Network outputs can change indirectly to unexpected values after any random batch update for input data not included in the batch, called churn in this paper.

The Wasserstein distance (WD) [27] is a suitable metric for this purpose due to possessing nonvanishing gradients.

Deep learning is booming driven by massive labeled data over the past few years, such as image classification[1,2],semanticsegmentation[3,4],objectdetection[5,6],naturallanguageprocessing [7,8].

Gradient Descent (GD) is a simple yet efficient baseline algorithm for solving nonconvex optimization problems.

We propose a fast algorithm for the probabilistic solution of boundary value problems (BVPs), which areordinary differential equations subject toboundary conditions.