4251a86130e46e9d8a5e52ffc1c7f86bbd7f3f8a,tensorforce/core/preprocessing/running_standardize.py,RunningStandardize,tf_process,#RunningStandardize#Any#,48

Before Change

            trainable=False
        )

        with tf.control_dependencies([tf.assign_add(ref=count, value=1)]):
            // TODO Do we want to allow an axis or always standardize along time
            update_mean = tf.reduce_sum(input_tensor=(state - mean_estimate), axis=self.axis)

            // Update implements: https://www.johndcook.com/blog/standard_deviation/
            // TODO check batch shapes?
            float_count = tf.cast(x=count, dtype=tf.float32)
            mean_estimate = tf.cond(
                pred=(count > 1),
                true_fn=(lambda: mean_estimate + update_mean / float_count),
                false_fn=(lambda: mean_estimate)
            )
            update_variance_sum = (state - mean_estimate) * (state - update_mean)
            variance_sum_estimate = variance_sum_estimate + \
                tf.reduce_sum(input_tensor=update_variance_sum, axis=self.axis)

            variance_estimate = tf.cond(
                pred=(count > 1),
                true_fn=(lambda: variance_sum_estimate / (float_count - 1.0)),
                false_fn=(lambda: variance_estimate)
            )
        // print("mean estimate shape = {}".format(tf.shape(mean_estimate)))

        return tf.cond(
            pred=(count > 1),
            true_fn=(lambda: (state - mean_estimate) / (tf.maximum(x=tf.sqrt(x=variance_estimate), y=util.epsilon))),
            false_fn=(lambda: state)

After Change

            trainable=False
        )

        assignment = tf.assign_add(ref=count, value=1.0)

        with tf.control_dependencies(control_inputs=(assignment,)):
            // Mean update
            mean = tf.reduce_sum(input_tensor=(tensor - mean_estimate), axis=0)
            assignment = tf.assign_add(ref=mean_estimate, value=(mean / count))

        with tf.control_dependencies(control_inputs=(assignment,)):

            def first_run():
                // No meaningful mean and variance yet

Instances