/
HG_ST_labcode.py
351 lines (319 loc) · 15.2 KB
/
HG_ST_labcode.py
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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import numpy as np
from Params import args
import Utils.TimeLogger as logger
from Utils.TimeLogger import log
import Utils.NNLayers as NNs
from Utils.NNLayers import FC, Regularize, Activate, Bias, defineParam, defineRandomNameParam
from DataHandler import DataHandler
import tensorflow as tf
from tensorflow.core.protobuf import config_pb2
import pickle
class Model:
def __init__(self, sess, handler):
self.sess = sess
self.handler = handler
self.metrics = dict()
mets = ['preLoss', 'microF1', 'macroF1']
for i in range(args.offNum):
mets.append('F1_%d' % i)
for met in mets:
self.metrics['Train' + met] = list()
self.metrics['Test' + met] = list()
def makePrint(self, name, ep, reses, save):
ret = 'Epoch %d/%d, %s: ' % (ep, args.epoch, name)
for metric in reses:
val = reses[metric]
ret += '%s = %.4f, ' % (metric, val)
tem = name + metric
if save and tem in self.metrics:
self.metrics[tem].append(val)
ret = ret[:-2] + ' '
return ret
def run(self):
self.prepareModel()
log('Model Prepared')
if args.load_model != None:
self.loadModel()
stloc = len(self.metrics['TrainpreLoss']) * args.tstEpoch
else:
stloc = 0
init = tf.global_variables_initializer()
self.sess.run(init)
log('Variables Inited')
bestRes = None
for ep in range(stloc, args.epoch):
test = (ep % args.tstEpoch == 0)
reses = self.trainEpoch()
log(self.makePrint('Train', ep, reses, test))
if test:
reses = self.testEpoch(self.handler.tstT, np.concatenate([self.handler.trnT, self.handler.valT], axis=1))
if bestRes is None or args.task == 'r' and bestRes['MAPE'] > reses['MAPE'] or args.task == 'c' and bestRes['macroF1'] > reses['macroF1']:
bestRes = reses
if ep % args.tstEpoch == 0:
self.saveHistory()
print()
reses = self.testEpoch(self.handler.tstT, np.concatenate([self.handler.trnT, self.handler.valT], axis=1))
log(self.makePrint('Test', args.epoch, reses, True))
if bestRes is None or args.task == 'r' and bestRes['MAPE'] > reses['MAPE'] or args.task == 'c' and bestRes['macroF1'] > reses['macroF1']:
bestRes = reses
log(self.makePrint('Best', args.epoch, bestRes, True))
self.saveHistory()
def spacialModeling(self, rows, cols, vals, embeds):
# edge, time, offense, latdim
rowEmbeds = tf.nn.embedding_lookup(embeds, rows)
colEmbeds = tf.nn.embedding_lookup(embeds, cols)
Q = defineRandomNameParam([args.latdim, args.latdim], reg=False)
K = defineRandomNameParam([args.latdim, args.latdim], reg=False)
V = defineRandomNameParam([args.latdim, args.latdim], reg=False)
q = tf.reshape(tf.einsum('etod,dl->etol', rowEmbeds, Q), [-1, args.temporalRange, args.offNum, 1, args.head, args.latdim//args.head])
k = tf.reshape(tf.einsum('etod,dl->etol', colEmbeds, K), [-1, args.temporalRange, 1, args.offNum, args.head, args.latdim//args.head])
v = tf.reshape(tf.einsum('etod,dl->etol', colEmbeds, V), [-1, args.temporalRange, 1, args.offNum, args.head, args.latdim//args.head])
att = tf.nn.softmax(tf.reduce_sum(q * k, axis=-1, keep_dims=True) / tf.sqrt(float(args.latdim//args.head)), axis=3)
attV = tf.reshape(tf.reduce_sum(att * v, axis=3), [-1, args.temporalRange, args.offNum, args.latdim])
ret = tf.math.segment_sum(attV * tf.nn.dropout(vals, rate=self.dropRate), rows)
return Activate(ret, 'leakyRelu') # area, time, offense, latdim
def temporalModeling(self, rows, cols, vals, embeds):
prevTEmbeds = tf.slice(embeds, [0, 0, 0, 0], [-1, args.temporalRange-1, -1, -1])
nextTEmbeds = tf.slice(embeds, [0, 1, 0, 0], [-1, args.temporalRange-1, -1, -1])
rowEmbeds = tf.nn.embedding_lookup(nextTEmbeds, rows)
