avishek Joey Bose, Andre Cianflone, William Hamilton
    PDF
  • Improving Exploration in Soft-Actor-Critic with Normalizing Flows Policies
    (ICML 2019 Invertible Neural Networks and Normalizing Flows Workshop)
    Patrick Nadeem Ward, Ariella Smofsky, Avishek Joey Bose
  • A Cross-Domain Transferable Neural Coherence Model
    (Association of Computational Linguistics 2019)
    Xu Peng, Hamidreza Saghir, JinSung Kang, Teng Long, Avishek Joey Bose,
    Yanshuai Cao, Jackie Cheung
  • Compositional Fairness Constraints for Graph Embeddings
    (International Conference on Machine Learning 2019)
    Avishek Joey Bose, William Hamilton
    PDF
  • Discrete off-policy policy gradient using continuous relaxations
    (Reinforcement Learning and Decision Making 2019)
    Andre Cianflone, Zafarali Ahmed, Riashat Islam, Avishek Joey Bose, William Hamilton
    PDF
  • Compositional Hard Negative Mining for Visual Semantic Embeddings via an Adversary
    (NeurIPS 2018 Visually Grounded Language and Interaction Workshop)
    Avishek Joey Bose, Huan Ling, Yanshuai Cao
    PDF
  • Compositional Fairness Constraints for Graph Embeddings
    (NeurIPS 2018 Relational Representational Learning Workshop)
    Avishek Joey Bose, William Hamilton
    PDF
  • Adversarial Attacks on Face Detectors using Neural Net based Constrained Optimization
    Avishek Joey Bose, Parham Aarabi
    PDF (IEEE MMSP) Best Paper Nominee
  • Adversarial Contrastive Estimation.
    (Association of Computational Linguistics 2018) ORAL
    Avishek Joey Bose, Huan Ling, Yanshuai Cao
    PDF
  • Walking through sight: Seeing the ability to see, in a 3-D augmediated reality environment.
    (IEEE GEM 2014)
    Ryan Janzen, Seyed Nima Yasrebi, Avishek Joey Bose, Arjun Subramanian, Steve Mann
    PDF

Course and other Projects

  • Active Learning through Adversarial Exploration in Contrastive Objectives
    CSC2547 Winter 2018
    PDF
  • Improving Visual Semantic Embedding By Adversarial Contrastive Estimation
    CSC2548 Winter 2018
    PDF
  • Crafting Adversarial Attacks with Adversarial Transformations
    CSC2541 Fall 2017
    PDF
  • An Evaluation of 2D Face Alignment models under extreme lighting conditions and blur.
    PDF