Testing Neural Network - Python AI in StarCraft II tutorial p.12

TL;DR

In this tutorial, the author demonstrates how to use a trained neural network model to make more strategic choices in Starcraft 2 gameplay.

Transcript

what is going everybody welcome to part 12 of the artificial intelligences in the Starcraft 2 with Python tutorial series in the previous tutorials we built and trained a neural network and now we want to put it to use so in order to do that it's not too difficult all we're gonna do is I'm headed to part 9 and scrolling on down to the very bottom w... Read More

Key Insights

• 😒 The tutorial demonstrates how to use a trained neural network model in Starcraft 2 gameplay.
• 🚂 The trained model outperforms a random model in defeating the medium AI opponent.
• 💄 Several potential improvements for the AI are discussed, including better scouting techniques and expanding the decision-making process.

Questions & Answers

Q: How does the author suggest using the trained neural network model in Starcraft 2?

The author suggests loading the model using Keras and then using the predict() function to make predictions based on the game data. The output of the prediction can be used to make decisions in the gameplay.

Q: What improvements are discussed for the AI in the tutorial?

The tutorial mentions several potential improvements for the AI, including using a worker unit as the initial scout instead of an observer, better scouting techniques, and expanding the decision-making process to include more choices and strategies.

Q: How does the trained model perform compared to a random model?

According to the author's testing, the trained model performed significantly better than a random model. The trained model defeated the medium AI in Starcraft 2 66% of the time, compared to 44% for the random model.

Q: Are there any other resources or communities mentioned in the tutorial?

Yes, the tutorial mentions a Discord channel dedicated to Starcraft 2 AI, where users can find helpful discussions and suggestions for improving their AI projects.

Summary & Key Takeaways

• The tutorial is about implementing a neural network model in Starcraft 2 gameplay using Python.

• The author explains how to load the model and use it to make decisions based on predictions.

• The tutorial also discusses potential improvements for the AI and future directions for the project.