Showing 1–1 of 1 results for author: Krishnamurthy, K
-
Emergence of robust memory manifolds
Authors:
Tankut Can,
Kamesh Krishnamurthy
Abstract:
The ability to store continuous variables in the state of a biological system (e.g. a neural network) is critical for many behaviours. Most models for implementing such a memory manifold require hand-crafted symmetries in the interactions or precise fine-tuning of parameters. We present a general principle that we refer to as {\it frozen stabilisation} (FS), which allows a family of neural network…
▽ More
The ability to store continuous variables in the state of a biological system (e.g. a neural network) is critical for many behaviours. Most models for implementing such a memory manifold require hand-crafted symmetries in the interactions or precise fine-tuning of parameters. We present a general principle that we refer to as {\it frozen stabilisation} (FS), which allows a family of neural networks to self-organise to a critical state exhibiting multiple memory manifolds without parameter fine-tuning or symmetries. Memory manifolds arising from FS exhibit a wide range of emergent relaxational timescales and can be used as general purpose integrators for inputs aligned with the manifold. Moreover, FS allows robust memory manifolds in small networks, and this is relevant to debates of implementing continuous attractors with a small number of neurons in light of recent experimental discoveries.
△ Less
Submitted 5 September, 2024; v1 submitted 8 September, 2021;
originally announced September 2021.
