def multiplyData(y,node,edge): y_ = [] node_ = [] edge_ = [] N = len(node) for l, n , e in zip(y,node,edge): samples = sampleSubSequences(n.shape[0],extra_samples,min_length_sequence) for s in samples: if copy_start_state: ll = [0] if s[0] > 0: ll = ll + range(s[0],s[1]) else: ll = range(s[0],s[1]) y_.append(l[ll]) node_.append(n[ll,:]) edge_.append(e[ll,:]) else: y_.append(l[s[0]:s[1]]) node_.append(n[s[0]:s[1],:]) edge_.append(e[s[0]:s[1],:]) N += 1 y = y + y_ edge = edge + edge_ node = node + node_ return N,y,node,edge
def multiplyData(features,sample_ratio):
N = 0
for action in actions:
new_samples = []
for f in features[action]:
N += 1
samples = sampleSubSequences(f.shape[0],int(sample_ratio[action]*extra_samples),min_length_sequence)
for s in samples:
N += 1
if copy_start_state:
ll = [0]
if s[0] > 0:
ll = ll + range(s[0],s[1])
else:
ll = range(s[0],s[1])
new_samples.append(f[ll,:])
else:
new_samples.append(f[s[0]:s[1],:])
features[action] = features[action] + new_samples
print '{0} {1}'.format(action,len(features[action]))
return N,features
def multiplyData(features,sample_ratio):
N = 0
for action in actions:
new_samples = []
for f in features[action]:
N += 1
samples = sampleSubSequences(f.shape[0],int(sample_ratio[action]*extra_samples),min_length_sequence)
for s in samples:
N += 1
if copy_start_state:
ll = [0]
if s[0] > 0:
ll = ll + range(s[0],s[1])
else:
ll = range(s[0],s[1])
new_samples.append(f[ll,:])
else:
new_samples.append(f[s[0]:s[1],:])
features[action] = features[action] + new_samples
print '{0} {1}'.format(action,len(features[action]))
return N,features
def multiplyData(y,node,edge,edge_intra,y_object,node_object,edge_object,edge_intra_object,edge_intra_object_human): y_ = [] y_anticipation_ = [] node_ = [] edge_ = [] edge_intra_ = [] y_object_ = [] y_object_anticipation_ = [] node_object_ = [] edge_object_ = [] edge_intra_object_ = [] edge_intra_object_human_ = [] N = len(node) for l in y: y_anticipation_.append(appendToArray(l[1:],11)) for l, n , e, ei, yo, no, eo, eio, eioh in zip(y,node,edge,edge_intra,y_object,node_object,edge_object,edge_intra_object,edge_intra_object_human): samples = sampleSubSequences(n.shape[0],extra_samples,min_length_sequence) for yo_, no_, eo_, eio_, eioh_ in zip(yo, no, eo, eio, eioh): y_object_.append(yo_) y_object_anticipation_.append(appendToArray(yo_[1:],13)) node_object_.append(no_) edge_object_.append(eo_) edge_intra_object_.append(eio_) edge_intra_object_human_.append(eioh_) for s in samples: if copy_start_state: ll = [0] if s[0] > 0: ll = ll + range(s[0],s[1]) else: ll = range(s[0],s[1]) y_.append(l[ll]) node_.append(n[ll,:]) edge_.append(e[ll,:]) edge_intra_.append(ei[ll,:]) new_list = [(x+1) for x in ll] y_anticipation_.append(appendToArray(l,11,new_list)) for yo_, no_, eo_, eio_, eioh_ in zip(yo, no, eo, eio, eioh): y_object_.append(yo_[ll]) node_object_.append(no_[ll,:]) edge_object_.append(eo_[ll,:]) edge_intra_object_.append(eio_[ll,:]) edge_intra_object_human_.append(eioh_[ll,:]) y_object_anticipation_.append(appendToArray(yo_,13,new_list)) else: y_.append(l[s[0]:s[1]]) node_.append(n[s[0]:s[1],:]) edge_.append(e[s[0]:s[1],:]) edge_intra_.append(ei[s[0]:s[1],:]) new_list = range(s[0]+1,s[1]+1) y_anticipation_.append(appendToArray(l,11,new_list)) for yo_, no_, eo_, eio_, eioh_ in