Skip to content

Drew Hudson | From Machine Learning to Machine Reasoning

Photo of
Hosted By
Martin G.


Virtual London Machine Learning Meetup - 12.10.20 @ 18:30

We would like to invite you to our next Virtual Machine Learning Meetup. We are taking the opportunity to change the format slightly and devote more time to Q&A. Please read the papers below and help us create a vibrant discussion.

The discussion will be facilitated by Vivek Natarajan, a researcher at Google working at the intersection of AI and healthcare. Vivek’s current research involves improving the data efficiency, robustness, generalization, safety and fairness of AI models in healthcare.

- 18:25: Virtual doors open
- 18:30: Talk
- 19:00: Q&A session
- 19:35: Close

*Sponsors* : Machines that Read - Intelligent data extraction from corporate and financial documents.

* Title: From Machine Learning to Machine Reasoning (Drew Hudson is a Ph.D. student at Stanford University)

* Papers:
Learning by Abstraction: The Neural State Machine
Compositional Attention Networks for Machine Reasoning
GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering

Abstract: In this talk, I will discuss compositionality and reasoning and the visual domain, and introduce two models we have developed: MAC and the Neural State Machine, which seek to integrate the competing views of neural and symbolic AI and leverage their complementary strengths for the task of visual reasoning. These models build semantic representations that combine both the visual and linguistic modalities, and then perform sequential computation over them to answer new questions or reach novel conclusions, while achieving transparency and modularity. We have evaluated these models on multiple VQA datasets such as CLEVR and GQA which involve multi-step inference and diverse reasoning skills, achieving state-of-the-art results in both cases. I will present further experiments that illustrate the models' generalization capacity across multiple dimensions, including novel compositions of concepts, changes in the answer distribution, and unseen linguistic structures, to demonstrate the efficiency of these approaches.

Bio: Drew Hudson is a Ph.D. student at Stanford University, where she's been working with Professors Christopher Manning and Jay Mcclelland. Her research focuses on compositionality and reasoning at the intersection of computer vision, deep learning, and natural language processing. Her projects were published at the ICLR, NeurIPS and CVPR conferences, where she introduced new datasets and models for real-world visual reasoning, sequential inference and question answering. She is a recipient of the Stanford SoE Graduate Fellowship, the Chais' Scholarship for Excellence, and the Google Anita Borg Scholarship for leading women in Computer Science.
Online event
This event has passed