Chaitanya K. Joshi
Chaitanya K. Joshi
Home
Publications
Blog
Contact
CV
Light
Dark
Automatic
Graph Neural Networks
Recent Advances in Deep Learning for Routing Problems
We characterize recent advances in deep learning for routing problems via a unified Neural Combinatorial Optimization pipeline and provide new directions to stimulate future research.
(Also published at the inaugural ICLR 2022 Blog Track)
Chaitanya K. Joshi
,
Rishabh Anand
Jan 12, 2022
20 min read
Cite
Slides
Video
ICLR Blog Track
Related Paper
Tweet
Learning TSP Requires Rethinking Generalization
We study zero-shot generalization to large-scale instances in neural network-driven solvers for the Travelling Salesman Problem: what architectures, inductive biases and learning paradigms enable better generalization?
(CP 2021)
Chaitanya K. Joshi
,
Quentin Cappart
,
Louis-Martin Rousseau
,
Thomas Laurent
PDF
Cite
Code
Slides
Video
DOI
Blog
Learning TSP Requires Rethinking Generalization
This talk discusses our recent work on deep learning for TSP and the challenge of zero-shot generalization for large-scale and real-world routing problems.
Jun 8, 2021 12:00 AM
CORS 2021 (Host: Dr. Maxime Gasse)
Chaitanya K. Joshi
Slides
Video
Multi-Graph Transformer for Free-Hand Sketch Recognition
Representation learning for free-hand drawings using GNNs and Transformers.
(IEEE TNNLS)
Peng Xu
,
Chaitanya K. Joshi
,
Xavier Bresson
PDF
Cite
Code
DOI
Blog
Graph Neural Networks: Benchmarks and Future Directions
I discuss state-of-the-art Graph Neural Network architectures and introduce our recent work on Benchmarking GNNs, along with some interesting future directions.
Sep 24, 2020 12:00 AM
DSO National Laboratories (Host: Dr. Chieu Hai Leong)
Chaitanya K. Joshi
Slides
Benchmarking Graph Neural Networks
Open-source benchmarking framework to identify scalable and powerful GNN architectures, and track the progress of graph representation learning research.
(500+ citations, 2000+ GitHub stars)
Vijay Prakash Dwivedi
,
Chaitanya K. Joshi
,
Luu Anh Tuan
,
Thomas Laurent
,
Yoshua Bengio
,
Xavier Bresson
PDF
Cite
Code
Slides
Press
JMLR
Transformers are Graph Neural Networks
Exploring the connection between Transformer models such as GPT and BERT for Natural Language Processing, and Graph Neural Networks.
(80,000+ readers on The Gradient, featured in Probabilistic ML textbook and classes at Stanford, Cambridge, Oxford.)
Chaitanya K. Joshi
Last updated on Jun 21, 2021
12 min read
Cite
Slides
The Gradient
Towards Data Science
Stanford CS224W
Cambridge L45
Probabilistic ML textbook
Graph Neural Networks for the Travelling Salesman Problem
This talk introduces a recent line of work using Graph Neural Networks to directly ‘learn’ good heuristics for TSP in an end-to-end manner.
Oct 22, 2019 12:00 AM
INFORMS Annual Meeting 2019 (Host: Prof. Quentin Cappart)
Chaitanya K. Joshi
Slides
On Learning Paradigms for the Travelling Salesman Problem
How do learning paradigms impact zero-shot generalization to large-scale instances in learning-driven TSP solvers?
(NeurIPS 2019 Workshop)
Chaitanya K. Joshi
,
Thomas Laurent
,
Xavier Bresson
PDF
Cite
Code
Poster
An Efficient Graph Convolutional Network Technique for the Travelling Salesman Problem
Deep Graph ConvNets paired with parallelized graph search can learn TSP up to few hundred cities, but fall short of classical solvers.
Chaitanya K. Joshi
,
Thomas Laurent
,
Xavier Bresson
PDF
Cite
Code
Slides
«
»
Cite
×