def R3_resize(R3, E_all):
E3, E2, E1 = E_all
R3_new = np.zeros(resized_shape(R3.shape, E_all), dtype=R3.dtype)
R3_new[-(R3.shape[0] - E3.center) :, :, :] = R3[
: end(E3.center), (E2.center + E3.center) : end(E2.center), E1.center : end(E1.center)
]
try:
return MetaArray(R3_new, ticks=resize_ticks(R3.ticks, E_all), rw_freq=R3.rw_freq)
except AttributeError:
return R3_new
item.reset_index().to_pickle('../feature/{}/f210_product.p'.format(folder))
#==============================================================================
# main
#==============================================================================
make(0)
make(1)
make(2)
#make(3)
#make(4)
#make(5)
make(-1)
utils.end(__file__)
convolution_operator(E2, L2),
# convolution_operator(E2, L2)[:end(E1.center), :],
convolution_operator(E3, L3, trim_left_boundary=False),
]
d1, d2, d3 = (len(conv_matrices[i]) for i in xrange(3))
M1 = L3 + L2 + d1
M2 = L3 + d2
M3 = d3
shapes = [(L3, L2, L1), (L3, L2, M1), (L3, M2, M1), (M3, M2, M1)]
slice_maps = [[], [], []]
for t3 in xrange(L3):
for t2 in xrange(L2):
slice_maps[0].append(
(
conv_matrices[0][E1.center : end(E1.center)],
(t3, t2, slice(t3 + t2 + E1.center, t3 + t2 + d1 - E1.center)),
(t3, t2),
)
)
# slice_maps[0].append((conv_matrices[0],
# (t3, t2, slice(t3 + t2, t3 + t2 + d1)),
# (t3, t2)))
for t3 in xrange(L3):
for sum12 in xrange(M1):
slice_maps[1].append((conv_matrices[1], (t3, slice(t3, t3 + d2), sum12), (t3, slice(None), sum12)))
for sum23 in xrange(M2):
for sum123 in xrange(M1):
sl = slice(max(0, sum23 + len(E3) + E2.center - L3), min(M3, sum23 + 1))
{
# fixed
'objective': 'binary',
'metric': 'auc',
'learning_rate': 0.2,
'max_bin': 100,
'nthread': 64,
'bagging_freq': 10,
# optimize
'max_depth': 3,
'num_leaves': 2**3 - 1,
'scale_pos_weight': 100,
'min_child_weight': 0.001,
'subsample': 0.1,
'colsample_bytree': 0.5,
'lambda_l1': 0,
'lambda_l2': 5,
# fixed?
'min_child_samples': 300,
'seed': np.random.randint(9999)
},
]
for param in params:
do_lgb(param)
#==============================================================================
utils.end(__file__)
convolution_operator(E1, L1),
convolution_operator(E2, L2),
# convolution_operator(E2, L2)[:end(E1.center), :],
convolution_operator(E3, L3, trim_left_boundary=False)
]
d1, d2, d3 = (len(conv_matrices[i]) for i in xrange(3))
M1 = L3 + L2 + d1
M2 = L3 + d2
M3 = d3
shapes = [(L3, L2, L1), (L3, L2, M1), (L3, M2, M1), (M3, M2, M1)]
slice_maps = [[], [], []]
for t3 in xrange(L3):
for t2 in xrange(L2):
slice_maps[0].append(
(conv_matrices[0][E1.center:end(E1.center)],
(t3, t2, slice(t3 + t2 + E1.center,
t3 + t2 + d1 - E1.center)), (t3, t2)))
# slice_maps[0].append((conv_matrices[0],
# (t3, t2, slice(t3 + t2, t3 + t2 + d1)),
# (t3, t2)))
for t3 in xrange(L3):
for sum12 in xrange(M1):
slice_maps[1].append(
(conv_matrices[1], (t3, slice(t3, t3 + d2), sum12),
(t3, slice(None), sum12)))
for sum23 in xrange(M2):
for sum123 in xrange(M1):
sl = slice(max(0, sum23 + len(E3) + E2.center - L3),
#==============================================================================
# DATA LOAD
#=============================================================================='''
)
pri = utils.read_df_pickle(path='../input/prior*.p')
tra = utils.read_df_pickle(path='../input/train*.p')
order = utils.read_df_pickle(path='../input/order*.p')
logger.info(f'''
#==============================================================================
# MAKE EDA TABLE
#=============================================================================='''
)
pri_eda = eda.df_info(pri)
pri_eda.to_csv('../eda/prior_eda.csv')
tra_eda = eda.df_info(tra)
tra_eda.to_csv('../eda/train_eda.csv')
order_eda = eda.df_info(order)
order_eda.to_csv('../eda/orders_eda.csv')
sys.exit()
clean_app(app)
clean_prev(pre)
# clean_pos(pos)
# clean_ins(ins)
# clean_ccb(ccb)
utils.end(sys.argv[0])
# pre_eda = eda.df_info(pre)
def end(message):
if message.chat.id != config.group_id:
config.the_bot.forward_message(config.group_id, message.chat.id,
message.message_id)
utils.end(message)
