Function Identification in Neuron Populations via Information Bottleneck

Abstract

It is plausible to hypothesize that the spiking responses of certain neurons represent functions of the spiking signals of other neurons. A natural ensuing question concerns how to use experimental data to infer what kind of a function is being computed. Model-based approaches typically require assumptions on how information is represented. By contrast, information measures are sensitive only to relative behavior: information is unchanged by applying arbitrary invertible transformations to the involved random variables. This paper develops an approach based on the information bottleneck method that attempts to find such functional relationships in a neuron population. Specifically, the information bottleneck method is used to provide appropriate compact representations which can then be parsed to infer functional relationships. In the present paper, the parsing step is specialized to the case of remapped-linear functions. The approach is validated on artificial data and then applied to recordings from the motor cortex of a macaque monkey performing an arm-reaching task. Functional relationships are identified and shown to exhibit some degree of persistence across multiple trials of the same experiment.

A Discrete Detection and Decoding of MLC NAND Flash Memory With Retention Noise

“…The LDPC code used in simulation is a (8000,7200) regular code with d c = 51 and d v = 5, and generated by PEG [37]. We assume that a decoding process is performed in a block with one upper page and one lower page, and LLR are calculated as (23) (24). The maximum number of decoding iterations is set to 50.…”

“…After that, several IB algorithmic approaches were proposed, such as agglomerative IB (Agg-IB), sequential IB (Seq-IB), deterministic IB (Det-IB) and Kullback-Leibler-means IB (KL-means-IB) [22]. Since its invention, IB has widely applied in different fields, such as documents classification [23], neuroscience [24], deep learning [25] and LDPC decoding [18].…”

Genetic Algorithms to Maximize the Relevant Mutual Information in Communication Receivers

“…This rule is designed to preserve so-called relevant mutual information I(X; T) ≤ I(X; Y), where X is a properly chosen relevant random variable of interest. The method is very generic and has numerous applications, for example, in image and speech processing, in astronomy and in neuroscience [2][3][4].…”

A novel tensor-information bottleneck method for multi-input single-output applications

“…Information Bottleneck (IB) method for data compression was introduced by Tishby et al in [21], where the key idea is to compress the observation while the output preserves most of the information of the relevant variables, i.e., the original source. Since then, various IB methods have been rapidly developed and utilized in many fields, such as neuroscience, image processing, and deep learning [22,23,24,25]. The partitioning principle of it is usually divided into soft [21] and hard [26] partitions of original source.…”

“…The partitioning principle of it is usually divided into soft [21] and hard [26] partitions of original source. Kartik et al [22] developed an approach based on IB that attempts to find functional relationships in a neuron population. New image segmentation algorithms based on the hard version of the information bottleneck theory are presented in [23].…”