def plot_corrected_depths_train_all(trained_models,
chromosome_number=1, max_y=2):
for trained_model in trained_models:
train_sampled_data, _, _ = utility_train.unpickle(trained_model['path'])
observed_depth_mean = _compute_observed_depth_mean(train_sampled_data)
_plot_corrected_depths(train_sampled_data, observed_depth_mean, chromosome_number,
title='sample of training data: ' + trained_model['annotation'], min_y=0, max_y=max_y)
def plot_costs_versus_training_size(trained_models,
semilogx=True, title=None):
training_set_sizes = []
costs_train = []
costs_dev = []
for trained_model in trained_models:
training_set_sizes.append(get_train_args(trained_model['path'])['total number of train examples'])
_, _, cost_versus_epoch = utility_train.unpickle(trained_model['path'])
last_record = cost_versus_epoch.to_dict('records')[-1]
costs_train.append(last_record['cost_train'])
costs_dev.append(last_record['cost_dev'])
fig = plt.figure()
format_fig(fig)
axis = fig.add_subplot(111)
plot = axis.semilogx if semilogx else axis.plot
plot(training_set_sizes, costs_train, 'o', label='train cost')
plot(training_set_sizes, costs_dev, 'o', label='dev cost')
axis.set_xlabel('training set size')
if title:
axis.set_title(title)
format_axis(axis)
plt.show()
def plot_corrected_depths_dev_all(trained_models,
chromosome_number=1, show_title=True):
for trained_model in trained_models:
train_sampled_data, dev_data, _ = utility_train.unpickle(trained_model['path'])
observed_depth_mean = _compute_observed_depth_mean(train_sampled_data)
title = 'dev data: ' + trained_model['annotation'] if show_title else ''
figure_file_name = trained_model['figure_file_name'] if 'figure_file_name' in trained_model else None
_plot_corrected_depths(dev_data, observed_depth_mean, chromosome_number,
title=title, min_y=0, max_y=3, figure_file_name=figure_file_name)
def plot_costs_all(trained_models,
marker='-', start_epoch=0.01, end_epoch=1000, min_cost=2, max_cost=200,
loglog=False, logx=False, logy=False):
for trained_model in trained_models:
train_sampled_data, _, cost_versus_epoch = utility_train.unpickle(trained_model['path'])
observed_depth_mean = _compute_observed_depth_mean(train_sampled_data)
print('observed_depth_mean', observed_depth_mean)
_plot_costs(cost_versus_epoch, marker,
_minimum_achievable_cost(train_sampled_data),
_feature_independent_cost(train_sampled_data, observed_depth_mean),
start_epoch, end_epoch, min_cost, max_cost, title=trained_model['annotation'],
loglog=loglog, logx=logx, logy=logy)
def plot_depths_dev_all(trained_models,
chromosome_number=1, min_depth=0, max_depth=None, logy_scale=False):
for trained_model in trained_models:
_, dev_sampled_data, _ = utility_train.unpickle(trained_model['path'])
if max_depth:
_plot_depths(dev_sampled_data, chromosome_number,
title=trained_model['annotation'],
min_depth=min_depth, max_depth=max_depth, logy_scale=logy_scale)
else:
_plot_depths(dev_sampled_data, chromosome_number,
title=trained_model['annotation'],
min_depth=min_depth, max_depth=trained_model['max_depth'], logy_scale=logy_scale)
def plot_corrected_depths_test_all(trained_models,
normalized_depths=True, corrected_depths=True,
chromosome_number=1, min_y=0, max_y=2,
show_title=True, grid=True):
for trained_model in trained_models:
if 'depth_file_name' in trained_model:
from utility import named_tuple
from load_preprocess_data import read_depths
depths = read_depths(named_tuple({'chromosome_number': chromosome_number,
'depth_file_name': trained_model['depth_file_name']}))
from load_preprocess_data import compute_observed_depth_mean
observed_depth_mean = compute_observed_depth_mean(depths, chromosome_number)
else:
train_sampled_data, _, _ = utility_train.unpickle(trained_model['path'])
observed_depth_mean = _compute_observed_depth_mean(train_sampled_data)
test_data = pd.read_pickle(os.path.join(trained_model['path'], trained_model['test_file_name']))
title = trained_model['annotation'] if show_title else ''
figure_file_name = trained_model['figure_file_name'] if 'figure_file_name' in trained_model else None
_plot_corrected_depths(test_data, observed_depth_mean, chromosome_number,
title=title, min_y=min_y, max_y=max_y,
normalized_depths=normalized_depths, corrected_depths=corrected_depths,
figure_file_name=figure_file_name,
grid=grid)