85afa40a630106d83907d0195911971ad3f130d8,art/classifiers/GPy.py,GPyGaussianProcessClassifier,class_gradient,#GPyGaussianProcessClassifier#Any#Any#Any#,64

Before Change

                 `(batch_size, 1, input_shape)` when `label` parameter is specified.
        :rtype: `np.ndarray`
        
        grads = np.zeros((np.shape(x)[0], 2, np.shape(x)[1]))
        for i in range(np.shape(x)[0]):
            // get gradient for the two classes GPC can maximally have
            for i_c in range(2):
                ind = self.predict(x[i].reshape(1, -1))[0, i_c]

After Change

            for i_c in range(2):
                ind = self.predict(x[i].reshape(1, -1))[0, i_c]
                sur = self.predict(np.repeat(x_preprocessed[i].reshape(1, -1),
                                             np.shape(x_preprocessed)[1], 0) + eps * np.eye(np.shape(x_preprocessed)[1]))[:, i_c]
                grads[i, i_c] = ((sur - ind) * eps).reshape(1, -1)

        grads = self._apply_preprocessing_gradient(x, grads)

Instances