Shift your ML model into overdrive with AWS DeepRacer analysis tools
summary
AWS DeepRacer adds Open and Pro divisions (Pro requires top 10% in a monthly virtual race) and shows concrete ways to improve lap times: use progress over steps to shave seconds by rewarding progress per step, inspect logs (~15 steps/sec, examples at 600–6,000 iterations) and download training logs into a Jupyter/SageMaker Notebook to visualize waypoints (153 on one track), action histograms and over 900 camera frames. Uses numpy/scipy/shapely to compute distances; adjust learning rate or action space when histograms show wasted actions.
AI-written summary · about
chapters (12)
- 0:00Start
- 1:01What's New for aws Deep Racer in 2021
- 1:32The Aws Deep Racer League
- 2:38Reward Function
- 5:06Combine Steps with Progress
- 5:47Evaluating Your Reward Function
- 6:58Waypoints
- 15:03Overfitting
- 17:34Hyper Parameters
- 19:45Training Log Data
- 22:58Analyze the Reward Distribution
- 23:48Histogram for each Action Distribution
related talks
30mShift your ML model into overdrive with AWS DeepRacer analysis tools
28mShift your ML model into overdrive with AWS DeepRacer analysis tools
30mPush Your ML Model to Top Speed (AWS DeepRacer)
29mGet rolling with ML on AWS DeepRacer
30mStart your ML journey with AWS DeepRacer for public sector
37mAWS DeepRacer: Train, Evaluate, and Tune
23mGet rolling with machine learning on AWS DeepRacer
24mGet Rolling on Machine Learning