No abstract

No abstract

Organizations across the globe gather more and more data, encouraged by easyto-use and cheap cloud storage services. Large datasets require new approaches to analysis and processing, which include methods based on machine learning. In particular, symbolic regression can provide many useful insights. Unfortunately, due to high resource requirements, use of this method for large-scale dataset analysis might be unfeasible. In this paper, we analyze a bottleneck in the open-source implementation of this method we call hubert. We identify that the evaluation of individuals is the most costly operation. As a solution to this problem, we propose a new evaluation service based on the Apache Spark framework, which attempts to speed up computations by executing them in a distributed manner on a cluster of machines. We analyze the performance of the service by comparing the evaluation execution time of a number of samples with the use of both implementations. Finally, we draw conclusions and outline plans for further research.

No abstract

Reinforcement learning has been recently a very active field of research. Thanks to combining it with Deep Learning, many newly designed algorithms improve the state of the art. In this paper we present the results of our attempt to use the recent advancements in Reinforcement Learning to automate the management of heterogeneous resources in an environment which hosts a compute-intensive evolutionary process. We describe the architecture of our system and present evaluation results. The experiments include autonomous management of a sample workload and a comparison of its performance to the traditional automatic management approach. We also provide the details of training of the management policy using the Proximal Policy Optimization algorithm. Finally, we discuss the feasibility to extend the presented approach to other scenarios.