44d81a0ec3
FIX: add except* in Splitter.init definitions
2015-02-23 08:19:10 +01:00
c03c01a504
ENH Bring sparse input support to tree-based methods
...
Author: Arnaud Joly <arnaud.v.joly@gmail.com>
Fares Hedayati <fares.hedayati@gmail.com>
2014-11-06 13:58:28 +01:00
7e9c093585
tree: add min_weight_fraction_leaf
2014-07-02 11:47:31 +02:00
cdd2279c09
ENH fix astype usage to prevent copying
...
Removed where possible; using copy=False (from utils.fixes) where needed.
Also some C integer type fixes to gradient boosting.
2014-06-18 11:01:28 +02:00
7ce38f4405
ENH: better API for passing weighted_n_samples
2014-04-12 13:36:16 +02:00
57a743a157
ENH: rename Node attributes to avoid confusion
2014-04-12 11:06:57 +02:00
61dd4b4be9
FIX: weight impurity improvement by the size of the node
2014-04-12 10:33:04 +02:00
84c3742cb3
ENH re-generate c file
2014-03-20 14:20:35 +01:00
e509724e7e
ENH avoid trying to split on constant features for presort best splitter
2014-03-10 12:16:42 +01:00
25e2c19573
FIX bug with BestSPlitter + ENH avoid searching for a split on constant features
2014-03-10 12:16:32 +01:00
57f93fac16
ENH avoid splitting on constant features for best splitter
2014-03-10 12:16:22 +01:00
5cc89b2d3a
FIX: Recompile _gradient_boosting.c
2014-02-06 19:35:13 +01:00
7c33502c2e
ENH faster heapsort in trees
...
Uses the heapsort version that I'm familiar with: linear-time heapify
followed by n delete-min operations. Also cache-friendlier by copying
active features into a temporary array.
Random forest training time on covertype:
now 34.4123s
before 60.1179s
2014-01-25 15:13:01 +01:00
87194f77cb
Remove 'locked', add comments
2014-01-25 22:10:18 +11:00
86fbebdadf
FIX dtype refcount; block resizing after building
2014-01-24 09:30:06 +11:00
e7e7133912
ENH/FIX Change Tree underlying data structure
...
* avoid segfault (#2726 ) by setting base on numpy arrays
* A struct array for tree nodes and values array (size indeterminate at
compile time) are the only underlying structure, so each node is locally
grouped memory, and joblib dumps will save only two numpy files per
tree.
* predict() uses the value array's take method, reducing code repetition
2014-01-16 09:32:16 +11:00
b3a9da908d
Regenerate C files
2014-01-09 10:05:20 +01:00
dd680d6a2f
malloc / realloc checks
...
correct rc values
2014-01-06 13:56:23 +01:00
fda6211367
pass impurity to ``impurity_improvement`` - now it is correct so that arbitrary Criteria can be used in best-first search.
2013-11-29 10:56:57 +01:00
cde390f2c3
remove split_impurity from node_split signature
2013-11-22 17:47:21 +01:00
9d485d78c2
fix: correct impurity_improvement formula (weight by n_left and n_right)
2013-11-22 16:53:26 +01:00
5847be48a5
removed Impurity struct, Criterion.child_impurity now returns impurity_left and impurity_right (not total), fix: moved pos >= end check to if branch
2013-11-22 11:28:33 +01:00
4b45668212
Best-first instead of branch heuristic - identical score as GBM on covertype
2013-11-21 13:43:54 +01:00
13e43d202b
Implement GBM's best-first heuristic tree growing procedure .
2013-11-20 21:10:16 +01:00
e62d196fe7
Merge branch 'master' into gbrt-enh-stackrec-greedy
...
Conflicts:
sklearn/ensemble/_gradient_boosting.c
sklearn/tree/_tree.c
sklearn/tree/_tree.pxd
sklearn/tree/_tree.pyx
sklearn/tree/tree.py
2013-11-08 14:35:33 +02:00
79a67d00ba
updated after tree.pxd change
2013-11-03 14:20:53 +02:00
e9cea77bd8
PriorityQueue for best-first tree growing
...
max_leaf_nodes instead of complete parameter
2013-11-03 09:40:24 +02:00
1192fe64e5
Tree: compute partition specific impurity and total impurity in criterion.children_impurity. Pass partition impurity to stack to avoid re-computation (saves some runtime).
...
New tree parameter: complete; specifies whether complete binary trees are grows or if a greedy branch of max_depth is grown with at most max_depth + 1 leafs.
2013-11-02 01:04:23 +02:00
30eb78de8d
FIX #1622 : OPTIM: remove obsolete random_state instance in the Tree class
...
The random_state instance is now directly handled by the Splitter class,
there is no need to store it as an attribute of the cython Tree class
anymore.
This significantly speeds up the time of unpickling the trees by not having
to reconstruct pickled numpy.random.RandomState instances.
2013-10-29 11:28:12 +01:00
6a5a65dede
ENH: use npy_int32 type
2013-10-09 13:36:02 +02:00
b9384c98cb
ENH: make node_reset/split/value gil-free
2013-10-09 13:35:52 +02:00
c7266ec712
ENH back-port rand_r from 4.4BSD
...
This function is not available on Windows and is deprecated by POSIX.
Fetched from OpenBSD:
http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/stdlib/rand.c
This reverts commit 6f737ad8d5b29a098df6b2ad0fdd1db38e801db8.
2013-07-22 12:20:01 +02:00
0c795872c8
WIP: revert use of rand_r
2013-07-22 12:20:01 +02:00
4c032ac02f
WIP: use re-entrant rand_r
2013-07-22 12:20:01 +02:00
e2363c9974
WIP: even cleaner splitter interface
2013-07-22 12:19:59 +02:00
bf8c786251
fix: proper dtype for SIZE_t
2013-07-22 12:19:57 +02:00
4742806106
gbrt now works with new DecisionTree implementation
2013-07-22 12:19:57 +02:00
67ece78da4
COSMIT various typofixes
...
As suggested by codespell <https://github.com/lucasdemarchi/codespell >
Signed-off-by: Stefano Lattarini <stefano.lattarini@gmail.com>
2013-04-11 21:22:14 +02:00
37c1d092f8
COSMIT shut up the build by calling np.import_array in Cython modules
...
$(make in 2>&1 | wc -l) down from 706 to 517.
Took the liberty of modernizing some of the Cython code while I was at it.
2013-01-10 18:00:21 +01:00
7945e9d7df
FIX: raise an exception if negative number of samples
2013-01-08 23:29:19 +01:00
8dcec88ce8
ENH: early binding + allocate features at tree creation (by @pprett)
2013-01-08 23:29:17 +01:00
cbb8df668b
FIX: roll back some changes
2013-01-08 23:28:52 +01:00
357f3c17d7
Call DecisionTreeRegressor instead of Tree
2012-12-11 09:33:10 +01:00
ef9161bb47
more explicit typing (int32, float64)
2012-11-13 19:57:05 +01:00
63bae39fc1
cosmit: docstrings
2012-11-11 12:37:59 +01:00
df909c74f3
rename dependency -> dependence
2012-10-30 08:45:34 +01:00
021e9f9a68
add learn rate to partial dependency function
2012-10-22 09:26:27 +02:00
7f203451a1
automatically create grid from training data
...
cosmit
docstrings
2012-10-14 19:57:39 +02:00
ab78baa280
cythonized and cosmit
2012-10-12 18:04:59 +02:00
20f654032e
implemented partial dependecy plot
2012-10-11 22:52:26 +02:00
Gilles Louppe