def Machine(jEE=1, extE=1, dist=3):
### Get Standard Parameters ###
info = numParam()
_, toDo = doParam()
### Specify Parameters ###
toDo["switch"] = 1
info["j_EE"] = jEE
info["extM"][0] = extE
info["dist"] = dist
info['recNum'] = 100 # info['sizeM'][0]
### Create constant inputs to function ###
jCon = nn.createjCon(info)
external = nn.createExt(info)
thresh = nn.createThresh(info["sizeM"], info["threshM"], toDo["doThresh"])
### Run System ###
(indiNeuronsDetailed, activeOT, fireOT,
labels_train) = nn.run_update(jCon, thresh, external, info, toDo)
activeOT = np.array(activeOT)
###
correctness = predic_readout(activeOT, labels_train, info)
contrib = np.mean(indiNeuronsDetailed[:info["recNum"]][:], axis=(0, 1))
overall_activity = contrib[3]
ratio_ext_int = contrib[3] / contrib[4]
useless(indiNeuronsDetailed, fireOT)
return correctness, overall_activity, ratio_ext_int, internal
def changeThresh_():
check_plt()
### Specify Parameters
name = "thresh"
info = numParam(name)
(drw, toDo) = doParam()
### Create constant inputs to function
jCon = nn.createjCon(info)
external = nn.createExt(info)
doThresh = ["constant", "gauss", "bound"]
for i in range(len(doThresh)):
toDo["doThresh"] = doThresh[i]
thresh = nn.createThresh(info["sizeM"], info["threshM"],
toDo["doThresh"])
(
indiNeuronsDetailed,
activeOT,
fireOT,
_ # label
) = nn.run_update(
jCon,
thresh,
external,
info,
toDo,
)
plot_machine(activeOT, fireOT, indiNeuronsDetailed, info, drw, toDo)
def changeExt_():
"""
"""
check_plt()
name = "m_vs_m0"
info = numParam(name)
toDo = doParam()[1]
info["meanExt"] = 0.04
info["timer"] = 50
jCon = nn.createjCon(info)
thresh = nn.createThresh(info["sizeM"], info["threshM"], toDo["doThresh"])
mE_List = np.arange(0.04, 0.32, .02)
# mE_List = np.arange(0.04, 0.36,.08)
meanList = []
meanList2 = []
for i in range(len(mE_List)):
external = nn.createExt(info, mE_List[i])
activeOT = nn.run_update(jCon, thresh, external, info,
toDo)[1] # 1:active, 2:fire
means = eva.analyzeMeanOT(activeOT, info["sizeM"])
meanList.append([mE_List[i]])
meanList2.append([mE_List[i]])
meanList[-1] += [np.mean(means[i][10:]) for i in range(2)]
meanList2[-1] += [np.mean(means[i]) for i in range(2)]
meanList = np.transpose(meanList)
meanList2 = np.transpose(meanList2)
print(meanList)
name = "m_vs_m0_K%d.npy" % info["K"]
np.save(name, meanList)
plots.figfolder_GLOBAL = info["figfolder"]
plots.mean_vs_ext(meanList)
def plain_vanilla(dur, start, swi=0):
info = numParam()
info['timer'] = dur #info['sizeM'][0]
info['meanStartActi'] = start
(drw, toDo) = doParam()
toDo["switch"] = swi
### Create constant inputs to function
jCon = nn.createjCon(info)
external = nn.createExt(info)
thresh = nn.createThresh(info["sizeM"], info["threshM"], toDo["doThresh"])
### Updating Function
(indiNeuronsDetailed, activeOT, fireOT, label) = nn.run_update(
jCon,
thresh,
external,
info,
toDo,
)
return activeOT
def Vanilla():
"""
Iterates over whole process once according to specifications in numParam and doParam
Records results.
"""
timestamp = [["start", time.time(), 0]]
### Specify Parameters
# mode_list = ['count_up', 'static', 'permute', 'rand']
mode = "rand"
name = "classic" + "_" + mode
info = numParam(name)
(drw, toDo) = doParam()
toDo["switch"] = 0
toDo["update_mode"] = mode
### Create constant inputs to function
jCon = nn.createjCon(info)
external = nn.createExt(info)
thresh = nn.createThresh(info["sizeM"], info["threshM"], toDo["doThresh"])
stamp(timestamp, "setup done")
### Updating Function
(indiNeuronsDetailed, activeOT, fireOT, label) = nn.run_update(
jCon,
thresh,
external,
info,
toDo,
)
stamp(timestamp, "calculations done")
### Save Results
valueFolder = plots.describe(toDo, info, 0, 1)
utils.saveResults(valueFolder, indiNeuronsDetailed, activeOT, fireOT, info,
toDo)
stamp(timestamp, "saving done")
def hamming():
flip = 1
name = "hamming3"
info = numParam()
(_, toDo) = doParam()
jCon = nn.createjCon(info)
external = nn.createExt(info)
thresh = nn.createThresh(info["sizeM"], info["threshM"], toDo["doThresh"])
info["timer"] = 50
hamm_comb = []
rand_dist = []
for i in range(10):
orig = nn.createNval(info["sizeM"], info["meanStartActi"])
changed = nn.createNval(info["sizeM"], info["meanStartActi"])
dist = np.mean(abs(orig - changed))
rand_dist.append(dist)
for control in range(2):
orig_list = []
chang_list = []
dist_list = []
for i in range(10):
rand_seed = i + 100
np.random.seed(rand_seed)
random.seed(rand_seed)
info["flip"] = 0
(indiNeuronsDetailed, activeOT, fireOT, label) = nn.run_update(
jCon,
thresh,
external,
info,
toDo,
)
orig = (transform2binary(activeOT, info["timer"]))
if not control:
np.random.seed(rand_seed)
random.seed(rand_seed)
info["flip"] = flip
(indiNeuronsDetailed, activeOT, fireOT, label) = nn.run_update(
jCon,
thresh,
external,
info,
toDo,
)
changed = (transform2binary(activeOT, info["timer"]))
dist = np.mean(abs(orig - changed), axis=1)
if control:
dist = np.concatenate(([rand_dist[i]], dist))
else:
dist = np.concatenate(([flip / np.sum(info["sizeM"])], dist))
orig_mean = np.concatenate(
([info['meanStartActi']], np.mean(orig, axis=1)))
chang_mean = np.concatenate(
([info['meanStartActi']], np.mean(changed, axis=1)))
orig_list.append(orig_mean)
chang_list.append(chang_mean)
dist = i
dist_list.append(dist)
hamming = np.array(dist_list)
hamm_comb.append(hamming)
print(hamming.shape)
print(hamm_comb)