Commit graph

267 commits

Author SHA1 Message Date
Ronny Bjarnason
e211354b42 Add Additive Regresson as a classifier (aka Gradient Boosted Decision Trees: GBDT) 2011-02-21 12:56:22 -07:00
Ben Mabey
4f8e7c11d2 adds dataset-set-name wrapper 2011-01-18 17:18:19 -07:00
Ben Mabey
af47246368 fixes incorrect type hinting that was causing the method not to be found 2011-01-11 13:49:05 -07:00
Ben Mabey
aa4048c580 yes.. more type hinting... 2011-01-04 17:23:41 -07:00
Ben Mabey
b066a65969 more type hinting 2011-01-04 17:06:09 -07:00
Ben Mabey
1678bf3b19 yep, you guessed it- more type hinting 2011-01-04 16:58:30 -07:00
Ben Mabey
0a2c30ccbe more type hinting.. 2011-01-04 16:45:29 -07:00
Ben Mabey
8bc6223e8f more type hinting/weka lack of Interfaces dance around 2011-01-04 16:38:17 -07:00
Ben Mabey
e909942d39 more type hinting fun...
totally awesome that Weka doesn't know how to use interfaces :/
2011-01-04 16:26:31 -07:00
Ben Mabey
0dfd143a62 adds attribute-labels-as-strings since keywording isn't always desired 2011-01-04 11:47:18 -07:00
Ben Mabey
8194b4531b faster version of attribute-labels 2011-01-04 11:40:09 -07:00
Ben Mabey
44df9744c1 adds dataset-as-vecs and dataset-as-lists 2011-01-04 10:02:48 -07:00
Ben Mabey
ece83e062b disables difftest since it is not compat with lein 1.4.2 2011-01-03 16:49:42 -07:00
Ben Mabey
8cb6620b6f adds dataset-class-name 2010-12-30 10:14:57 -07:00
Ben Mabey
7cd8e08446 makes attribute-labels fn that returns set of labels and renames old one 2010-12-29 15:08:05 -07:00
Ben Mabey
ef4207ee88 extracts attribute-labels for general use 2010-12-29 14:52:16 -07:00
Ben Mabey
228c58f74c adds headers-only fn to extract the headers of a weka dataset 2010-12-22 14:23:44 -07:00
Ben Mabey
68cf83b712 fixes dataset-format to return nominal labels in correct order
Also made it more idiomatic and concise by using reduce...
2010-12-16 16:58:53 -07:00
Ben Mabey
54567b5611 fixes supervised-nominal-to-binary filter fn 2010-12-16 15:37:54 -07:00
Ben Mabey
dc6abbcce4 Merge branch 'master' of github.com:leadtune/clj-ml 2010-12-14 16:45:49 -07:00
Ben Mabey
46d437a059 updating the project repo name for maven purposes... 2010-12-14 16:45:32 -07:00
Ronny Bjarnason
9829c0eea4 fix typo decission->decision in tests 2010-12-14 15:16:39 -07:00
Ronny Bjarnason
c89e559ace additional options to weka models 2010-12-14 15:04:42 -07:00
Ronny Bjarnason
44f63508c9 1. Create weka model wrappers for PaceRegression, RandomForests, M5P Trees
and boosted stumps (LogitBoost).
2. Added functionality in MultilayerPerceptron to specify hidden layers
3. corrected misspelling decission->decision
2010-12-13 17:01:25 -07:00
Ben Mabey
89578f113c comment 2010-12-08 16:14:28 -07:00
Ben Mabey
4d18af461e not keywording the instance nominal values by default due to performance
On large datasets if you use dataset-as-maps too much time is taken up
in interning the nominal values time and time again.  For the vast
majority of use cases string values should be fine.  If we were in
clojure-land 100% then keywords would be the better option.. but we
aren't.
2010-12-08 16:10:45 -07:00
Ben Mabey
5d59785f3b type hints 2010-12-08 16:05:46 -07:00
Ben Mabey
81cda11ed2 DRYs up some fns and replaces custom code with enumeration-seq 2010-12-08 16:05:41 -07:00
Ben Mabey
b47e794176 removes custom (and slow) nominal lookup code- just use weka's API 2010-12-08 14:29:25 -07:00
Ben Mabey
a34fe4913f adds linear and logistic regression wrapeprs 2010-12-07 17:24:22 -07:00
Ben Mabey
652c54aaf7 standardizes how attributes can be asked for (by index or name) 2010-12-07 14:14:20 -07:00
Ben Mabey
99048715f1 adds tests for dataset label helper fns 2010-12-07 13:48:52 -07:00
Ben Mabey
6a30d86841 adds dataset-nominal? 2010-12-07 09:32:27 -07:00
Ben Mabey
138ee7f016 adds dataset-class-values 2010-12-06 17:36:36 -07:00
Ben Mabey
0ceb8020fa changes values -> labels where appropriate to be consitent with weka terminology 2010-12-06 17:02:11 -07:00
Ben Mabey
49f99e3218 adds classifier-copy-and-train as a non-side-effect alternative 2010-12-06 16:47:50 -07:00
Ben Mabey
df8fcd467c type hint 2010-12-03 14:31:57 -07:00
Ben Mabey
b6deb0421e improves filters API by making filter fns that create and apply filters 2010-12-02 11:44:38 -07:00
Ben Mabey
3a393ebcf9 adds make-apply-filters 2010-12-01 16:22:19 -07:00
Ben Mabey
ca7281d214 allows attributes in filters to be specified by name in addition to index 2010-12-01 14:36:05 -07:00
Ben Mabey
69a682f111 exluding deps that went missing and caused things to break 2010-12-01 11:43:30 -07:00
Ben Mabey
78fc93c2a4 allow for dataset-set-class to take the name of attribute 2010-12-01 11:37:27 -07:00
Ben Mabey
93846bcaea type hint classifiers as much as possible... 2010-11-29 10:08:57 -07:00
Ben Mabey
31348dd1e0 reworks dataset-as-maps to be faster by avoiding redundant string interning 2010-11-24 12:12:41 -07:00
Ben Mabey
1f3900a628 replaces some loop/recurs with more conciser idiomatic clojure 2010-11-24 12:05:30 -07:00
Ben Mabey
88d65f0e24 removes custom key-to-str fn in favor for clojure.core/name 2010-11-23 16:09:13 -07:00
Ben Mabey
ccc12c2d5c adds dataset-replace-attribute! 2010-11-08 10:37:42 -07:00
Ben Mabey
285e45717a shorter, and more idomatic version of index-attr 2010-11-08 10:37:42 -07:00
Ben Mabey
2d274d2598 adds into-fast-vector util fn 2010-11-08 10:37:42 -07:00
Ben Mabey
19a827d8b3 sets the output-format for ClojureBatchFilter to use the process fn format 2010-11-08 10:37:42 -07:00