4b081b92f1
Merge pull request #45 from LanderTome/feature-LeakyReLU
...
Add Leaky ReLU activation function.
2019-02-14 01:22:51 +03:00
b8ad378a79
Add Leaky ReLU activation.
...
Differentiation with autograd package confirmed to work correctly.
2019-02-13 13:51:13 -08:00
602d3b6fd9
Fix reduction test
2019-01-25 12:27:34 +03:00
dbf7159841
Merge pull request #44 from Timehsw/master
...
update gitignore
2019-01-25 11:48:28 +03:00
3352ca819b
update gitignore
2019-01-25 15:02:14 +08:00
be53f3eaff
Add AUTHORS
2019-01-07 14:08:47 +03:00
3d41c9f9bb
Merge pull request #43 from keineahnung2345/kmeans++
...
kmeans++
2019-01-07 14:00:15 +03:00
cd6689c12a
kmeans++
...
It seems the codes do not match with the comment starting from line 33.
Revise it according to the comments.
2019-01-07 17:02:36 +08:00
e253211090
Get the derivative of the loss function out of the for loop ( #42 )
...
I think getting the derivative of the loss function should be out of the for loop. There is no need to calculate the derivative function for max iteration times.
2018-10-23 08:18:21 +03:00
5fbfeca8c4
Update test_regression_accuracy.py
2018-10-21 23:46:12 +03:00
64d7c51d00
Fix the formula of l2 penalty ( #40 )
2018-10-21 23:45:45 +03:00
6b729f043e
update the penalty function in linear models ( #36 )
2018-06-30 16:48:41 +03:00
d398777642
Fix softplus overflow (attempt 2) ( #31 )
...
* actually fix numeric overflow
2017-08-17 11:56:34 +03:00
4b3c24dfce
Fix for Numeric Overflow of softplus implementation ( #30 )
...
* fix possible numeric overflow for softplus
* english
2017-08-12 07:24:05 +03:00
7b1cbfd833
Fix typo
2017-08-09 11:50:58 +03:00
2b798428d8
Update rbm.py
2017-03-27 11:48:39 +07:00
b7364b560d
Fix the CD-1 training code and add comments in RBM ( #29 )
...
1. `negative_visible` should also be sampled, only the last hidden can use the expectation. Though trivial modification, it follows the original algorithm, and theorically it should also be done (refer to Kevin Murphy's MLaPP P944).
2. Some comments are added in the code, for more friendly reading. At the first sight I didn't find it's CD :)
2017-03-27 08:13:32 +07:00
1fbc443166
Fix markdown
2017-03-23 09:07:32 +07:00
96397f5ff8
Fix typo ( #28 )
2017-03-23 08:09:30 +07:00
aa41a68676
fixes missing log ( #26 )
2017-02-15 17:54:25 +03:00
12e7c8770e
Update some comments
2017-01-23 06:09:34 +03:00
11ef425874
Remove rl directory
2017-01-08 16:35:43 +03:00
7f0836b22a
Fix mlp test
2017-01-08 05:51:51 +03:00
766ee72969
Add some comments to DQN module
2017-01-08 04:50:45 +03:00
63b00ff95f
Add initial version of Deep Q learning (DQN)
2017-01-08 04:37:44 +03:00
6b46afa7cc
Update neural network interface, FIX bug in optimization
2017-01-08 04:35:12 +03:00
a64a2e4580
Remove TODO list, update contributing section
2017-01-01 14:03:40 +03:00
0e89fb762d
Add t-SNE example
2016-12-28 21:44:24 +03:00
f24fe1d4d8
Add initial t-SNE implementation
2016-12-28 19:04:40 +03:00
cd261f9afc
Update Readme, fix test
2016-12-26 22:30:39 +03:00
6528cf5d89
Add initial version of rbm
2016-12-26 22:28:48 +03:00
b060229349
Fix entropy function, #22
2016-12-21 00:02:41 +03:00
2a41d55fc4
Update tree.py to default to use all features ( #21 )
2016-12-06 09:57:58 +03:00
9dc76dfaff
Add method. ( #20 )
2016-12-04 14:22:06 +03:00
829c74cf7d
Fixing on KMeans plot and Gaussian Mixture ( #19 )
...
* Add Gaussian Mixture Model and example
* Fixing KMeans.plot and a trivial change on GMM code
2016-11-29 16:36:24 +03:00
b6d1489cef
Add Gaussian Mixture Model ( #18 )
2016-11-28 16:37:29 +03:00
8f111282c9
Add Gaussian Mixture Model and example ( #18 )
2016-11-28 16:36:30 +03:00
bdbd2bfbf4
fit pca w training data, not entire dataset ( #17 )
...
* fit pca w training data, not entire dataset
* remove eggs
2016-11-26 16:27:09 +03:00
21b090dafe
Fix neuralnet test
2016-11-26 03:23:34 +03:00
862bcdfae9
Unittest for reduction ( #16 )
...
* add test_DR
* include pca fitting
* include pca fitting
2016-11-26 03:11:31 +03:00
41abfd310f
add docker configuration ( #13 )
...
* add docker configuration
Why:
* it's a nice way to work in a disposable environment
This change addresses the need by:
* add dockerfile
* update readme to explain usage
Reference Links:
*
* switch to using official python image as a base
Why:
* removes a lot of custom setup
This change addresses the need by:
* swap out buildpack-deps:trusty with python:3
* remove setup lines that are no longer required
* clean up comment to describe the RUN command
Reference Links:
* https://hub.docker.com/r/library/python/
2016-11-24 02:14:58 +03:00
830f68a905
Add some comments to the neuralnets
2016-11-20 21:53:26 +03:00
2c15265007
Small fixes for k-means, Naive bayes
2016-11-19 22:21:38 +03:00
33d2662530
Ensemble methods: Optimize split search, update comments
2016-11-19 22:03:10 +03:00
2e4398de5e
Remove unused dataset, update requirements
2016-11-18 18:25:27 +03:00
9fbb7114bc
Merge branch 'master' of https://github.com/rushter/MLAlgorithms
2016-11-18 18:09:29 +03:00
2027d3347a
Update Naive Bayes classifier
2016-11-18 18:08:53 +03:00
507a5ef9b0
Implemented k-NN regressor ( #11 )
...
* Implemented a KNN Regressor.
* changed a comment and used generator instead of list
2016-11-18 17:29:16 +03:00
acfef46f8e
Remove redundant suffixes from tests
2016-11-18 02:19:26 +03:00
6d75555ab3
Update knn example, add test
2016-11-18 02:16:35 +03:00
Artem Golubin