def predict_w_noise(self, xs, stochastic=True, **kwargs):
ts_ys, ts_ms, _ = self.ts_predict_all(array_to_ts(xs),
stochastic=stochastic,
**kwargs)
ys, ms = ts_to_array(ts_ys), ts_to_array(ts_ms)
ns = ys - ms
return ys, ms
def kl(self, other, xs, reversesd=False, **kwargs):
""" Return the KL divergence for each data point in the batch xs. """
return ts_to_array(self.ts_kl(other, array_to_ts(xs), reversesd=reversesd))
def mean_variable(self):
return flatten(ts_to_array(super().ts_variables))
def grad(self, x, **kwargs):
return flatten(ts_to_array(self.ts_grad(array_to_ts(x), **kwargs)))
def fun(self, x, **kwargs):
return ts_to_array(self.ts_fun(array_to_ts(x)), **kwargs)
def grad(self, xs, **kwargs):
""" Derivative with respect to xs. """
return ts_to_array(self.ts_grad(array_to_ts(xs), **kwargs))