Gneural Network - Tasks: task #14200, Implement random_search training...
You are not allowed to post comments on this tracker with your current authentication level.
task #14200: Implement random_search training for nnet.
Submitter: | Ray Dillinger <rayd> | ||
Submitted: | Sun 30 Oct 2016 04:37:32 PM UTC | ||
Should Start On: | Sun 30 Oct 2016 07:00:00 AM UTC | Should be Finished on: | Wed 30 Nov 2016 08:00:00 AM UTC |
Category: | None | Priority: | 7 - High |
Status: | None | Privacy: | Public |
Assigned to: | None | Percent Complete: | 0% |
Open/Closed: | Open | Effort: | 0.00 |
No files currently attached
Depends on the following items: None found
Items that depend on this one: None found
Carbon-Copy List
There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.
Follows 1 latest change.
Date | Changed by | Updated Field | Previous Value | => | Replaced by |
---|---|---|---|---|---|
2016-10-30 | rayd | Should be Finished on | 2016-10-30 | 2016-11-30 |
Powered by Savane 3.13-02a9.
Corresponding source code
The current implementation (for nn) simply makes a series of random guesses as to the weights of the whole network, checking each to see if it has lower error than the best yet discovered - a pure, weight-only, exploration strategy.
It's possible to do much better by selecting randomly among a set of different operations that include a mix of many possible exploration and exploitation strategies. If feeling particularly ambitious, it could also include topology elaboration and reduction strategies.