def init_sdrs():
problems = []
folder_name = os.path.join('Data', 'encoded')
for sub_folder in os.listdir(folder_name):
f = os.path.join(folder_name, sub_folder)
problems += parse_images.get_problems(f)
for problem in problems:
assert problem['SDRs'].shape == (9, 32, 32)
problem['Input'] = problem['SDRs'][:3].reshape((3, -1))
problem['Output'] = problem['SDRs'][3:].reshape((6, -1))
num_cols = len(problems[0]['Input'][0])
bckTM = BTM(numberOfCols=num_cols,
cellsPerColumn=15,
initialPerm=0.5,
connectedPerm=0.5,
minThreshold=10,
newSynapseCount=10,
activationThreshold=10,
pamLength=10)
layers = [(bckTM, bk_tm_compute)]
return (layers, problems)
def main(args):
if args.data == 'imgs':
folder_name = 'Data/Problems'
problems = parse_images.get_problems(folder_name)
elif args.data == 'sdrs':
folder_name = 'Data/Problems_sdr'
problems = parse_images.get_problems(folder_name)
elif args.data == 'symbolic':
folder_name = 'Data/Problems_txt'
problems = read_symbolic_problems.get_problems(folder_name)
cell_size = len(problems[0]['Input'][0])
agent = LSTMAgent(cell_size)
print('Train')
agent.train(problems[:5], 30)
print('Test')
agent.run(problems[:5])
def main(args):
problems = []
if args.data == 'imgs':
#folder_name = 'Data/Problems'
#problems = parse_images.get_problems(folder_name)
folder_name = 'Data/raw'
for sub_folder in os.listdir(folder_name):
f = os.path.join(folder_name, sub_folder)
problems += parse_images.get_problems(f)
elif args.data == 'sdrs':
#folder_name = 'Data/Problems_sdr'
#problems = parse_images.get_problems(folder_name)
folder_name = 'Data/encoded'
for sub_folder in os.listdir(folder_name):
f = os.path.join(folder_name, sub_folder)
problems += parse_images.get_problems(f)
elif args.data == 'symbolic':
if args.nn == 'conv':
print(
"Please use the fully connected network for this type of data."
)
return
folder_name = 'Data/desc'
problems = read_symbolic_problems.get_yml_problems(folder_name)
if args.nn == 'fc':
# fully connected
problems = reshape_fc(problems)
print(problems[0]['Input'].shape)
agent = Agent()
agent.init_fc(len(problems[0]['Input'][0]))
errs = agent.train(problems, args.nn, 1500, args.ex)
if args.nn == 'conv':
# conv
problems = reshape_fc(problems)
print(problems[0]['Input'].shape)
agent = Agent()
agent.init_conv(len(problems[0]['Input'][0]))
errs = agent.train(problems, args.nn, 500, args.ex)
def init_imgs(write_output):
#problems = parse_images.get_problems(folder_name)
folder_name = 'Data/raw'
problems = []
for sub_folder in os.listdir(folder_name):
f = os.path.join(folder_name, sub_folder)
problems += parse_images.get_problems(f)
print(len(problems))
for problem in problems:
problem['Input'] = problem['Input'].reshape((3, -1))
problem['Output'] = problem['Output'].reshape((6, -1))
# dimenstions
num_cols1 = len(problems[0]['Input'][0])
tm_cols = num_cols1
layers = []
if not write_output:
num_cols2 = num_cols1 / 2
num_cols3 = num_cols2 / 2
sp1 = SpatialPooler(inputDimensions=(num_cols1, ),
columnDimensions=(num_cols2, ),
numActiveColumnsPerInhArea=-1,
localAreaDensity=0.05)
sp2 = SpatialPooler(inputDimensions=(num_cols2, ),
columnDimensions=(num_cols3, ),
numActiveColumnsPerInhArea=-1,
localAreaDensity=0.05)
tm_cols = num_cols3
layers = [(sp1, sp_compute), (sp2, sp_compute)]
bckTM = BTM(numberOfCols=tm_cols,
cellsPerColumn=10,
initialPerm=0.5,
connectedPerm=0.5,
minThreshold=10,
newSynapseCount=10,
activationThreshold=10,
pamLength=10)
layers += [(bckTM, bk_tm_compute)]
return (layers, problems)
_, cost = self.sess.run([self.train_step, self.err],
feed_dict={
self.X: window,
self.lr: learning_rate
})
errs += [cost]
return errs
if __name__ == '__main__':
folder_name = 'Data/raw'
problems = []
for sub_folder in os.listdir(folder_name):
f = os.path.join(folder_name, sub_folder)
problems += get_problems(f)
print(len(problems))
#folder_name = 'Data/Problems'
#problems = get_problems(folder_name)
input_windows = get_windows(problems)
(num_windows, height, width) = input_windows.shape
print("Get Windows Done")
'''
for win in input_windows:
print(np.sum(win) * 100 / (height * width))
'''
enc = Encoder(height * width)
print("Encoder Done")
errs = enc.train(input_windows)
print("Train Done")
print("Show decoded data")