def calc_metrics(self, data, history, dataset, logs):
y_true = []
predictions = []
B = self.batch_size
for i in range(0, len(data[0]), B):
if self.verbose == 1:
print("\r\tdone {}/{}".format(i, len(data[0])))
if self.target_repl:
(x, y, y_repl) = (data[0][i:i + B], data[1][0][i:i + B],
data[1][1][i:i + B])
else:
(x, y) = (data[0][i:i + B], data[1][i:i + B])
outputs = self.model.predict(x, batch_size=B)
if self.target_repl:
predictions += list(np.array(outputs[0]).flatten())
else:
predictions += list(np.array(outputs).flatten())
y_true += list(np.array(y).flatten())
print("\n")
predictions = np.array(predictions)
predictions = np.stack([1 - predictions, predictions], axis=1)
ret = metrics.print_metrics_binary(y_true, predictions)
for k, v in ret.iteritems():
logs[dataset + '_' + k] = v
history.append(ret)
def calc_metrics(self, data_gen, history, dataset, logs):
y_true = []
predictions = []
for i in range(data_gen.steps):
if self.verbose == 1:
print "\r\tdone {}/{}".format(i, data_gen.steps),
(x, y) = next(data_gen)
pred = self.model.predict(x, batch_size=self.batch_size)
if self.deep_supervision:
for m, t, p in zip(x[1].flatten(), y.flatten(), pred.flatten()):
if np.equal(m, 1):
y_true.append(t)
predictions.append(p)
else:
y_true += list(y.flatten())
predictions += list(pred.flatten())
print "\n"
predictions = np.array(predictions)
predictions = np.stack([1 - predictions, predictions], axis=1)
ret = metrics.print_metrics_binary(y_true, predictions)
for k, v in ret.iteritems():
logs[dataset + '_' + k] = v
history.append(ret)
def calc_metrics(self, data_gen, history, dataset, logs):
y_true = []
predictions = []
for i in range(data_gen.steps):
if self.verbose == 1:
print "\r\tdone {}/{}".format(i, data_gen.steps),
(x, y) = next(data_gen)
pred = self.model.predict(x, batch_size=self.batch_size)
if self.deep_supervision:
for m, t, p in zip(x[1].flatten(), y.flatten(), pred.flatten()):
if np.equal(m, 1):
y_true.append(t)
predictions.append(p)
else:
y_true += list(y.flatten())
predictions += list(pred.flatten())
print "\n"
predictions = np.array(predictions)
predictions = np.stack([1 - predictions, predictions], axis=1)
ret = metrics.print_metrics_binary(y_true, predictions)
for k, v in ret.iteritems():
logs[dataset + '_' + k] = v
history.append(ret)
def calc_metrics(self, data, history, dataset, logs):
y_true = []
predictions = []
B = self.batch_size
for i in range(0, len(data[0]), B):
if self.verbose == 1:
print "\r\tdone {}/{}".format(i, len(data[0])),
if self.target_repl:
(x, y, y_repl) = (data[0][i:i + B], data[1][0][i:i + B], data[1][1][i:i + B])
else:
(x, y) = (data[0][i:i + B], data[1][i:i + B])
outputs = self.model.predict(x, batch_size=B)
if self.target_repl:
predictions += list(np.array(outputs[0]).flatten())
else:
predictions += list(np.array(outputs).flatten())
y_true += list(np.array(y).flatten())
print "\n"
predictions = np.array(predictions)
predictions = np.stack([1 - predictions, predictions], axis=1)
ret = metrics.print_metrics_binary(y_true, predictions)
for k, v in ret.iteritems():
logs[dataset + '_' + k] = v
history.append(ret)
def calc_metrics(self, data, history, dataset, logs):
y_true = []
predictions = []
for i in range(0, len(data[0]), self.batch_size):
if self.verbose == 1:
print "\r\tdone {}/{}".format(i, len(data[0])),
