def config_and_test(index):
global file_name, DSTR, add_strat_nn, writer, scan_spec
# checks if we have any more params to scan
if index < len(params):
# gets the current param, for instance epochs: [100,200,300] 100 200 and 300 are param
for param in scan_spec[params[index]]:
# metaprogramming stuff, just sets the param value, e.g. epoch = 100, then recurses down to the next index
exec (params[index] + '=' + str(param))
config_and_test(index + 1)
else:
print settings.strategies
# this part is broken, it should print what the current params are but I can't get it to work
print settings.scan_spec
# the file name is the current time
file_name = datetime.datetime.now().strftime("%Y%m%d%H%M%S")
full_name = os.path.join(os.path.join(os.path.dirname(__file__), 'test_csv'), file_name + '.csv')
with open(full_name, 'wb') as csvfile:
# initialize the writer, DSTR, neural network for each config we want to test
writer = csv.writer(csvfile)
writer.writerow(['Output Format Version', '20150807'])
DSTR = Distribution()
ADD.main()
add_strat_nn = counting_network()
test(settings.n_problems)
def exec_strategy():
global writer
global strat_list
ADD.PPA()
# try getting a random number from a list above the confidence criterion
retrieval = add_strat_nn.guess(ADD.ADDEND.ad1, ADD.ADDEND.ad2, 0, 13)
SOLUTION = 0
if retrieval is not None:
#trp(1, "Used Retrieval")
SOLUTION = retrieval
writer.writerow(["used","retrieval", ADD.ADDEND.ad1, ADD.ADDEND.ad2, SOLUTION])
else:
# retrieval failed, so we get try to get a strategy from above the confidence criterion and use hands to add
strat_num = add_strat_nn.guess(ADD.ADDEND.ad1, ADD.ADDEND.ad2, 13, 13 + len(settings.strategies))
if strat_num is None:
strat_num = randint(0, len(settings.strategies) - 1)
else:
strat_num -= 13
SOLUTION = ADD.exec_strategy(settings.strategies[strat_num])
# !!! WWW WARNING: This gets displayed even if Dynamic
# Retrieval was used. You have to Analyze this distinction out
# of the log at the end by seeing that a DR message appeared!
writer.writerow(["used",settings.strategies[strat_num], ADD.ADDEND.ad1, ADD.ADDEND.ad2, SOLUTION])
# update the neural networks based on if the strategy worked or not
add_strat_nn.update(ADD.ADDEND.ad1, ADD.ADDEND.ad2, SOLUTION, 13 + strat_num, 13, 13 + len(settings.strategies))
add_strat_nn.update(ADD.ADDEND.ad1, ADD.ADDEND.ad2, SOLUTION, ADD.ADDEND.ad1 + ADD.ADDEND.ad2, 0, 13)
add_strat_nn.fit(add_strat_nn.X, add_strat_nn.y, settings.learning_rate, settings.epoch)
add_strat_nn.update_y()
# add method here to get what strategy is used
return [ADD.ADDEND.ad1, ADD.ADDEND.ad2, SOLUTION]
def testFloats(self):
for x in xrange(-10, 10):
for y in xrange(-10, 10):
x = x / 10.0
y = y / 10.0
p = ADD.add(x, y)
self.failUnless(p == x + y, 'Float multiplication faild')
def compute_filamentarity(self):
if self.nodata:
return
if args.verbose:
print("-------------------> Computing Additional Parameters!")
self.filamentarity = ADD.filamentarity(self.segmap)
if args.verbose:
print("Computed Filamentarity :\t F=%.3f"%(self.filamentarity))
print("-------------------> Done computing Filamentarity!")
import ADD
def add(a, b, c, d):
return a + b + c + d
print ADD.add(1, 2)
print add(1, 2, 3, 4)
#print a.add(1,2)
def testIntergers(self):
for x in xrange(-10, 10):
for y in xrange(-10, 10):
p = ADD.add(x, y)
self.failUnless(p == x + y, 'Inter multiplication faild')