colEmbeds = tf.nn.embedding_lookup(prevTEmbeds, cols)
Q = defineRandomNameParam([args.latdim, args.latdim], reg=False)
K = defineRandomNameParam([args.latdim, args.latdim], reg=False)
V = defineRandomNameParam([args.latdim, args.latdim], reg=False)
q = tf.reshape(tf.einsum('etod,dl->etol', rowEmbeds, Q), [-1, args.temporalRange-1, args.offNum, 1, args.head, args.latdim//args.head])
k = tf.reshape(tf.einsum('etod,dl->etol', colEmbeds, K), [-1, args.temporalRange-1, 1, args.offNum, args.head, args.latdim//args.head])
v = tf.reshape(tf.einsum('etod,dl->etol', colEmbeds, V), [-1, args.temporalRange-1, 1, args.offNum, args.head, args.latdim//args.head])
att = tf.nn.softmax(tf.reduce_sum(q * k, axis=-1, keep_dims=True) / tf.sqrt(float(args.latdim//args.head)), axis=3)
attV = tf.reshape(tf.reduce_sum(att * v, axis=3), [-1, args.temporalRange-1, args.offNum, args.latdim])
ret = tf.math.segment_sum(attV * tf.nn.dropout(vals, rate=self.dropRate), rows)
ret = tf.concat([tf.slice(embeds, [0, 0, 0, 0], [-1, 1, -1, -1]), ret], axis=1)
return Activate(ret, 'leakyRelu') # area, time, offense, latdim
def hyperGNN(self, adj, embeds):
tpadj = tf.transpose(adj)
hyperEmbeds = Activate(tf.einsum('hn,ntod->htod', tf.nn.dropout(adj, rate=self.dropRate), embeds), 'leakyRelu')
retEmbeds = Activate(tf.einsum('nh,htod->ntod', tf.nn.dropout(tpadj, rate=self.dropRate), hyperEmbeds), 'leakyRelu')
return retEmbeds
def ours(self):
offenseEmbeds = defineParam('offenseEmbeds', [1, 1, args.offNum, args.latdim], reg=False)
initialEmbeds = offenseEmbeds * tf.expand_dims(self.feats, axis=-1) # area, time, offense, latdim
areaEmbeds = defineParam('areaEmbeds', [args.areaNum, 1, 1, args.latdim], reg=False)
embeds = [initialEmbeds]# + areaEmbeds]
for l in range(args.spacialRange):
embed = embeds[-1]
embed = self.spacialModeling(self.rows, self.cols, self.vals, embed)
embed = self.hyperGNN(self.hyperAdj, embed) + embed
embeds.append(embed)
embed = tf.add_n(embeds) / args.spacialRange
embeds = [embed]
for l in range(args.temporalGnnRange):
embeds.append(self.temporalModeling(self.rows, self.cols, self.vals, embeds[-1]))
embed = tf.add_n(embeds) / args.temporalGnnRange
embed = tf.reduce_mean(embed, axis=1) # area, offense, latdim
W = defineParam('predEmbeds', [1, args.offNum, args.latdim], reg=False)
if args.task == 'c':
allPreds = Activate(tf.reduce_sum(embed * W, axis=-1), 'sigmoid') # area, offense
elif args.task == 'r':
allPreds = tf.reduce_sum(embed * W, axis=-1)
return allPreds, embed
def prepareModel(self):
self.rows = tf.constant(self.handler.rows)
self.cols = tf.constant(self.handler.cols)
self.vals = tf.reshape(tf.constant(self.handler.vals, dtype=tf.float32), [-1, 1, 1, 1])
self.hyperAdj = defineParam('hyperAdj', [args.hyperNum, args.areaNum], reg=True)
self.feats = tf.placeholder(name='feats', dtype=tf.float32, shape=[args.areaNum, args.temporalRange, args.offNum])
self.dropRate = tf.placeholder(name='dropRate', dtype=tf.float32, shape=[])
self.labels = tf.placeholder(name='labels', dtype=tf.float32, shape=[args.areaNum, args.offNum])
self.preds, embed = self.ours()
if args.task == 'c':
posInd = tf.cast(tf.greater(self.labels, 0), tf.float32)
negInd = tf.cast(tf.less(self.labels, 0), tf.float32)
posPred = tf.cast(tf.greater_equal(self.preds, args.border), tf.float32)
negPred = tf.cast(tf.less(self.preds, args.border), tf.float32)
NNs.addReg('embed', embed * tf.expand_dims(posInd + negInd, axis=-1))
self.preLoss = tf.reduce_sum(-(posInd * tf.log(self.preds + 1e-8) + negInd * tf.log(1 - self.preds + 1e-8))) / (tf.reduce_sum(posInd) + tf.reduce_sum(negInd))