(x, y) = (data[0][i:i + self.batch_size],
data[1][i:i + self.batch_size])
if len(y) == 2: # target replication
y_true += list(y[0].flatten())
else:
y_true += list(np.array(y).flatten())
outputs = self.model.predict(x, batch_size=self.batch_size)
if len(outputs) == 2:
predictions += list(outputs[0].flatten())
else:
predictions += list(outputs.flatten())
print "\n"
predictions = np.array(predictions)
predictions = np.stack([1 - predictions, predictions], axis=1)
ret = metrics.print_metrics_binary(y_true, predictions)
for k, v in ret.iteritems():
logs[dataset + '_' + k] = v
history.append(ret)
def calc_metrics(self, data_gen, history, dataset, logs):
ihm_y_true = []
decomp_y_true = []
los_y_true = []
pheno_y_true = []
ihm_pred = []
decomp_pred = []
los_pred = []
pheno_pred = []
for i in range(data_gen.steps):
if self.verbose == 1:
print("\r\tdone {}/{}".format(i, data_gen.steps))
(X, y, los_y_reg) = data_gen.next(return_y_true=True)
outputs = self.model.predict(X, batch_size=self.batch_size)
ihm_M = X[1]
decomp_M = X[2]
los_M = X[3]
if not data_gen.target_repl: # no target replication
(ihm_p, decomp_p, los_p, pheno_p) = outputs
(ihm_t, decomp_t, los_t, pheno_t) = y
else: # target replication
(ihm_p, _, decomp_p, los_p, pheno_p, _) = outputs
(ihm_t, _, decomp_t, los_t, pheno_t, _) = y
los_t = los_y_reg # real value not the label
# ihm
for (m, t, p) in zip(ihm_M.flatten(), ihm_t.flatten(),
ihm_p.flatten()):
if np.equal(m, 1):
ihm_y_true.append(t)
ihm_pred.append(p)
# decomp
for (m, t, p) in zip(decomp_M.flatten(), decomp_t.flatten(),
decomp_p.flatten()):
if np.equal(m, 1):
decomp_y_true.append(t)
decomp_pred.append(p)
# los
if los_p.shape[-1] == 1: # regression
for (m, t, p) in zip(los_M.flatten(), los_t.flatten(),
los_p.flatten()):
if np.equal(m, 1):
los_y_true.append(t)
los_pred.append(p)
else: # classification
for (m, t, p) in zip(los_M.flatten(), los_t.flatten(),
los_p.reshape((-1, 10))):
if np.equal(m, 1):
los_y_true.append(t)
los_pred.append(p)
# pheno
for (t, p) in zip(pheno_t.reshape((-1, 25)),
pheno_p.reshape((-1, 25))):
pheno_y_true.append(t)
pheno_pred.append(p)
print("\n")
# ihm
print("\n ================= 48h mortality ================")
ihm_pred = np.array(ihm_pred)
ihm_pred = np.stack([1 - ihm_pred, ihm_pred], axis=1)
ret = metrics.print_metrics_binary(ihm_y_true, ihm_pred)
for k, v in ret.iteritems():
logs[dataset + '_ihm_' + k] = v
# decomp
print("\n ================ decompensation ================")
decomp_pred = np.array(decomp_pred)
decomp_pred = np.stack([1 - decomp_pred, decomp_pred], axis=1)
ret = metrics.print_metrics_binary(decomp_y_true, decomp_pred)
for k, v in ret.iteritems():
logs[dataset + '_decomp_' + k] = v
# los
print("\n ================ length of stay ================")
if self.partition == 'log':
los_pred = [metrics.get_estimate_log(x, 10) for x in los_pred]
ret = metrics.print_metrics_log_bins(los_y_true, los_pred)
if self.partition == 'custom':
los_pred = [metrics.get_estimate_custom(x, 10) for x in los_pred]
ret = metrics.print_metrics_custom_bins(los_y_true, los_pred)
if self.partition == 'none':
ret = metrics.print_metrics_regression(los_y_true, los_pred)
for k, v in ret.iteritems():
logs[dataset + '_los_' + k] = v
# pheno
print("\n =================== phenotype ==================")
