Paradoxical noise preference in RNNs
In recurrent neural networks (RNNs) used to model biological neural networks, noise is typically introduced during training to emulate biological variability and regularize learning. The expectation is that removing the noise at test time should preserve or improve performance. Contrary to this intuition, we find that continuous-time RNNs (CTRNNs) often perform best at or near the training noise level. This noise preference typically arises when noise is injected inside the neural activation function; networks trained with noise injected outside the activation function perform best with zero noise. The phenomenon arises robustly in diverse tasks for large enough training noise; we also show the phenomenon arising in feedforward neural networks, not just in RNNs. Our analyses show that the phenomenon stems from noise-induced shifts of fixed points (stationary distributions) in the underlying stochastic dynamics of the RNNs. These fixed point shifts are noise-level dependent and bias the network outputs when the noise is removed, degrading performance. Analytical and numerical results show that the bias arises when neural states operate near activation-function nonlinearities, where noise is asymmetrically attenuated, and that performance optimization incentivizes operation near these nonlinearities; such performance incentives exist for networks with noise inside, but not outside, the activation function, explaining why only noise-in networks show the preference. Thus, networks can overfit to the training noise itself rather than just to the input-output data. The phenomenon is distinct from stochastic resonance, wherein nonzero noise enhances signal processing. Our findings reveal that training noise can become an integral part of the computation learned by neural networks, with implications for understanding neural population dynamics and for the design of robust artificial RNNs.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
The Physics of Preference: Unravelling Imprecision of Human Preferences through Magnetisation Dynamics
Paradoxical decision-making behaviours such as preference reversal often arise from imprecise or noisy human preferences. Harnessing the physical principle of magnetisation reversal in ferromagnetic nanostructures, we de…
Decision MakingVariational Adaptive Noise and Dropout towards Stable Recurrent Neural Networks
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 o…
Imitation LearningLearning TheoryVariational InferenceDo RNNs learn human-like abstract word order preferences?
RNN language models have achieved state-of-the-art results on various tasks, but what exactly they are representing about syntax is as yet unclear. Here we investigate whether RNN language models learn humanlike word ord…
Language ModelingLanguage ModellingSentenceNoise-Based Regularizers for Recurrent Neural Networks
Recurrent neural networks (RNNs) are powerful models for sequential data. They can approximate arbitrary computations, and have been used successfully in domains such as text and speech. However, the flexibility of RNNs …
Language ModelingLanguage ModellingRegularizing Recurrent Networks - On Injected Noise and Norm-based Methods
Advancements in parallel processing have lead to a surge in multilayer perceptrons' (MLP) applications and deep learning in the past decades. Recurrent Neural Networks (RNNs) give additional representational power to fee…