Representation-based-Multilabel-Learning/train_rep.py at master · algorithmdog/Representation-based-Multilabel-Learning · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
#!/bin/python
import sys
import os
path  = os.path.split(os.path.realpath(__file__))[0]

from latent_factor import *
from arffio        import *
from common        import *
import copy
import logging, Logger
import pickle
import numpy as np
import scipy.sparse as sp
import sampler
import random
import time
from common       import *
from train_common import *

np.random.seed(0)
random.seed(0)

def printUsages():
    print "Usage: train_rep.py [options] train_file model_file"
    print "options"
    print "  -h  hidden_space_dimension: set the hidden space dimension (default 100)"
    print "  -ha hidden_activation: set the hidden activation(default 0)"
    print "       0 -- tanh"
    print "       1 -- linear"
    print "       2 -- relu"
    print "  -oa output_activation: set the output activation(default 0)"
    print "       0 -- sgmoid"
    print "       1 -- linear"
    print "  -l  loss_function: set the loss function(default 0)"
    print "       0 -- negative_log_likelihood"
    print "       1 -- least_sqaure"
    print "  -l2 l2_regularization: set the l2 regularization(default 0.001)"
    print "  -b  batch_size: set the batch size (default 100)"
    print "  -i  number_of_iter: set the number of iteration(default 10)"
    print "  -st using_sampling: set whether using the sampling scheme(default 1)"
    print "       0 -- not using sampling scheme"
    print "       1 -- using sampling scheme"
    print "  -sr sampling_ratio: set the sampling ratio(default 5)"
    print "  -sp sparse_threhold: set the threhold (default 0.01)"
    print "  -m  using_external_memory: set using the external memory (default 0). Now you can't use external memory and we will implement the function as soon as possible"
    print "       0 -- not using external memory"
    print "       1 -- using external memory"
    print "  -r  learning_rate: set the learning rate (default 0.001)"


def parseParameter(argv):
    if len(argv) < 3: #at least 4 paramters: train.py train_file model_file
        printUsages()
        exit(1)

    parameters = copy.deepcopy(rep_default_params)
    parameters["train_file"]   = argv[len(argv) - 2]
    parameters["model_file"]   = argv[len(argv) - 1]

    i = 1
    while i + 1 < len(argv) - 2:
        if  "-h" == argv[i]:
            parameters["h"]     = int(argv[i+1])
        elif "-ha" == argv[i]:
            parameters["ha"]    = ha_map[int(argv[i+1])]
        elif "-oa" == argv[i]:
            parameters["oa"]    = oa_map[int(argv[i+1])]
        elif "-l" == argv[i]:
            parameters["l"]     = lo_map[int(argv[i+1])]
        elif  "-l2" == argv[i]:
            parameters["l2"]    = float(argv[i+1])
        elif "-b" == argv[i]:
            parameters["b"]     = int(argv[i+1])
        elif "-i" == argv[i]:
            parameters["i"]     = int(argv[i+1])
        elif "-st" == argv[i]:
            parameters["st"]    = st_map[int(argv[i+1])];
        elif "-sr" == argv[i]:
            parameters["sr"]    = float(argv[i+1])
        elif "-sp" == argv[i]:
            parameters["sp"]    = float(argv[i+1])
        elif "-m"  == argv[i]:
            parameters["m"]     = m_map[int(argv[i+1])]
        elif "-r"  == argv[i]:
            parameters["r"]     = float(argv[i+1])
        else:
            print argv[i]
            printUsages()
            exit(1)
        i += 2

    if False == checkParamValid(parameters):
        printUsages()
        exit(1)

    return parameters


def main(argv):
    parameters = parseParameter(argv)

    train_file  = parameters["train_file"]
    model_file  = parameters["model_file"]

    # read a instance to know the number of features and labels
    train_reader       = SvmReader(train_file, 1)
    x, y, has_next     = train_reader.read()
    parameters["nx"]   = x.shape[1]
    parameters["ny"]   = y.shape[1]
    train_reader.close()

    model        = Model(parameters)
    rater        = AdaGrad(model)
    model.rater  = rater
    thrsel       = ThresholdSel()
    model.thrsel = thrsel

    if m.internal_memory == parameters["m"]:
        model = train_internal(model, train_file, parameters)
    elif m.external_memory == parameters["m"]:
        model = train_external(model, train_file, parameters)
    else:
        logger = logging.getLogger(Logger.project_name)
        logger.error("Invalid m param")
        raise Exception("Invalid m param")

    #write the model
    #model.clear_for_save()
    model.save(model_file)
    #s = pickle.dumps(model)
    #f = open(model_file, "w")
    #f.write(s)
    #f.close()

if __name__ == "__main__":
    main(sys.argv)