Future Industrial Operations with Machine Learning

Описание: This talk will discuss decision making for industrial operations with Self-Supervised Learning (SSL) and Reinforcement Learning (RL). Self-supervised learning holds a lot of promise to build robust machine learning models without massive amounts of labelled data. Reinforcement learning presents a formal framework for sequential decision-making. Thanks to the recent success of deep neural networks, research into both SSL and RL has enjoyed remarkable achievements.
This talk will briefly share our journey on using SSL and RL to automate and optimize industrial operations. It will also highlight some prescriptive guidance on accelerating production-ready machine learning pipelines on AWS.
website: https://www.aicamp.ai/event/eventdeta...
Resources:

Amazon miniSCOT open source
https://github.com/amzn/supply-chain-...

AWS Energy Storage open source
https://github.com/aws-samples/sagema...

TSP Open sources
https://github.com/chaitjo/graph-conv...
https://github.com/aws-samples/amazon...

An intuitive RL tutorial (Markov Decision Process)
http://karpathy.github.io/2016/05/31/rl/

Difference between RL and Supervised learning
https://web.stanford.edu/class/cs234/...

AlphaGo paper (think fast and think slow)
https://www.nature.com/articles/natur...

Self-supervised Learning papers to read
https://arxiv.org/abs/1810.04805 (BERT)
https://arxiv.org/abs/1909.11942 (ALBERT)
https://arxiv.org/abs/2012.06659 (CoDeAR)

https://arxiv.org/abs/2002.05709 (SimCLR)