-
Notifications
You must be signed in to change notification settings - Fork 1.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add a parallel algorithm to improve the performance of TPE with large concurrency #1052
Comments
I've run experiments on some test functions with different experimental configuration ( trial concurrency, constant_liar_type etc.). Experiment on Ackley FunctionAckley Function
ExperimentSetupTwo contrast experiments: Exp 1 with Exp 2 and Exp 3 with Exp 4 to see if there's any improvement.
ResultExp 1 and Exp 2
Exp 3 and Exp 4
In this experiment on Ackley Function, TPE in parallel mode tends to explore more and find a better 'best result'. Strange PhenomenonWith parallel optimize mode on, starting from the second trial, TPE will generate several trial configurations that are totally the same. This might be a potential bug. Experiment on Rastrigin FunctionRastrigin Function
ExperimentSetup
ResultExp 1 and Exp 2Exp 3 and Exp 4Exp 5 and Exp 6 |
|
solved in #1373 |
What would you like to be added: When the number of concurrencies is large, the points selected by the TPE will be concentrated, we will add excellent parallel algorithms to solve this problem.
Why is this needed: Better usage of computing resource
Components that may involve changes: A new algorithm to handle parallel TPE and related doc
The text was updated successfully, but these errors were encountered: