"lg_XBA.pkl": 0,
"lg_XBC.pkl": 0,
"lg_XCC.pkl": 0,
"lg_XBCCR.pkl": 0,
"lg_XBCR.pkl": 0,
"lg_XCCR.pkl": 0,
"lg_XBCOV.pkl": 0,
"lg_XB.pkl": 0,
"lg_XC.pkl": 0
}
path = "/storage/gablab001/data/genus/fs_cog/pred_diag/lg/results"
for key, val in respath.items():
respath[key] = os.path.join(path, key)
auc_df = pd.DataFrame(
columns=[key.split('_')[1].replace('.pkl', '') for key in respath.keys()])
for key, val in respath.items():
col = key.split('_')[1].replace('.pkl', '')
data = utils.read_pickle(val)['auc']
auc_df[col] = [i[1] for i in data]
order = ['XBA', 'XBC', 'XBCCR', 'XBCOV', 'XBCR', 'XB', 'XCC', 'XCCR', 'XC']
sns.violinplot(auc_df, order=order)
plt.title("Distributions of AUC")
fig = plt.gcf()
fig.set_size_inches(12, 7)
fig.savefig('lg_auc_dist.png', dpi=400, bbox_inches="tight")
plt.close()
result_name = sys.argv[1]
header_name = sys.argv[2]
save_name = sys.argv[3]
brain = Brain("fsaverage", "split", "inflated",
views=['lat', 'med'], background="white")
hp = "/storage/gablab001/data/genus/fs_cog/pred_diag/column_headers"
rp = "/storage/gablab001/data/genus/fs_cog/pred_diag/lg/results"
with open(os.path.join(hp, header_name), "r") as tmp:
headers = np.array(tmp.read().strip("\n").split("\n"))
coefs = utils.read_pickle(os.path.join(rp,"{}".format(result_name)))['coef']
nz = lambda x: np.nonzero(x)[0]
features_selected_lh = []
features_selected_rh = []
features_all = []
from collections import Counter
import pandas as pd
for coef in coefs:
idx = nz(coef[1])
features_all.extend(headers[idx].tolist())
for col in headers[idx].tolist():
if 'lh' in col:
features_selected_lh.append(col)
elif 'rh' in col:
header_name = sys.argv[2]
save_name = sys.argv[3]
brain = Brain("fsaverage",
"split",
"inflated",
views=['lat', 'med'],
background="white")
hp = "/storage/gablab001/data/genus/fs_cog/pred_diag/column_headers"
rp = "/storage/gablab001/data/genus/fs_cog/pred_diag/lg/results"
with open(os.path.join(hp, header_name), "r") as tmp:
headers = np.array(tmp.read().strip("\n").split("\n"))
coefs = utils.read_pickle(os.path.join(rp, "{}".format(result_name)))['coef']
nz = lambda x: np.nonzero(x)[0]
features_selected_lh = []
features_selected_rh = []
features_all = []
from collections import Counter
import pandas as pd
for coef in coefs:
idx = nz(coef[1])
features_all.extend(headers[idx].tolist())
for col in headers[idx].tolist():
if 'lh' in col:
features_selected_lh.append(col)
elif 'rh' in col:
"et_XBC.pkl": 0,
"et_XCC.pkl": 0,
"et_XBCCR.pkl": 0,
"et_XBCR.pkl": 0,
"et_XCCR.pkl": 0,
"et_XBCOV.pkl": 0,
"et_XB.pkl": 0,
"et_XC.pkl": 0
}
path = "/storage/gablab001/data/genus/fs_cog/pred_diag/extra_trees/results/"
for key, val in respath.items():
respath[key] = os.path.join(path, key)
auc_df = pd.DataFrame(
columns=[key.split('_')[1].replace(".pkl", "") for key in respath.keys()])
for key, val in respath.items():
col = key.split('_')[1].replace(".pkl", "")
data = utils.read_pickle(val)
#auc_df[col] = [v for k, v in data.items() if "auc" in k]
auc_df[col] = [a[1] for a in data['auc']]
order = ['XBA', 'XBC', 'XBCCR', 'XBCOV', 'XBCR', 'XB', 'XCC', 'XCCR', 'XC']
sns.violinplot(auc_df, order=order)
plt.title("Distributions of AUC")
fig = plt.gcf()
fig.set_size_inches(12, 7)
fig.savefig('et_auc_dist.png', dpi=400, bbox_inches="tight")
plt.close()
#def extract(res, k2):
# k1 = k2[-1]
# tmp = [(int(key.split(k1)[1]), val) for
# key, val in res.items() if k2 in key]
# return sorted(tmp, key=lambda tup: tup[0])
def mask2imp(mask, imp):
mask_copy = mask.copy().astype(float)
mask_copy[mask_copy > 0] = imp
return mask_copy
for key_i, val_i in dnh.items():
result = utils.read_pickle(os.path.join(dp, key_i))
header = np.genfromtxt(os.path.join(hp, val_i), dtype=str)
#coef = extract(result, "coef")
#aucs = [i[1] for i in extract(result, "auc")]
#coef = [c[1] for c in result['coef']]
aucs = [a[1] for a in result['auc']]
#fimp_mat = np.array([coef[i][1] for i in range(len(coef))])
fimp_mat = np.array([c[1] for c in result['coef']])
wf_imp = (np.abs(fimp_mat) * np.array(aucs)[:, None]).sum(0) / np.sum(aucs)
df = pd.DataFrame(columns=header)
df.loc[0, :] = wf_imp
fn = equate_n_sort(all_headers, header)
data_save.append(pd.concat([fn, df]).fillna(0).sum(0))
def most_imp(data, nth_p):
from custom import utils
from collections import Counter
import matplotlib.pyplot as plt
# command line inputs
result_name = sys.argv[1]
header_name = sys.argv[2]
save_name = sys.argv[3]
# headers
hp = "/storage/gablab001/data/genus/fs_cog/pred_diag/column_headers"
with open(os.path.join(hp, header_name), "r") as tmp:
headers = np.array(tmp.read().strip("\n").split("\n"))
coefs = utils.read_pickle("../{}".format(result_name))['coef']
nz = lambda x: np.nonzero(x)[0]
cog_scores = ['SOP', 'RPS', 'VLM']
cols = []
for coef in coefs:
selected = headers[nz(coef[1])]
cols.extend(selected.tolist())
counts = dict(Counter(cols))
to_keep = {}
for key, val in counts.items():
s = 'DOMAIN'
if ('SOP_' + s in key) or ('VLM_' + s in key) or ('RPS_' + s in key):
to_keep[key] = val
# imp = imp.tolist()
# for idx in range(max_mask):
# if mask[idx]:
# out_imp.append(imp[idx])
# else:
# imp.insert(idx, 0)
# out_imp.append(0)
# return np.array(out_imp)
def mask2imp(mask, imp):
mask_copy = mask.copy().astype(float)
mask_copy[mask_copy > 0] = imp
return mask_copy
for key_i, val_i in dnh.items():
result = utils.read_pickle(os.path.join(dp, key_i))
header = np.genfromtxt(os.path.join(hp, val_i), dtype=str)
#fimp = extract(result, "fimp")
#aucs = [i[1] for i in extract(result, "auc")]
#masks = extract(result, "mask")
fimp = np.array([f[1] for f in result['fimp']])
aucs = [a[1] for a in result['auc']]
masks = np.array([m[1] for m in result['mask']])
fimp_mat = np.array([mask2imp(masks[i], fimp[i]) for i in range(len(fimp))])
#fimp_mat = np.array([fimp[i] for i in range(len(fimp))])
wf_imp =(np.abs(fimp_mat)*np.array(aucs)[:, None]).sum(0)/np.sum(aucs)
df = pd.DataFrame(columns=header)
df.loc[0, :] = wf_imp
fn = equate_n_sort(all_headers, header)
data_save.append(pd.concat([fn, df]).fillna(0).sum(0))
respath = {
"lg_XBA.pkl":0,
"lg_XBC.pkl":0,
"lg_XCC.pkl":0,
"lg_XBCCR.pkl":0,
"lg_XBCR.pkl":0,
"lg_XCCR.pkl":0,
"lg_XBCOV.pkl":0,
"lg_XB.pkl":0,
"lg_XC.pkl":0
}
path = "/storage/gablab001/data/genus/fs_cog/pred_diag/lg/results"
for key, val in respath.items():
respath[key] = os.path.join(path, key)
auc_df = pd.DataFrame(columns=[key.split('_')[1].replace('.pkl', '') for key in respath.keys()])
for key, val in respath.items():
col = key.split('_')[1].replace('.pkl', '')
data = utils.read_pickle(val)['auc']
auc_df[col] = [i[1] for i in data]
order = ['XBA','XBC','XBCCR','XBCOV','XBCR','XB','XCC','XCCR','XC']
sns.violinplot(auc_df, order=order)
plt.title("Distributions of AUC")
fig = plt.gcf()
fig.set_size_inches(12, 7)
fig.savefig('lg_auc_dist.png', dpi=400, bbox_inches="tight")
plt.close()
from custom import utils
from collections import Counter
import matplotlib.pyplot as plt
# command line inputs
result_name = sys.argv[1]
header_name = sys.argv[2]
save_name = sys.argv[3]
# headers
hp = "/storage/gablab001/data/genus/fs_cog/pred_diag/column_headers"
with open(os.path.join(hp, header_name), "r") as tmp:
headers = np.array(tmp.read().strip("\n").split("\n"))
coefs = utils.read_pickle("../{}".format(result_name))['coef']
nz = lambda x: np.nonzero(x)[0]
cog_scores = ['SOP','RPS','VLM']
cols = []
for coef in coefs:
selected = headers[nz(coef[1])]
cols.extend(selected.tolist())
counts = dict(Counter(cols))
to_keep = {}
for key, val in counts.items():
s = 'DOMAIN'
if ('SOP_'+s in key) or ('VLM_'+s in key) or ('RPS_'+s in key):
to_keep[key] = val