Search Results for author: Biswa Sengupta

Found 23 papers, 3 papers with code

Learning Multi-Agent Communication with Contrastive Learning

no code implementations3 Jul 2023 Yat Long Lo, Biswa Sengupta, Jakob Foerster, Michael Noukhovitch

By examining the relationship between messages sent and received, we propose to learn to communicate using contrastive learning to maximize the mutual information between messages of a given trajectory.

Contrastive Learning

Learning to Infer Belief Embedded Communication

no code implementations15 Mar 2022 Guo Ye, Han Liu, Biswa Sengupta

In multi-agent collaboration problems with communication, an agent's ability to encode their intention and interpret other agents' strategies is critical for planning their future actions.

Text Generation

Switch Trajectory Transformer with Distributional Value Approximation for Multi-Task Reinforcement Learning

no code implementations14 Mar 2022 Qinjie Lin, Han Liu, Biswa Sengupta

Our results also demonstrate the advantage of the switch transformer model for absorbing expert knowledge and the importance of value distribution in evaluating the trajectory.

reinforcement-learning Reinforcement Learning (RL)

The Multi-Agent Pickup and Delivery Problem: MAPF, MARL and Its Warehouse Applications

no code implementations14 Mar 2022 Tim Tsz-Kit Lau, Biswa Sengupta

We study two state-of-the-art solutions to the multi-agent pickup and delivery (MAPD) problem based on different principles -- multi-agent path-finding (MAPF) and multi-agent reinforcement learning (MARL).

Multi-Agent Path Finding Multi-agent Reinforcement Learning +2

Reinforcement Learning for Location-Aware Scheduling

no code implementations7 Mar 2022 Stelios Stavroulakis, Biswa Sengupta

Recent techniques in dynamical scheduling and resource management have found applications in warehouse environments due to their ability to organize and prioritize tasks in a higher temporal resolution.

Management reinforcement-learning +2

Learning to Ground Decentralized Multi-Agent Communication with Contrastive Learning

no code implementations7 Mar 2022 Yat Long Lo, Biswa Sengupta

For communication to happen successfully, a common language is required between agents to understand information communicated by one another.

Contrastive Learning

Hierarchically Structured Scheduling and Execution of Tasks in a Multi-Agent Environment

no code implementations6 Mar 2022 Diogo S. Carvalho, Biswa Sengupta

In this work, we set ourselves on a problem that presents itself with a hierarchical structure: the task-scheduling, by a centralised agent, in a dynamic warehouse multi-agent environment and the execution of one such schedule, by decentralised agents with only partial observability thereof.

Management reinforcement-learning +2

Neural dynamics under active inference: plausibility and efficiency of information processing

no code implementations22 Jan 2020 Lancelot Da Costa, Thomas Parr, Biswa Sengupta, Karl Friston

We then show that these neuronal dynamics approximate natural gradient descent, a well-known optimisation algorithm from information geometry that follows the steepest descent of the objective in information space.

How Robust are Deep Neural Networks?

no code implementations30 Apr 2018 Biswa Sengupta, Karl J. Friston

In this paper, we evaluate the robustness of three recurrent neural networks to tiny perturbations, on three widely used datasets, to argue that high accuracy does not always mean a stable and a robust (to bounded perturbations, adversarial attacks, etc.)

Distributed non-parametric deep and wide networks

no code implementations ICLR 2018 Biswa Sengupta, Yu Qian

In recent work, it was shown that combining multi-kernel based support vector machines (SVMs) can lead to near state-of-the-art performance on an action recognition dataset (HMDB-51 dataset).

Action Recognition Gaussian Processes +1

LatentPoison -- Adversarial Attacks On The Latent Space

no code implementations ICLR 2018 Antonia Creswell, Biswa Sengupta, Anil A. Bharath

Robustness and security of machine learning (ML) systems are intertwined, wherein a non-robust ML system (classifiers, regressors, etc.)

General Classification Reinforcement Learning (RL)

Adversarial Information Factorization

1 code implementation ICLR 2019 Antonia Creswell, Yumnah Mohamied, Biswa Sengupta, Anil A. Bharath

We propose a novel generative model architecture designed to learn representations for images that factor out a single attribute from the rest of the representation.

Attribute Facial Attribute Classification +2

LatentPoison - Adversarial Attacks On The Latent Space

1 code implementation8 Nov 2017 Antonia Creswell, Anil A. Bharath, Biswa Sengupta

Robustness and security of machine learning (ML) systems are intertwined, wherein a non-robust ML system (classifiers, regressors, etc.)

General Classification reinforcement-learning +1

GeoSeq2Seq: Information Geometric Sequence-to-Sequence Networks

no code implementations ICLR 2018 Alessandro Bay, Biswa Sengupta

The Fisher information metric is an important foundation of information geometry, wherein it allows us to approximate the local geometry of a probability distribution.

Image Captioning Translation

Generative Adversarial Networks: An Overview

2 code implementations19 Oct 2017 Antonia Creswell, Tom White, Vincent Dumoulin, Kai Arulkumaran, Biswa Sengupta, Anil A. Bharath

Generative adversarial networks (GANs) provide a way to learn deep representations without extensively annotated training data.

General Classification Image Generation +2

StackSeq2Seq: Dual Encoder Seq2Seq Recurrent Networks

no code implementations11 Oct 2017 Alessandro Bay, Biswa Sengupta

A widely studied non-deterministic polynomial time (NP) hard problem lies in finding a route between the two nodes of a graph.

Translation

Approximating meta-heuristics with homotopic recurrent neural networks

no code implementations7 Sep 2017 Alessandro Bay, Biswa Sengupta

We show the viability of recurrent neural network solutions on a graph that has over 300 nodes and argue that a sequence-to-sequence network rather than other recurrent networks has improved approximation quality.

Reinforcement Learning (RL)

Pillar Networks++: Distributed non-parametric deep and wide networks

no code implementations18 Aug 2017 Biswa Sengupta, Yu Qian

In recent work, it was shown that combining multi-kernel based support vector machines (SVMs) can lead to near state-of-the-art performance on an action recognition dataset (HMDB-51 dataset).

Action Recognition Gaussian Processes +1

Multi-kernel learning of deep convolutional features for action recognition

no code implementations21 Jul 2017 Biswa Sengupta, Yu Qian

Image understanding using deep convolutional network has reached human-level performance, yet a closely related problem of video understanding especially, action recognition has not reached the requisite level of maturity.

Action Recognition Temporal Action Localization +1

Approximate Bayesian inference as a gauge theory

no code implementations17 May 2017 Biswa Sengupta, Karl Friston

In a published paper [Sengupta, 2016], we have proposed that the brain (and other self-organized biological and artificial systems) can be characterized via the mathematical apparatus of a gauge theory.

Bayesian Inference Variational Inference

Energetically Optimal Action Potentials

no code implementations NeurIPS 2011 Martin B. Stemmler, Biswa Sengupta, Simon Laughlin, Jeremy Niven

For a given height of an action potential, the least energy is consumed when the underlying currents obey the bang-bang principle: the currents giving rise to the spike should be intense, yet short-lived, yielding spikes with sharp onsets and offsets.

Cannot find the paper you are looking for? You can Submit a new open access paper.