analyse_replay = True
TFR = True
save = False
verbose = True
f_in = "wmx_%s_%.1f_linear.pkl" % (
STDP_mode, place_cell_ratio) if linear else "wmx_%s_%.1f.pkl" % (
STDP_mode, place_cell_ratio)
PF_pklf_name = os.path.join(base_path, "files", "PFstarts_%s_linear.pkl" %
place_cell_ratio) if linear else None
dir_name = os.path.join(
base_path, "figures", "%.2f_replay_det_%s_%.1f" %
(1, STDP_mode, place_cell_ratio)) if linear else None
wmx_PC_E = load_wmx(os.path.join(base_path, "files",
f_in)) * 1e9 # *1e9 nS conversion
SM_PC, SM_BC, RM_PC, RM_BC, selection, StateM_PC, StateM_BC = run_simulation(
wmx_PC_E, STDP_mode, cue=cue, save=save, seed=seed, verbose=verbose)
_ = analyse_results(SM_PC,
SM_BC,
RM_PC,
RM_BC,
selection,
StateM_PC,
StateM_BC,
seed=seed,
multiplier=1,
linear=linear,
pklf_name=PF_pklf_name,
dir_name=dir_name,
analyse_replay=analyse_replay,
Ap = 0.01
Am = -Ap
wmax = 4e-8 # S
scale_factor = 1.27
elif STDP_mode == "sym":
taup = taum = 62.5 * ms
Ap = Am = 4e-3
wmax = 2e-8 # S
scale_factor = 0.62
Ap *= wmax; Am *= wmax # needed to reproduce Brian1 results
f_name = os.path.join(base_path, "files", f_in)
spiking_neurons, spike_times = load_spike_trains(f_name)
pklf_name = os.path.join(base_path, "files", f_in_wmx)
intermediate_weightmx = load_wmx(pklf_name) / scale_factor # (scale only once, at the end)
weightmx = learning_2nd_env(spiking_neurons, spike_times, taup, taum, Ap, Am, wmax, intermediate_weightmx)
weightmx *= scale_factor # quick and dirty additional scaling! (in an ideal world the STDP parameters should be changed to include this scaling...)
pklf_name = os.path.join(base_path, "files", f_out)
save_wmx(weightmx, pklf_name)
plot_wmx(weightmx, save_name=f_out[:-4])
plot_wmx_avg(weightmx, n_pops=100, save_name="%s_avg"%f_out[:-4])
plot_w_distr(weightmx, save_name="%s_distr"%f_out[:-4])
selection = np.array([500, 2400, 4000, 5500, 7015])
plot_weights(save_selected_w(weightmx, selection), save_name="%s_sel_weights"%f_out[:-4])
plt.show()
return np.max(EPSP) - Vrest_Pyr/mV
if __name__ == "__main__":
try:
n = int(sys.argv[1])
except:
n = 500 # number of random weights
v_hold = -70. # mV
i_hold = 20.967 # pA (calculated by `clamp_cell.py`)
f_in = "wmx_sym_0.5_linear.pkl"
pklf_name = os.path.join(base_path, "files", "paper", f_in)
wmx_PC_E = load_wmx(pklf_name) * 1e9 # *1e9 nS conversion
nonzero_weights = wmx_PC_E[np.nonzero(wmx_PC_E)]
print "mean(nonzero weights): %s (nS)"%np.mean(nonzero_weights)
np.random.seed(12345)
weights = np.random.choice(nonzero_weights, n, replace=False)
EPSPs = np.zeros((n, 4000)); EPSCs = np.zeros((n, 4000)) # 4000 is hard coded for sim length
peak_EPSPs = np.zeros(n); peak_EPSCs = np.zeros(n)
for i, weight in enumerate(tqdm(weights)):
t_, EPSP, EPSC = sym_paired_recording(weight, i_hold)
EPSPs[i,:] = EPSP; EPSCs[i,:] = EPSC
peak_EPSPs[i] = get_peak_EPSP(t_, EPSP, i_hold, v_hold)
peak_EPSCs[i] = np.max(EPSC)
if __name__ == "__main__":
try:
STDP_mode = sys.argv[1]
except:
STDP_mode = "sym"
assert STDP_mode in ["sym", "asym"]
place_cell_ratio = 0.5
linear = True
f_in = "wmx_%s_%.1f_linear.pkl" % (
STDP_mode, place_cell_ratio) if linear else "wmx_%s_%.1f.pkl" % (
STDP_mode, place_cell_ratio)
pklf_name = os.path.join(base_path, "files", f_in)
wmx_orig = load_wmx(pklf_name)
#wmx_modified = shuffle(wmx_orig); f_out = "%s_shuffled_linear.pkl"%f_in[:-11] if linear else "%s_shuffled.pkl"%f_in[:-4]
#wmx_modified = column_shuffle(wmx_orig); f_out = "%s_cshuffled_linear.pkl"%f_in[:-11] if linear else "%s_cshuffled.pkl"%f_in[:-4]
#wmx_modified = shuffle_blocks(wmx_orig); f_out = "%s_block_shuffled_linear.pkl"%f_in[:-11] if linear else "%s_block_shuffled.pkl"%f_in[:-4]
wmx_modified = binarize(wmx_orig)
f_out = "%s_binary_linear.pkl" % f_in[:
-11] if linear else "%s_binary.pkl" % f_in[:
-4]
assert np.shape(wmx_modified) == (
nPCs, nPCs), "Output shape is not %i*%i" % (nPCs, nPCs)
assert (wmx_modified >=
0.0).all(), "Negative weights in the modified matrix!"
pklf_name = os.path.join(base_path, "files", f_out)