self.truePos = tf.reduce_sum(posPred * posInd, axis=0)
self.falseNeg = tf.reduce_sum(negPred * posInd, axis=0)
self.trueNeg = tf.reduce_sum(negPred * negInd, axis=0)
self.falsePos = tf.reduce_sum(posPred * negInd, axis=0)
elif args.task == 'r':
self.mask = tf.placeholder(name='mask', dtype=tf.float32, shape=[args.areaNum, args.offNum])
self.preLoss = tf.reduce_sum(tf.square(self.preds - self.labels) * self.mask) / tf.reduce_sum(self.mask)
self.sqLoss = tf.reduce_sum(tf.square(self.preds - self.labels) * self.mask, axis=0)
self.absLoss = tf.reduce_sum(tf.abs(self.preds - self.labels) * self.mask, axis=0)
self.tstNums = tf.reduce_sum(self.mask, axis=0)
posMask = self.mask * tf.cast(tf.greater(self.labels, 0.5), tf.float32)
self.apeLoss = tf.reduce_sum(tf.abs(self.preds - self.labels) / (self.labels + 1e-8) * posMask, axis=0)
self.posNums = tf.reduce_sum(posMask, axis=0)
NNs.addReg('embed', embed * tf.expand_dims(self.mask, axis=-1))
self.regLoss = args.reg * Regularize() + args.spreg * tf.reduce_sum(tf.abs(self.hyperAdj))
self.loss = self.preLoss + self.regLoss
globalStep = tf.Variable(0, trainable=False)
learningRate = tf.train.exponential_decay(args.lr, globalStep, args.decay_step, args.decay, staircase=True)
self.optimizer = tf.train.AdamOptimizer(learningRate).minimize(self.loss, global_step=globalStep)
def sampleTrainBatch(self, batIds):
idx = batIds[0]
label = self.handler.trnT[:, idx, :]# area, offNum
if args.task == 'c':
negRate = args.negRate#np.random.randint(1, args.negRate*2)
elif args.task == 'r':
negRate = 0
posNums = np.sum(label != 0, axis=0) * negRate
retLabels = (label != 0) * 1
if args.task == 'r':
mask = retLabels
retLabels = label
for i in range(args.offNum):
temMap = label[:, i]
negPos = np.reshape(np.argwhere(temMap==0), [-1])
sampedNegPos = np.random.permutation(negPos)[:posNums[i]]
# sampedNegPos = negPos
if args.task == 'c':
retLabels[sampedNegPos, i] = -1
elif args.task == 'r':
mask[sampedNegPos, i] = 1
feat = self.handler.trnT[:, idx-args.temporalRange: idx, :]
if args.task == 'c':
return self.handler.zScore(feat), retLabels
elif args.task == 'r':
return self.handler.zScore(feat), retLabels, mask
def trainEpoch(self):
ids = np.random.permutation(list(range(args.temporalRange, args.trnDays)))
epochLoss, epochPreLoss, epochAcc = [0] * 3
num = len(ids)
steps = int(np.ceil(num / args.batch))
for i in range(steps):
st = i * args.batch
ed = min((i+1) * args.batch, num)
batIds = ids[st: ed]
tem = self.sampleTrainBatch(batIds)
if args.task == 'c':
feats, labels = tem
elif args.task == 'r':
feats, labels, mask = tem
targets = [self.optimizer, self.preLoss, self.loss]
feeddict = {self.feats: feats, self.labels: labels, self.dropRate: args.dropRate}
if args.task == 'r':
feeddict[self.mask] = mask
res = self.sess.run(targets, feed_dict=feeddict, options=config_pb2.RunOptions(report_tensor_allocations_upon_oom=True))
preLoss, loss = res[1:]
epochLoss += loss
epochPreLoss += preLoss
log('Step %d/%d: preLoss = %.4f ' % (i, steps, preLoss), save=False, oneline=True)
ret = dict()
ret['Loss'] = epochLoss / steps
ret['preLoss'] = epochPreLoss / steps
return ret
def sampTestBatch(self, batIds, tstTensor, inpTensor):
idx = batIds[0]
label = tstTensor[:, idx, :]# area, offNum
if args.task == 'c':
retLabels = ((label > 0) * 1 + (label == 0) * (-1)) * self.handler.tstLocs
elif args.task == 'r':
retLabels = label
mask = self.handler.tstLocs * (label > 0)
if idx - args.temporalRange < 0:
temT = inpTensor[:, idx-args.temporalRange:, :]
temT2 = tstTensor[:, :idx, :]
feats = np.concatenate([temT, temT2], axis=1)
else:
feats = tstTensor[:, idx-args.temporalRange: idx, :]
if args.task == 'c':
