LGROJun 2, 2025

Variational Adaptive Noise and Dropout towards Stable Recurrent Neural Networks

arXiv:2506.01350v1h-index: 1ICDL
Originality Incremental advance
AI Analysis

This addresses stability issues in RNNs for robotics applications, representing an incremental improvement by combining and adapting existing stabilization techniques.

The paper tackles the problem of stabilizing recurrent neural networks (RNNs) by proposing variational adaptive noise and dropout (VAND), which reinterprets RNN optimization as variational inference to derive noise and dropout simultaneously, resulting in successful imitation of sequential and periodic behaviors in a mobile manipulator scenario.

This paper proposes a novel stable learning theory for recurrent neural networks (RNNs), so-called variational adaptive noise and dropout (VAND). As stabilizing factors for RNNs, noise and dropout on the internal state of RNNs have been separately confirmed in previous studies. We reinterpret the optimization problem of RNNs as variational inference, showing that noise and dropout can be derived simultaneously by transforming the explicit regularization term arising in the optimization problem into implicit regularization. Their scale and ratio can also be adjusted appropriately to optimize the main objective of RNNs, respectively. In an imitation learning scenario with a mobile manipulator, only VAND is able to imitate sequential and periodic behaviors as instructed. https://youtu.be/UOho3Xr6A2w

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes