AWS DeepRacer Expert Boot Camp at AWS re:Invent 2019 – Programme

Read the Boot Camp announcement here.

When

Monday December 2nd, 4:00PM – 7:00PM PST,

Tuesday December 3rd, 12:00PM – 6:00PM PST,

Wednesday December 4th, 12:00 PM – 6:00PM PST

Where

Aria, Level 1, Pinyon 2

Sessions

All times are in Pacific Standard Time.

MON 12/2 4:00PM – 4:30PM VIP Welcome and Swag Pick Up (location TBA)

MON 12/2 4:00PM – 7:00PM Elizabeth Varland: Under the hood: how I built a gas-to-electric conversion go-kart with AWS DeepRacer

---

TUE 12/3 12:00PM – 12:45PM Free Racing / Ask the Experts Anything

TUE 12/3 12:45PM – 1:00PM Lyndon Leggate & Tomasz Ptak: Kickoff

TUE 12/3 1:00PM – 1:30PM Adrian Cockroft: Innovating at speed: AWS and Formula 1

TUE 12/3 1:30PM – 2:00PM Kire Galev: Customizing AWS DeepRacer action space. Learn how to design better performing action space for one or more tracks using my track analysis and action space analysis tools.

TUE 12/3 2:00PM – 2:30PM Cyrus Wong and Kitson Cheung: AWS DeepRacer tips and tricks: How to build a powerful rewards function with AWS Lambda and Adobe Photoshop

TUE 12/3 2:30PM – 3:00PM Tomasz Ptak: Through the logs and into the AWS DeepRacer’s mind with the AWS DeepRacer log analysis tool. In this talk, we will discuss how AWS DeepRacer logs can be used to look at the car’s performance from a new perspective to plan and improve our training.

TUE 12/3 3:00PM – 3:30PM Sunil Mallya: Interpretability of AWS DeepRacer models. When building neural networks its desirable to understand as to why the model make the predictions they do. In this session we’ll explore techniques to visualize where the DeepRacer RL model chooses to focus while taking its navigation decisions.

TUE 12/3 3:30PM – 4:00PM Guy Ernest: AWS DeepRacer hyperparameter tuning with Amazon SageMaker. In this session you will learn how to both optimize your model and build, train, and deploy production grade models for scale

TUE 12/3 4:00PM – 4:30PM Alex Schultz: Hyperparmeter mechanics and optimization. AWS DeepRacer hyperparameters provide a complicated set of levers that can either enable your model to move quickly or not at all. In this session you’ll learn from AWS DeepRacer experts how each hyperparameter works and the range of effects you can employ to maximize your model’s performance.

TUE 12/3 4:30PM – 5:00PM Sahika Genc, Prem Ranga, Anthony Navarro: Using reinforcement learning and simulation for broad experimentation and advanced concepts

TUE 12/3 5:00PM – 6:00PM Sunil Mallya: Ask the Experts Anything

---

WED 12/4 12:00PM – 12:30PM Lyndon Leggate: Kickoff

WED 12/4 12:30PM – 1:00PM Elizabeth Varland: Under the hood: how I built a gas-to-electric conversion go-kart with AWS DeepRacer

WED 12/4 1:00PM – 1:30PM Jouni Luoma: Minimum math introduction to machine learning side of DeepRacer. In this session you will learn how the DeepRacer actually learns to drive in simulator and what happens when you the model to a real AWS DeepRacer device.

WED 12/4 1:30PM – 2:00PM Ray Goh: From simulator to reality: how to design and test your model for real world top performance. Your model proved itself in the simulator, but will it perform on a real AWS DeepRacer device and physical track? Learn what works (and what doesn’t) when moving from the simulator to the real world and how to prepare for real-world racing.

WED 12/4 2:00PM – 2:30PM Chris Thompson: Spelunking AWS DeepRacer. An under the hood deep dive into the AWS DeepRacer stack including, OpenAI Gym, Intel Coach, ROS, Gazebo, Docker, RaceLines, papers, K1999 algorithm, and the Jupyter race line calculation.

WED 12/4 2:30PM – 3:00PM Arthur Pace: Find the best trajectory for your model. Default waypoints can bias the way the model learns. You’ll learn how to find the best waypoints for faster training and a perfect trajectory.

WED 12/4 3:00PM – 3:30PM Karl Stiegler: DeepRacer Hacks – How I won 3 races in the 2019 virtual circuit. In this in-depth, I will share how I approached the DeepRacer Virtual Circuit races, including action space selection, what behaviours to reward, training analysis and the training hacks and optimizations I built to dominate the online league.

WED 12/4 3:30PM – 4:00PM Scott Pletcher: Roll your own AWS DeepRacer event. In this session, you will learn how to host your own AWS DeepRacer event with everything from building a track, logistics, getting support from AWS, planning, leaderboards and more.

WED 12/4 4:00PM – 4:30PM Martin Paradesi and Sujoy Dutta: How Capital One trains internal associates on reinforcement learning using AWS DeepRacer. AWS DeepRacer is a great educational tool for learning about Machine Learning and Reinforcement Learning. This session explores the lessons that Capital One learned while teaching reinforcement learning concepts using AWS DeepRacer internally.

WED 12/4 4:30PM – 5:00PM Tim O’Brien: How much training is enough? In this lightning talk, you can view the results of training models with different training periods and perhaps optimize your model without spending a lot or overfitting.

WED 12/5 5:00PM – 5:30PM Mohammad Al-Ansari and Weiling Gu: How to win as a team and the secret source of a champion model.

WED 12/4 5:30PM – 6:00PM Lyndon Leggate, Tomasz Ptak and Friends: Ask the Experts Anything

See you!

Update 18.11.2019: Adjusted Monday times, updated Tuesday Kickoff