AI-Driven Application Optimization
summary
Covers how multi-armed bandit algorithms trade exploration and exploitation to pick winning thumbnails, headlines or routes, and how exploration rate increases when the environment shifts; shows results on advertising where one banner yields 10 clicks per 100 and explains why always serving that banner fails after a trend change. Demonstrates a GitHub simulator that compares popular algorithms, applies to thousands or millions of options and runs continuously for live adaptation.
AI-written summary · about
chapters (10)
- 0:00Exploration vs exploitation
- 1:05Bandit algorithms in practice
- 3:33Ads with Bandit algorithms
- 5:20Biological learning exploration
- 6:05Environmental exploitation changes
- 7:31Balancing exploration exploitation
- 9:17Scaling Bandit algorithms
- 10:41Bandit strategies resources
- 11:45Key takeaways AI systems
- 13:33Q&A Bandit vs A/B testing
related talks
21mLeveraging Multi-Armed Bandit Algorithms for Dynamic Decision Making
20mWinning the Game: Leveraging Contextual Bandits for Creative Optimization
32mMulti-Armed Bandits in Recommendations
21mRise of the Machines: Removing the Human-in-the-Loop
57mSearch Relevance: A/B testing to Reinforcement Learning
38mA Guide to Modern Hyperparameters Turning Algorithms
27mBeing Smart with Art
30mPractical Multi Armed Bandits