return self.handler.zScore(feats), retLabels
elif args.task == 'r':
return self.handler.zScore(feats), retLabels, mask
def testEpoch(self, tstTensor, inpTensor):
ids = np.random.permutation(list(range(tstTensor.shape[1])))
epochLoss, epochPreLoss, = [0] * 2
if args.task == 'c':
epochTp, epochFp, epochTn, epochFn = [np.zeros(4) for i in range(4)]
elif args.task == 'r':
epochSqLoss, epochAbsLoss, epochTstNum, epochApeLoss, epochPosNums = [np.zeros(4) for i in range(5)]
num = len(ids)
steps = int(np.ceil(num / args.batch))
for i in range(steps):
st = i * args.batch
ed = min((i+1) * args.batch, num)
batIds = ids[st: ed]
tem = self.sampTestBatch(batIds, tstTensor, inpTensor)
if args.task == 'c':
feats, labels = tem
elif args.task == 'r':
feats, labels, mask = tem
if args.task == 'c':
targets = [self.preLoss, self.regLoss, self.loss, self.truePos, self.falsePos, self.trueNeg, self.falseNeg]
feeddict = {self.feats: feats, self.labels: labels, self.dropRate: 0.0}
elif args.task == 'r':
targets = [self.preds, self.preLoss, self.regLoss, self.loss, self.sqLoss, self.absLoss, self.tstNums, self.apeLoss, self.posNums]
feeddict = {self.feats: feats, self.labels: labels, self.dropRate: 0.0, self.mask: mask}
res = self.sess.run(targets, feed_dict=feeddict, options=config_pb2.RunOptions(report_tensor_allocations_upon_oom=True))
if args.task == 'c':
preLoss, regLoss, loss, tp, fp, tn, fn = res
epochTp += tp
epochFp += fp
epochTn += tn
epochFn += fn
elif args.task == 'r':
preds, preLoss, regLoss, loss, sqLoss, absLoss, tstNums, apeLoss, posNums = res
epochSqLoss += sqLoss
epochAbsLoss += absLoss
epochTstNum += tstNums
epochApeLoss += apeLoss
epochPosNums += posNums
epochLoss += loss
epochPreLoss += preLoss
log('Step %d/%d: loss = %.2f, regLoss = %.2f ' % (i, steps, loss, regLoss), save=False, oneline=True)
ret = dict()
ret['preLoss'] = epochPreLoss / steps
if args.task == 'c':
temSum = 0
for i in range(args.offNum):
ret['F1_%d' % i] = epochTp[i] * 2 / (epochTp[i] * 2 + epochFp[i] + epochFn[i])
temSum += ret['F1_%d' % i]
ret['microF1'] = temSum / args.offNum
ret['macroF1'] = np.sum(epochTp) * 2 / (np.sum(epochTp) * 2 + np.sum(epochFp) + np.sum(epochFn))
elif args.task == 'r':
for i in range(args.offNum):
ret['RMSE_%d' % i] = np.sqrt(epochSqLoss[i] / epochTstNum[i])
ret['MAE_%d' % i] = epochAbsLoss[i] / epochTstNum[i]
ret['MAPE_%d' % i] = epochApeLoss[i] / epochPosNums[i]
ret['RMSE'] = np.sqrt(np.sum(epochSqLoss) / np.sum(epochTstNum))
ret['MAE'] = np.sum(epochAbsLoss) / np.sum(epochTstNum)
ret['MAPE'] = np.sum(epochApeLoss) / np.sum(epochPosNums)
return ret
def calcRes(self, preds, temTst, tstLocs):
hit = 0
ndcg = 0
for j in range(preds.shape[0]):
predvals = list(zip(preds[j], tstLocs[j]))
predvals.sort(key=lambda x: x[0], reverse=True)
shoot = list(map(lambda x: x[1], predvals[:args.shoot]))
if temTst[j] in shoot:
hit += 1
ndcg += np.reciprocal(np.log2(shoot.index(temTst[j])+2))
return hit, ndcg
def saveHistory(self):
if args.epoch == 0:
return
with open('History/' + args.save_path + '.his', 'wb') as fs:
pickle.dump(self.metrics, fs)
saver = tf.train.Saver()
saver.save(self.sess, 'Models/' + args.save_path)
log('Model Saved: %s' % args.save_path)
def loadModel(self):
saver = tf.train.Saver()
saver.restore(sess, 'Models/' + args.load_model)
with open('History/' + args.load_model + '.his', 'rb') as fs:
self.metrics = pickle.load(fs)
log('Model Loaded')
if __name__ == '__main__':
logger.saveDefault = True
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
log('Start')
handler = DataHandler()
log('Load Data')
with tf.Session(config=config) as sess:
model = Model(sess, handler)
model.run()