Task #10366 (closed)
Opened 11 years ago
Closed 11 years ago
Consider saving WND-CHRM classifier instead of generating on the fly
Reported by: | spli | Owned by: | spli |
---|---|---|---|
Priority: | minor | Milestone: | OMERO-4.4.7 |
Component: | General | Version: | n.a. |
Keywords: | n.a. | Cc: | analysis@… |
Resources: | n.a. | Referenced By: | n.a. |
References: | n.a. | Remaining Time: | n.a. |
Sprint: | 2013-03-12 (6)) |
Description
The WND-CHRM classifier has slightly complex storage requirements as it uses nearest neighbour, so all training samples must be stored. At present the classifier is re-trained from the saved features everytime a set of predictions are made, maybe it would be better to separate the training and prediction stages, and either store the classifier state in OMERO.tables or (easier) jsut dump it to a Python pickle file saved as an OriginalFile?.
Change History (3)
comment:1 Changed 11 years ago by jburel
- Sprint changed from 2013-02-12 (5) to 2013-03-12 (6))
comment:2 Changed 11 years ago by spli
See parent story #10230
comment:3 Changed 11 years ago by spli
- Resolution set to fixed
- Status changed from new to closed
Referencing ticket #10230 has changed sprint.