zip(yo, no, eo, eio, eioh): y_object_.append(yo_[s[0]:s[1]]) node_object_.append(no_[s[0]:s[1],:]) edge_object_.append(eo_[s[0]:s[1],:]) edge_intra_object_.append(eio_[s[0]:s[1],:]) edge_intra_object_human_.append(eioh_[s[0]:s[1],:]) y_object_anticipation_.append(appendToArray(yo_,13,new_list)) N += 1 y = y + y_ edge = edge + edge_ node = node + node_ edge_intra = edge_intra + edge_intra_ y_anticipation = y_anticipation_ y_object = y_object_ y_object_anticipation = y_object_anticipation_ edge_object = edge_object_ node_object = node_object_ edge_intra_object = edge_intra_object_ edge_intra_object_human = edge_intra_object_human_ return N,y,node,edge,edge_intra,y_object,node_object,edge_object,edge_intra_object,edge_intra_object_human,y_anticipation,y_object_anticipation
def multiplyData(y, node, edge, edge_intra, y_object, node_object, edge_object,
edge_intra_object, edge_intra_object_human):
y_ = []
y_anticipation_ = []
node_ = []
edge_ = []
edge_intra_ = []
y_object_ = []
y_object_anticipation_ = []
node_object_ = []
edge_object_ = []
edge_intra_object_ = []
edge_intra_object_human_ = []
N = len(node)
for l in y:
y_anticipation_.append(appendToArray(l[1:], 11))
for l, n, e, ei, yo, no, eo, eio, eioh in zip(y, node, edge, edge_intra,
y_object, node_object,
edge_object,
edge_intra_object,
edge_intra_object_human):
samples = sampleSubSequences(n.shape[0], extra_samples,
min_length_sequence)
for yo_, no_, eo_, eio_, eioh_ in zip(yo, no, eo, eio, eioh):
y_object_.append(yo_)
y_object_anticipation_.append(appendToArray(yo_[1:], 13))
node_object_.append(no_)
edge_object_.append(eo_)
edge_intra_object_.append(eio_)
edge_intra_object_human_.append(eioh_)
for s in samples:
if copy_start_state:
ll = [0]
if s[0] > 0:
ll = ll + range(s[0], s[1])
else:
ll = range(s[0], s[1])
y_.append(l[ll])
node_.append(n[ll, :])
edge_.append(e[ll, :])
edge_intra_.append(ei[ll, :])
new_list = [(x + 1) for x in ll]
y_anticipation_.append(appendToArray(l, 11, new_list))
for yo_, no_, eo_, eio_, eioh_ in zip(yo, no, eo, eio, eioh):
y_object_.append(yo_[ll])
node_object_.append(no_[ll, :])
edge_object_.append(eo_[ll, :])
edge_intra_object_.append(eio_[ll, :])
edge_intra_object_human_.append(eioh_[ll, :])
y_object_anticipation_.append(
appendToArray(yo_, 13, new_list))
else:
y_.append(l[s[0]:s[1]])
node_.append(n[s[0]:s[1], :])
edge_.append(e[s[0]:s[1], :])
edge_intra_.append(ei[s[0]:s[1], :])
new_list = range(s[0] + 1, s[1] + 1)
y_anticipation_.append(appendToArray(l, 11, new_list))
for yo_, no_, eo_, eio_, eioh_ in zip(yo, no, eo, eio, eioh):
y_object_.append(yo_[s[0]:s[1]])
node_object_.append(no_[s[0]:s[1], :])
edge_object_.append(eo_[s[0]:s[1], :])
edge_intra_object_.append(eio_[s[0]:s[1], :])
edge_intra_object_human_.append(eioh_[s[0]:s[1], :])
y_object_anticipation_.append(
appendToArray(yo_, 13, new_list))
N += 1
y = y + y_
edge = edge + edge_
node = node + node_
edge_intra = edge_intra + edge_intra_
y_anticipation = y_anticipation_
y_object = y_object_
y_object_anticipation = y_object_anticipation_
edge_object = edge_object_
node_object = node_object_
edge_intra_object = edge_intra_object_
edge_intra_object_human = edge_intra_object_human_
return N, y, node, edge, edge_intra, y_object, node_object, edge_object, edge_intra_object, edge_intra_object_human, y_anticipation, y_object_anticipation
