def _compute(self, data_1, data_2):
if self._sigma is None:
# modification of libSVM heuristics
self._sigma = float(data_1.shape[1])
dists_sq = euclidean_dist_matrix(data_1, data_2)
return np.exp(-np.sqrt(dists_sq) / self._sigma)
def _compute(self, data_1, data_2):
if self._sigma is None:
# modification of libSVM heuristics
self._sigma = float(data_1.shape[1])
dists_sq = euclidean_dist_matrix(data_1, data_2)
return np.exp(-np.sqrt(dists_sq) / self._sigma)
def _compute(self, data_1, data_2):
if self._gamma is None:
# libSVM heuristics
self._gamma = 1. / data_1.shape[1]
dists_sq = euclidean_dist_matrix(data_1, data_2)
return np.exp(-self._gamma * dists_sq)
def _compute(self, data_1, data_2):
if self._gamma is None:
# libSVM heuristics
self._gamma = 1./data_1.shape[1]
dists_sq = euclidean_dist_matrix(data_1, data_2)
return np.exp(-self._gamma * dists_sq)
def _compute(self, data_1, data_2):
dists_sq = euclidean_dist_matrix(data_1, data_2)
return 1. - (dists_sq / (dists_sq + self._c))
def _compute(self, data_1, data_2):
return -euclidean_dist_matrix(data_1, data_2)**self._d / 2.
def _compute(self, data_1, data_2):
dists = np.sqrt(euclidean_dist_matrix(data_1, data_2))
return 1 / (1 + dists**self._d)
def _compute(self, data_1, data_2):
dists_sq = euclidean_dist_matrix(data_1, data_2)
return 1. - (dists_sq / (dists_sq + self._c))
def _compute(self, data_1, data_2):
return - euclidean_dist_matrix(data_1, data_2) ** self._d / 2.
def _compute(self, data_1, data_2):
dists = np.sqrt(euclidean_dist_matrix(data_1, data_2))
return 1 / (1 + dists ** self._d)