pheno_pred = np.array(pheno_pred)
ret = metrics.print_metrics_multilabel(pheno_y_true, pheno_pred)
for k, v in ret.iteritems():
logs[dataset + '_pheno_' + k] = v
history.append(logs)
def on_epoch_end(self, epoch, logs={}):
print("\n==>predicting on validation")
yhat = self.model.predict(self.val_X, batch_size=self.batch_size)
yhat = np.array(yhat)[:, 0]
result = metrics.print_metrics_binary(np.array(self.val_y), yhat)
self.results.append(result)
def calc_metrics(self, data_gen, history, dataset, logs):
ihm_y_true = []
decomp_y_true = []
los_y_true = []
pheno_y_true = []
ihm_pred = []
decomp_pred = []
los_pred = []
pheno_pred = []
for i in range(data_gen.steps):
if self.verbose == 1:
print "\r\tdone {}/{}".format(i, data_gen.steps),
(X, y, los_y_reg) = data_gen.next(return_y_true=True)
outputs = self.model.predict(X, batch_size=self.batch_size)
ihm_M = X[1]
decomp_M = X[2]
los_M = X[3]
if not data_gen.target_repl: # no target replication
(ihm_p, decomp_p, los_p, pheno_p) = outputs
(ihm_t, decomp_t, los_t, pheno_t) = y
else: # target replication
(ihm_p, _, decomp_p, los_p, pheno_p, _) = outputs
(ihm_t, _, decomp_t, los_t, pheno_t, _) = y
los_t = los_y_reg # real value not the label
# ihm
for (m, t, p) in zip(ihm_M.flatten(), ihm_t.flatten(), ihm_p.flatten()):
if np.equal(m, 1):
ihm_y_true.append(t)
ihm_pred.append(p)
# decomp
for (m, t, p) in zip(decomp_M.flatten(), decomp_t.flatten(), decomp_p.flatten()):
if np.equal(m, 1):
decomp_y_true.append(t)
decomp_pred.append(p)
# los
if los_p.shape[-1] == 1: # regression
for (m, t, p) in zip(los_M.flatten(), los_t.flatten(), los_p.flatten()):
if np.equal(m, 1):
los_y_true.append(t)
los_pred.append(p)
else: # classification
for (m, t, p) in zip(los_M.flatten(), los_t.flatten(), los_p.reshape((-1, 10))):
if np.equal(m, 1):
los_y_true.append(t)
los_pred.append(p)
# pheno
for (t, p) in zip(pheno_t.reshape((-1, 25)), pheno_p.reshape((-1, 25))):
pheno_y_true.append(t)
pheno_pred.append(p)
print "\n"
# ihm
print "\n ================= 48h mortality ================"
ihm_pred = np.array(ihm_pred)
ihm_pred = np.stack([1 - ihm_pred, ihm_pred], axis=1)
ret = metrics.print_metrics_binary(ihm_y_true, ihm_pred)
for k, v in ret.iteritems():
logs[dataset + '_ihm_' + k] = v
# decomp
print "\n ================ decompensation ================"
decomp_pred = np.array(decomp_pred)
decomp_pred = np.stack([1 - decomp_pred, decomp_pred], axis=1)
ret = metrics.print_metrics_binary(decomp_y_true, decomp_pred)
for k, v in ret.iteritems():
logs[dataset + '_decomp_' + k] = v
# los
print "\n ================ length of stay ================"
if self.partition == 'log':
los_pred = [metrics.get_estimate_log(x, 10) for x in los_pred]
ret = metrics.print_metrics_log_bins(los_y_true, los_pred)
if self.partition == 'custom':
los_pred = [metrics.get_estimate_custom(x, 10) for x in los_pred]
ret = metrics.print_metrics_custom_bins(los_y_true, los_pred)
if self.partition == 'none':
ret = metrics.print_metrics_regression(los_y_true, los_pred)
for k, v in ret.iteritems():
logs[dataset + '_los_' + k] = v
# pheno
print "\n =================== phenotype =================="
pheno_pred = np.array(pheno_pred)
ret = metrics.print_metrics_multilabel(pheno_y_true, pheno_pred)
for k, v in ret.iteritems():
logs[dataset + '_pheno_' + k] = v
history.append(logs)
