• Home
  • From Pandas, Koalas and other Fauna to Train your AI to play a generic game

This evening DataWorkz and CodeBeez are hosting a meetup together with two interesting talks, one from CodeBeez and one from DataWorkz

From Pandas, Koalas and other Fauna to Train your AI to play a generic game

From Pandas, Koalas and other Fauna to Train your AI to play a generic game
  • LOCATION

    Dataworkz offices

  • DATUM

    27 Feb, 2020

  • TIJD

    17:00

  • KOSTEN

    free

This evening DataWorkz and CodeBeez are hosting a meetup together with two interesting talks, one from CodeBeez and one from DataWorkz

Program:

18:00 food
18:30 talk 1
19:30 short break
19:45 talk 2
20:45 drinks

Talk 1: Pythons, Pandas, Koalas and other Fauna
Speaker: Luuk van der Velden

Although data is universal, data formats are not. Whether it is in memory or on disk, there are many lifeforms wandering about. What does the Machine Learning Engineer do to make all these creatures live in harmony? How do we work with Pandas as well as Spark dataframes in models and unit tests? And how does this relate to the lifecycle of an AI project?
Abstract
Data is universal, but data formats are not. Using Python we are often the glue between various computing systems with different internal data representations. To be productive and to integrate such systems we need to understand what the strengths of each system is and how to marry them in one application. The machine learning engineer needs to transition between the various data memory and storage formats. In addition, the choices they make are influenced by the lifecycle of the data driven product. For instance, in the exploration phase we can choose processing logic that is fast to prototype but hard to scale. In the production phase we focus on scaling and maintainability. How do we make the rights choices among techniques, such as Pandas, Koalas and Spark? and what role does PyArrow play?

Talk 2: Train your AI to play a generic game
Speaker: Laurens Koppenol

In this talk we will look at a simple 2D racing game. We will see how such a game is developed (in python!) and how we can spawn virtual babies into the game. Much like real babies, these virtual babies do not know how to drive a car. Extending the parallel with real babies, our virtual spawn does not know what the ultimate goal of their existence is. We will help them with the latter, and have them solve the driving issue themselves. This will also give us insight in the tough life of an all-powerful puppetmaster, really pushing the life lessons we can learn from this.
This session is not a code-along. The code will be made freely available on github. The following topics will be discussed: game development (pygame), paint, reinforcement learning, genetic algorithms, neural nets

Event sprekers

Laurens Koppenol

Luuk van der Velden

Andere events