6e5cf676b113af8d70e34224f23bc8eace618856,mloop/nnlearner.py,SingleNeuralNet,fit,#SingleNeuralNet#Any#Any#,120

Before Change

                                               })

        self.log.debug("Fit neural network with total training cost "
                + str(self.tf_session.run(
                        self.loss_func,
                        feed_dict={self.input_placeholder: params,
                                   self.output_placeholder: [[c] for c in costs],
                                   self.regularisation_coefficient_placeholder: self.regularisation_coefficient,
                                   }))
                + ", with unregularized cost "
                + str(self.tf_session.run(
                        self.loss_func,
                        feed_dict={self.input_placeholder: params,
                                   self.output_placeholder: [[c] for c in costs],
                                   self.regularisation_coefficient_placeholder: 0,

After Change

        // - if the new loss is greater than the threshold then we haven"t improved much, so stop
        // - else start from the top
        while True:
            threshold = 0.8 * self._loss(params, costs)[0]
            for i in range(self.train_epochs):
                // Split the data into random batches, and train on each batch
                indices = np.random.permutation(len(params))
                for j in range(math.ceil(len(params) / self.batch_size)):

Instances