tcRIP/ML_AutoEncoder.py at master · innate2adaptive/tcRIP · 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
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
# -*- coding: utf-8 -*-
"""
Created on Thu Aug  3 11:58:21 2017

@author: lewismoffat


This script runs an AutoEncoder network on all the data for a specified Feat.
Engineering methods.

"""

#==============================================================================
# Module Imports
#==============================================================================

import numpy as np
import dataProcessing as dp
import pdb
import sklearn as sk

from sklearn.model_selection import train_test_split
from scipy.sparse import hstack

import UL_autoencoderModel as ae

#==============================================================================
# Get the data
#==============================================================================

method ="tuple" #"Li" #
loadCDR12=False
tupsize= 2
length = 14
# are we doing the full set
singlePatient=False

# which patient to get data from
patient=['Complete']
chain = "beta"


if singlePatient:
    print('Patient: '+patient[0])
    delim = ["naive",chain]+patient #other delimiters
else:
    print('Patient: All')
    delim = ["naive",chain] #other delimiters


seqs, vj = dp.loadAllPatients(delim) # these gets all the sequences and vj values

# filter out joint sequences
seqs[0], seqs[1], vj[0], vj[1], joint = dp.removeDup(seqs[0], seqs[1], vj[0], vj[1])

#==============================================================================
# Helper Functions
#==============================================================================
def createCDR(seqs,length):
    c1=[]
    c2=[]
    c3=[]
    for idx, seq in enumerate(seqs):
        c1.append(seq[:length])
        c2.append(seq[length:(length+6)])
        c3.append(seq[(length+6):])
    return c1,c2,c3


#==============================================================================
# Feature Engineering
#==============================================================================

if method=="Li":
    # filter the sequnces
    seqs[0]=dp.filtr(seqs[0], 14)
    seqs[1]=dp.filtr(seqs[1], 14)


    if loadCDR12:
        # add extrac cds
        cddict=dp.extraCDs()
        seqs[0], vj[0]=dp.addCDextra(seqs[0],vj[0],cddict)
        seqs[1], vj[1]=dp.addCDextra(seqs[1],vj[1],cddict)
        seqs[0]=dp.filtr(seqs[0], 14+5+6)
        seqs[1]=dp.filtr(seqs[1], 14+5+6)

    seqs[0]=dp.seq2fatch(seqs[0])
    seqs[1]=dp.seq2fatch(seqs[1])

    # use function to create data
    X, y = dp.dataCreator(seqs[0],seqs[1])
else:

    if loadCDR12:
        length=length+5+6
        # add extrac cds
        cddict=dp.extraCDs()
        seqs[0], vj[0]=dp.addCDextra(seqs[0],vj[0],cddict)
        seqs[1], vj[1]=dp.addCDextra(seqs[1],vj[1],cddict)
        #filter
        seqs[0]=dp.filtr(seqs[0], length)
        seqs[1]=dp.filtr(seqs[1], length)

        l4=len(seqs[0])
        l8=len(seqs[1])

        seqs[0]=createCDR(seqs[0],length=(length-11))
        seqs[1]=createCDR(seqs[1],length=(length-11))
        # each seqs[0] has three dimensions now, each containing the cdrs

        c1=dp.char2ptuple(seqs[0][0]+seqs[1][0], n=tupsize)
        c2=dp.char2ptuple(seqs[0][1]+seqs[1][1], n=tupsize)
        c3=dp.char2ptuple(seqs[0][2]+seqs[1][2], n=tupsize)

        print("Size CDR3: {}".format(c1.shape[1]))
        print("Size CDR2: {}".format(c2.shape[1]))
        print("Size CDR1: {}".format(c3.shape[1]))


        X_init=hstack([c1,c2,c3])
        y4=np.zeros((l4))
        y8=np.ones((l8))
        y=np.concatenate((y4,y8))
        X_init=X_init.tocsr()
        # shuffle data
        X, y = sk.utils.shuffle(X_init,y)

    else:
        # use function to create data
        X, y = dp.dataCreator(seqs[0],seqs[1])

        X=dp.char2ptuple(X, n=tupsize)
        X=X.tocsr()

        # shuffle data
        X, y = sk.utils.shuffle(X,y)

X=X[:20000]
y=y[:20000]

X=X.toarray()


#==============================================================================
# Model Spec.
#==============================================================================


aeControllerParams={'batch_size':256,
                     'epochs':1000,
                     'learningRate':0.1}

ModelParams={'learningRate':0.1,
                'embedding_size':10,
                'vocab_size':22,
                'cell_size':128,
                'LSTM':False, # if false it uses GRU
                'stacked':False,
                'dropout':True,
                'unidirectional':True,
                'attention':True,
                'atten_len':2,
                'regu':False, # not being used
                'batch_norm':False, # uses batch norm on the last affine layers
                'onlyLinear':False,
                'conv':False,
                'embed':False,
                'save':True,
                'load':False}


ModelParams['maxLen']=X.shape[1]

xTrain, xVal, yTrain, yVal = train_test_split(X, y, test_size=0.20)

# Spool up warp drives! This gets the rnn controller class going
aeMain = ae.Controller(aeControllerParams, ModelParams)
print("Training AE")
xTrain, xVal = aeMain.train(xTrain, yTrain, xVal, yVal)