# data file
index = 0 # sys.argv[1]
rat = datafiles[int(index)]
print rat
f = '/tigress/fdamani/neuro_data/data/raw/allrats_withmissing_limitedtrials/csv/'
f += rat
rat = f.split('/')[-1].split('.csv')[0]
# add to dir name
savedir += '__obs' + str(num_obs_samples)
savedir += '__' + rat
# os.mkdir(savedir)
# f = '../data/W066_short.csv'
x, y, rw = read_and_process(num_obs_samples, f, savedir=savedir)
rw = torch.tensor(rw, dtype=dtype, device=device)
dim = x.shape[2]
T = x.shape[0]
init_prior = ([0.0] * dim, [math.log(1.0)] * dim)
transition_scale = [math.log(1.0)] #* dim
num_future_steps = 1
category_tt_split = 'single'
num_mc_samples = 10
y, x, y_future, x_future = train_future_split(y, x, num_future_steps)
y_train, y_test, test_inds = train_test_split(y, x, cat=category_tt_split)
x = torch.tensor(x, dtype=dtype, device=device)
y_train = torch.tensor(y_train, dtype=dtype, device=device)
y_test = torch.tensor(y_test, dtype=dtype, device=device)
test_inds = torch.tensor(test_inds, dtype=torch.long, device=device)
os.makedirs(output_file + '/plots')
savedir = output_file
raw_datasets = []
for ind in inds:
index = ind
rat = datafiles[int(index)]
print rat
if torch.cuda.is_available():
f = '/tigress/fdamani/neuro_data/data/raw/allrats_withmissing_limitedtrials/csv/'
else:
f = '../data/'
f += rat
rat = f.split('/')[-1].split('.csv')[0]
x, y, rw = read_and_process(num_obs_samples,
f,
savedir=savedir,
proficient=False)
# x = x[0:500]
# y = y[0:500]
# rw = rw[0:500]
rw = torch.tensor(rw, dtype=dtype, device=device)
z_true = None
true_model_params = None
dim = x.shape[2]
T = x.shape[0]
raw_datasets.append((x, y, rw, z_true, true_model_params, dim, T, rat))
datasets = []
for dataset in raw_datasets:
x, y, rw, z_true, true_model_params, dim, T, rat = dataset
# rat = f.split('/')[-1].split('.csv')[0]
# num_obs_samples=1
#output_file += '/'+rat
output_file += '/kfold_single_alpha_results_2k'
os.makedirs(output_file)
folds = np.arange(1, len(datafiles))
held_out_marginal_lhs = []
for fd in folds:
rat = datafiles[fd]
f = '/tigress/fdamani/neuro_data/data/raw/allrats_withmissing_limitedtrials/csv/'
#f = '../data/'
#rat = 'W073.csv'
f += rat
num_obs_samples = 1
x, y, rw = read_and_process(num_obs_samples, f, savedir=output_file)
T = 2000
#T = 50
#num_particles = 100
num_particles = 1000
x = x[0:T]
y = y[0:T]
x = torch.tensor(x, dtype=dtype, device=device)
y = torch.tensor(y, dtype=dtype, device=device)
dim = x.shape[-1]
md = LearningDynamicsModel(dim)
ev = HeldOutRat([y, x], md)
model_params_file = '/tigress/fdamani/neuro_output/5.5/single_alpha_shared_model_kfold_leave_out_' + str(