Commit graph

30 commits

Author SHA1 Message Date
06325a80bf Fix remaining bugs in unified training mode 2019-08-06 13:18:40 +01:00
6d21146b05 [server/DatasetFetcher] Don't include distance in unified training mode 2019-08-06 12:46:07 +01:00
858f53c588 Include the original training mode in the index file 2019-08-06 12:45:37 +01:00
082eba42ec Scan the input data for invalid values 2019-08-06 12:10:47 +01:00
5e78f5c788 Tweak logging messages 2019-08-05 17:08:38 +01:00
3ebc54ef74 Add unified training mode, but apparently we're now crashing the client 🤔 2019-08-05 17:06:31 +01:00
45541db0fa Tidy up new datasetfetcher logging 2019-08-02 14:04:34 +01:00
a2e4551b75 Mess about with the zapping system & add another secret geojson generator 2019-08-02 14:00:31 +01:00
6844cdc545 Implement initial false negative zapping system
...but it's not having the effect I'd hoped.
2019-08-02 13:17:17 +01:00
f18023cf1a Fix more training and prediction bugs, but its still not working as intended. 2019-07-30 18:56:45 +01:00
dea7940b2e Bugfix: Fix client-side Brain.js crashes 2019-07-30 18:38:44 +01:00
623af65aef Remove unused imports 2019-07-30 16:40:12 +01:00
c9114cc4c4 Use new tidied normalising code on the client-side 2019-07-30 15:55:15 +01:00
4493fbb0fa Tidy up normalising code 2019-07-30 15:42:37 +01:00
25bb302033 Bugfix: Fix Brain.js training issues 2019-07-30 15:28:31 +01:00
7d709c0fa5 Bugfix: Fix crashes in DatasetFetcher 2019-07-30 15:15:34 +01:00
03c1cbb97f Finish refactoring to use Brain.js, but it's untested. 2019-07-29 18:06:50 +01:00
cce0761fed Continue brain.js refactoring, but it's still not finished yet. 2019-07-29 17:02:34 +01:00
2a81ac792e Start replacing Tensorflow with Brain.js, but it's not finished yet
Note to self: bug them to upgrade ot GPU.js 2 for future projects
2019-07-29 14:36:52 +01:00
43f99d6b5d Try to add an additional input to the neural network, but it's not working. 2019-07-29 14:17:28 +01:00
61edaf6fa3 Write AI layer, but it's untested. 2019-07-23 15:45:29 +01:00
f0694de02c Bugfix: Create 1 model per-gateway 2019-07-23 15:14:50 +01:00
d158868c3f AITrainer: Save gateway locations to the index file 2019-07-23 13:44:00 +01:00
c2810d4871 Improve format of index 2019-07-22 12:42:04 +01:00
d9cf650019 Write index JSON file to AI output directory 2019-07-22 12:40:19 +01:00
92574bc98c Serialise trained AIs and save them to disk 2019-07-22 11:53:30 +01:00
5c7cc1e906 Fix bugs in AI trainer. It works, but precisely it's doing si up for debate.... 2019-07-18 17:22:37 +01:00
192e31a925 Write the the data input, but it's untested.
There's also some confusion over the batch size. If it doesn't work, we 
probably need to  revise our understanding of the batch size in 
different contexts.
2019-07-18 16:34:25 +01:00
6dbb6c3b87 Hook up a dataset importer for training the AIs, but it's untested.
Also, we don't have any code that actually does the training itself 
either yet.
2019-07-17 15:15:31 +01:00
f7e2d77daa Start writing the AI training code, but it's not finished yet 2019-07-17 14:16:24 +01:00