Follow Gregory Piatetsky, No. 1 on LinkedIn Top Voices in Data Science & Analytics

KDnuggets Home » News » 2017 » Jan » Software » Eat Melon: A Deep Q Reinforcement Learning Demo in your browser ( 17:n03 )

Eat Melon: A Deep Q Reinforcement Learning Demo in your browser


 
  http likes 47

Check "Eat Melon demo", a fun way to gain familiarity with the Deep Q Learning algorithm, which you can do in your browser.



By Rajiv Shah, Caterpillar/UIC.

The Eat Melon demo is a fun way to gain familiarity with the Deep Q Learning algorithm. The Deep Q learning algorithm falls in the class of reinforcement learning algorithms. In reinforcement learning, an agent learns by trying to maximize a cumulative reward. In this way, the agent teaches itself.

The most well-known use of the Deep Q algorithm is training an agent to play Atari 2600 video games. The agent learned how to play a number of different games using the same model architecture. For some of the games, it was able to surpass a human expert. OpenAI has released a toolkit, so you can run reinforcement learning algorithms on Atari games.

The Eat Melon demo lets you learn about the deep Q learning algorithm in your browser. The goal is to train an agent using three different rewards. There is a reward for avoiding the walls, going straight, and for eating food. By adjusting the rewards and model parameters, you affect how quickly the agent learns. In the process, you can gain some insight into the process of using reinforcement learning.

The code for this demo is based on Karpathy's ConvNetJS Deep Q Learning Demo.

Eat Melon Deep Learning Demo

Bio: Rajiv Shah is a data scientist at Caterpillar as well as a professor at University of Illinois at Chicago. He has a PhD from the University of Illinois at Urbana-Champaign. He enjoys exploring the latest developments in data science with an interest in telematics, sports analytics, and deep learning.

Sign Up