1385038d7f08f101a7597548196de45ca7a59679,brainiak/eventseg/event.py,EventSegment,fit,#EventSegment#Any#Any#,108

Before Change

        // Double-check that data is z-scored in time
        for i in range(n_train):
            X[i] = stats.zscore(X[i], axis=1, ddof=1)

        // Initialize variables for fitting
        log_gamma = []
        for i in range(n_train):

After Change

                logprob = self._logprob_obs(X[i], mean_pat, iteration_var)
                log_gamma[i], self.ll_[-1, i] = self._forward_backward(logprob)

            if step > 1 and self.split_merge:
                curr_ll = np.mean(self.ll_[-1, :])
                self.ll_[-1, :], log_gamma, mean_pat = \
                    self._split_merge(X, log_gamma, iteration_var, curr_ll)

            // If log-likelihood has started decreasing, undo last step and stop
            if np.mean(self.ll_[-1, :]) < best_ll:
                self.ll_ = self.ll_[:-1, :]
                break

Instances