5dd5b8652f28426322309c4e983c94a3caa8fa8e,libact/query_strategies/variance_reduction.py,VarianceReduction,A,#VarianceReduction#,19

Before Change

        grad = pi[c] *  np.ones((feature_count*label_count))
        for i in range(feature_count):
            for j in range(label_count):
                if c == j:
                    grad[i*label_count+j] *= (1-pi[c]) * x[i]
                    //grad.append(pi[c] * (1-pi[c]) * x[i])
                else:
                    grad[i*label_count+j] *= (-pi[j]) * x[i]
                    //grad.append(-pi[c] * pi[j] * x[i])
        //grad = np.array(grad)
        //print(np.shape(grad))

        return np.dot(grad.T, grad)

    def Fisher(self, pi, x, label_count, feature_count):
        fisher = np.ones((label_count*feature_count, label_count*feature_count))

After Change

        print(s-time.time())
        print(np.shape(grad))
        
        _pi = -1 * np.array(pi)
        _pi[c] += 1
        grad = pi[c] * np.tile(np.array([x]).T, (1, label_count)) *\
                        np.tile(np.array([_pi]), (feature_count, 1))
        grad = grad.reshape((feature_count*label_count))

        return np.dot(grad.T, grad)

    def Fisher(self, pi, x, label_count, feature_count):

Instances