A framework for Bayesian optimality of psychophysical laws

By John Z. Sun, Grace I. Wang, Vivek K Goyal, L. Varshney. 2012

73 citations recorded by the source corpus.

Abstract The Weber–Fechner law states that perceived intensity is proportional to physical stimuli on a logarithmic scale. In this work, we formulate a Bayesian framework for the scaling of perception and find logarithmic and related scalings are optimal under expected relative error fidelity. Therefore, the Weber–Fechner law arises as being information theoretically efficient under the constraint of limited representability. An even stronger connection is drawn between the Weber–Fechner law and a Bayesian framework when neural storage or communication is the dominant concern, such as for numerosity. Theoretical results and experimental verification for perception of sound intensity are both presented.

Subfields: Cognitive Neuroscience

Citations in the corpus

• Natural Compression for Distributed Deep Learning (2019, neutral): “Outside of the optimization for machine learning, compression operators are very relevant to optimal quantization theory and control theory (Elia and Mitter, 2001; Sun and Goyal, 2011; Sun et al., 2012).”
• A Survey of Quantization Methods for Efficient Neural Network Inference (2021, neutral): “We refer the reader to [220] for a thorough review of related work in neuroscience literature.”