Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for VLN

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=lqeeoqlaDiw


Duration: 1:21
1,182 views
19

Vision-Language Navigation is the task of navigating an embodied agent to carry out natural language instructions inside real 3D environments. We propose a novel Reinforced Cross-Modal Matching (RCM) approach that enforces cross-modal grounding both locally and globally via reinforcement learning (RL) and further introduce a Self-Supervised Imitation Learning (SIL) method to explore unseen environments by imitating its own past, good decisions.



Other Videos By Microsoft Research


2019-07-03Physics ∩ ML Workshop (Day 1): Short Talks
2019-07-03Higher Fidelity Systems for Online Discussion
2019-07-03The Blessings of Multiple Causes
2019-07-03Data Efficient Reinforcement learning for Autonomous Robots with Simulated and Off-policy Data
2019-06-26The brave new world of cloud-scale systems and networking with Dr. Lidong Zhou
2019-06-19Designing and Building Multi-Device Interfaces
2019-06-19On The Hardness of Reinforcement Learning With Value-Function Approximation
2019-06-19Lattice Attacks for Variants of LWE
2019-06-19Models as Code: Differentiable Programming with Zygote
2019-06-19Game on with Dr. Chris Bishop and Phil Spencer
2019-06-18Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for VLN
2019-06-17Eyes First – Match Two
2019-06-17Eyes First – Tile Slide
2019-06-17Eyes First – Maze
2019-06-17Eyes First Games - Double Up
2019-06-17Eyes First Games - Getting Started
2019-06-12All Data AI with Dr. Andrew Fitzgibbon
2019-06-11VNLA via Imitation Learning with Indirect Intervention
2019-06-07Installing Microsoft SEAL on Windows
2019-06-07Dissecting Racial Bias in an Algorithm that Guides Health Decisions for Millions
2019-06-06Installing Microsoft SEAL on Linux & macOS


Tags:
microsoft research