According to the 8th Kapitel (Play) book I need to prepare the models and programs to be shrunk models. This can lead to make the train. One of the tasks that needs to be done is to understand and apply the sample codes below.
8. Kapitel also tells it in order, I tried to set it up but I got stuck after a while, I tried the sample codes in Python (in Linux-Ubuntu system), but I could not get out of it.
3 steps needs to be taken:
1. enclosed is a new topic: it is about WorldModels, which is a reinforcement learning system that uses a network similar to the U-net (a variational autoencoder) to find a world model. An optimal action is then searched in this model. (If you need the description please inform me to send)
The advantage is that the complete system is already implemented. A critical aspect of these systems is that they often only work well under very limited conditions. It would be a matter of finding out whether the system, which copes well with classic game environments, it is also suitable for concrete applications in robotics. If the subject turns out to be too big or too difficult, it would have the advantage that we can switch to a part, such as providing training samples for such a system for a robotic application.
2. For any reference that you need I can send it for your further knowledge.
Task of the work is:
- Understanding the theory and the ongoing papers
- Comprehend and implementing a sample code (in practice this is always more than just copying the code)
- Applying it to something new (we should talk about it when you understand everything)
Explain the code in details, the new one and the codes are still in use in faster operating system like Ubuntu. (Ask for further information)
3. scale down the models a bit.
Don’t hesitate to contact for any information you need and the existing codes
Hello,
I understand you need a freelancer to help you in Machine Learning projects
My Skills:
Machine Learning, Data Science, Python, Artificial Intelligence
Hey there! I'm a PhD candidate, studying Deep Reinforcement Learning algorithms and applications. WorldModels is a complicated algorithm, I've studied it, but I believe I can get it to work as you wish. Let me know if you're interested.