/autoscaling

Fuzzy Deep Reinforcement Learning for autoscaling on clouds

Primary LanguagePythonThe UnlicenseUnlicense

Fuzzy Deep Reinforcement Learning

Inspired by the recent variants in Deep RL, Fuzzy Deep RL is developed by using Fuzzy Logic as a data representation method and Deep Q-Networks for autoscaling problem

References:

The used dataset from Clarknet Traces ftp://ita.ee.lbl.gov/traces/

Simulation code refined from Feedback Control for Computer Systems by Philipp K. Janert (O'Reilly Media)

Fuzzy Q-Learning corrected from FQL

Deep QN architecture derived from Hands-on Deep Reinforcement Learning, published by Packt

For the use of fuzzy logic with DQN, please see Fuzzy DQN and Fuzzy Q Learning FQL

If you use this code please cite as:

Doan D.N., Zaharie D., Petcu D. (2020) Auto-scaling for a Streaming Architecture with Fuzzy Deep Reinforcement Learning. In: Schwardmann U. et al. (eds) Euro-Par 2019: Parallel Processing Workshops. Euro-Par 2019. Lecture Notes in Computer Science, vol 11997. Springer, Cham. https://doi.org/10.1007/978-3-030-48340-1_37