 emeric
|
e0711861b5
|
Simplified the feature settings
|
2019-12-15 14:29:44 +01:00 |
|
|
emeric
|
74a5f6389d
|
Added a cache for track features, made the crossover ratio configurable, now using a fitness proportional selection
|
2019-12-03 13:47:48 +01:00 |
|
|
emeric
|
93cd3b6020
|
Made the number of samples per neuron configurable
|
2019-11-29 13:37:41 +01:00 |
|
|
emeric
|
85129d4a40
|
WIP, first working genetic algorithm to train the neural network
|
2019-11-29 13:24:03 +01:00 |
|
|
emeric
|
07eae9cf96
|
WIP, reworking features
|
2019-11-08 13:34:07 +01:00 |
|
|
emeric
|
38ce2600d2
|
Rework DB session + transactions. Now can handle multiple read only transactions in parallel
|
2019-07-16 13:17:23 +02:00 |
|
|
emeric
|
08988b17a2
|
Prevent the media scanner from never quitting
|
2019-06-13 14:29:39 +02:00 |
|
|
emeric
|
0aef48f3bc
|
Fallback on the cluster similarity searcher if no song is classified
|
2019-04-19 13:01:57 +02:00 |
|
|
emeric
|
368766974e
|
Similarity: fallback on tags if the artist or release has no track containing features
|
2019-03-21 13:06:25 +01:00 |
|
|
emeric
|
45b4468c01
|
Refactored the inputvector part of the som implementation
|
2019-03-08 15:46:29 +01:00 |
|
|
emeric
|
11a0ed9f12
|
Breakable training, added cache
|
2019-02-27 13:57:43 +01:00 |
|
|
emeric
|
6fcb693261
|
Reorganized sources, better accuracy for similarities based on features
|
2019-02-05 14:08:46 +01:00 |
|