Mercurial > pylearn

comparison doc/v2_planning/API_learner.txt @ 1174:fe6c25eb1e37

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
merge
author pascanur
date Fri, 17 Sep 2010 16:13:58 -0400
parents 77b6ed85d3f7
children 805e7c369fd1
comparison
equal deleted inserted replaced
1173:a0f178bc9052 1174:fe6c25eb1e37
1 # A list of "task types"
2
3 '''
4 List of tasks types:
5 Attributes
6
7 sequential
8 spatial
9 structured
10 semi-supervised
11 missing-values
12
13
14 Supervised (x,y)
15
16 classification
17 regression
18 probabilistic classification
19 ranking
20 conditional density estimation
21 collaborative filtering
22 ordinal regression ?= ranking
23
24 Unsupervised (x)
25
26 de-noising
27 feature learning ( transformation ) PCA, DAA
28 density estimation
29 inference
30
31 Other
32
33 generation (sampling)
34 structure learning ???
35
36
37 Notes on metrics & statistics:
38 - some are applied to an example, others on a batch
39 - most statistics are on the dataset
40 '''
41
42
43 class Learner(Object):
44 '''
45 Takes data as inputs, and learns a prediction function (or several).
46
47 A learner is parametrized by hyper-parameters, which can be set from the
48 outside (a "client" from Learner, that can be a HyperLearner, a
49 Tester,...).
50
51 The data can be given all at a time as a data set, or incrementally.
52 Some learner need to be fully trained in one step, whereas other can be
53 trained incrementally.
54
55 The question of statistics collection during training remains open.
56 '''
57 #def use_dataset(dataset)
58
59 # return a dictionary of hyperparameters names(keys)
60 # and value(values)
61 def get_hyper_parameters()
62 def set_hyper_parameters(dictionary)
63
64
65
66
67 # Ver B
68 def eval(dataset)
69 def predict(dataset)
70
71 # Trainable
72 def train(dataset) # train until complition
73
74 # Incremental
75 def use_dataset(dataset)
76 def adapt(n_steps =1)
77 def has_converged()
78
79 #
80
81
82 # Some example cases
83
84 class HyperLearner(Learner):
85
86 ### def get_hyper_parameter_distribution(name)
87 def set_hyper_parameters_distribution(dictionary)
88
89
90 def bagging(learner_factory):
91 for i in range(N):
92 learner_i = learner_factory.new()
93 # todo: get dataset_i ??
94 learner_i.use_dataset(dataset_i)
95 learner_i.train()