Shift your ML model into overdrive with AWS DeepRacer analysis tools
summary
Announces two 2021 race divisions (open and pro) and shows concrete tuning techniques: reward-by-progress over steps (about 15 steps/sec) produces inside-line racing that cuts lap time, using waypoint ranges (example waypoints 5–9 and 10–20) to change behavior mid‑corner, and leveraging dumpee Sai pi and shapely modules to compute distance/angle instead of hand trig. Provides log analysis notebook with ~900 sample images, per-step logs (episode/step entries), action‑space diagnostics (e.g. nine actions) and heatmaps to spot overfitting or unused actions.
AI-written summary · about
chapters (13)
- 0:00Introduction to AWS DeepRacer
- 1:202021 Update and Divisions
- 2:20Reward function fundamentals
- 4:08Advanced reward techniques
- 7:00Using waypoints and math
- 11:59Interpreting training graphs
- 14:20Training vs evaluation lines
- 16:58Avoiding overfitting
- 19:08Hyperparameter adjustments
- 20:24Deep dive into log analysis
- 22:29Analyzing log data
- 24:12Action space analysis
- 27:18Visualizing model decisions
related talks
28mShift 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
27mGet rolling with machine learning on AWS DeepRacer
30mStart your ML journey with AWS DeepRacer for public sector
37mAWS DeepRacer: Train, Evaluate, and Tune
24mGet Rolling on Machine Learning