def update(self, samples, axis=0) -> None:
"""
Update the online calculation of the mean and variance.
Notes:
Modifies the metrics in place.
Args:
samples: data samples
Returns:
None
"""
# compute the sample size and sample mean
n = len(samples)
sample_mean = be.tsum(samples, axis=axis) / max(n, 1)
sample_square = be.tsum(be.square(be.subtract(sample_mean, samples)),
axis=axis)
if self.mean is None:
self.mean = be.zeros_like(sample_mean)
self.square = be.zeros_like(sample_square)
self.var = be.zeros_like(sample_square)
self.num = 0
delta = sample_mean - self.mean
new_num = self.num + n
correction = n * self.num * be.square(delta) / max(new_num, 1)
self.square += sample_square + correction
self.var = self.square / max(new_num - 1, 1)
self.mean = (self.num * self.mean + n * sample_mean) / max(new_num, 1)
self.num = new_num
def update(self, samples, axis=0) -> None:
"""
Update the online calculation of the mean.
Notes:
Modifies the metrics in place.
Args:
samples: data samples
Returns:
None
"""
# compute the sample size and sample mean
n = len(samples)
sample_mean = be.tsum(samples, axis=axis) / n
# initialize the num and mean attributes if necessary
if self.mean is None:
self.mean = be.zeros_like(sample_mean)
self.num = 0
# update the num and mean attributes
tmp = self.num * self.mean + n * sample_mean
self.num += n
self.mean = tmp / self.num
def compute_reconstructions(rbm,
v_data,
fit,
n_recon=10,
vertical=False,
num_to_avg=1):
v_model = be.zeros_like(v_data)
# Average over n reconstruction attempts
for k in range(num_to_avg):
data_state = State.from_visible(v_data, rbm)
visible = data_state.units[0]
reconstructions = fit.DrivenSequentialMC(rbm)
reconstructions.set_state(data_state)
dropout_scale = State.dropout_rescale(rbm)
reconstructions.update_state(1, dropout_scale)
v_model += rbm.deterministic_iteration(1, reconstructions.state,
dropout_scale).units[0]
v_model /= num_to_avg
idx = numpy.random.choice(range(len(v_model)), n_recon, replace=False)
grid = numpy.array(
[[be.to_numpy_array(visible[i]),
be.to_numpy_array(v_model[i])] for i in idx])
if vertical:
return grid
else:
return grid.swapaxes(0, 1)
def compute_reconstructions(rbm, v_data, n_recon=10, vertical=False, num_to_avg=1):
v_model = be.zeros_like(v_data)
# Average over n reconstruction attempts
for k in range(num_to_avg):
reconstructions = rbm.compute_reconstructions(v_data)
v_model += reconstructions.get_visible() / num_to_avg
idx = np.random.choice(range(len(v_model)), n_recon, replace=False)
grid = np.array([[be.to_numpy_array(v_data[i]),
be.to_numpy_array(v_model[i])] for i in idx])
if vertical:
return grid
else:
return grid.swapaxes(0,1)