def _predict(self, x, confidence=False): ps = np.array([predict(x, self.W[i], self.b[i]) for i in range(self.k)]) c = np.argmax(ps) if confidence: return c, ps[c] else: return c
def _predict(self, x, confidence=False): ps = np.array([predict(x, self.W[i], self.b[i]) for i in range(self.k)]) c = np.argmax(ps) if confidence: return c,ps[c] else: return c
def probdist(self, x): """The probability distribution of class assignment. Transforms the confidence scores into a probability via a logit function :math:`\exp{\mathbf{w}^T \mathbf{x} + b} / Z`. :return: a `k`-dimensional probability vector. """ ps = np.array([np.exp(predict(x, self.W[i], self.b[i])) for i in range(self.k)]) Z = np.sum(ps) return ps / Z
def __call__(self, x, confidence=False): """Predicts the target value for the given example. :arg x: An instance in dense or sparse representation. :arg confidence: whether to output confidence scores. :returns: The class assignment and optionally a confidence score. """ if x.dtype != sparsedtype: x = dense2sparse(x) p = predict(x, self.w, self.bias) return p