wassname/attentive-neural-processes
1
0
Fork 0
mirror of https://github.com/wassname/attentive-neural-processes.git synced 2026-06-27 18:03:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

improved LSTM results

Browse Source
This commit is contained in:
wassname
2020-02-16 21:42:37 +08:00
parent 2ec7b99bb7
commit 3e277c22a2
2 changed files with 1121 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files
readme.md
+4 -3
View File
@@ -54,11 +54,12 @@ Results on [*Smartmeter* prediction](./smartmeters-ANP-RNN.ipynb) (lower is bett
|Model|val_np_loss|val_mse_loss|
|--|--|--|
| **ANP-RNN(impr)(MCDropout)**| **-1.48**|
|**ANP-RNN_imp**|**-1.38**|.00423
|ANP-RNN|-1.27|0.0047|
|ANP|-1.3|0.0072|
|NP|-1.3|0.0040|
|LSTM|-0.066| 0.010 |
|LSTM|-0.78| 0.0074 |
### Example LSTM baseline
@@ -132,8 +133,8 @@ The model tries to estimate the how unsure it is, but what about when it is out
|Name|val_loss (n=100) [lower is better]|
|--|--|
|MCDropout| -1.40|
|Normal| -0.64|
|MCDropout| -1.31|
|Normal| -1.04|
We can estimate additional uncertainty by using Monte Carlo Dropout to see how uncertain the model acts in the presence of dropout. This doesn't capture all uncertainty, but I found that is does improve (decrease) the validation loss. The loss is calculated by the negative overlap of the output distribution and the target value so this improvement in the loss shows that MCDropout improved the estimation of the uncertainty.
smartmeters-lstm_std.ipynb
+1117 -99
View File
File diff suppressed because one or more lines are too long