Lubomir Kostal

The role of stimulus parameterization in neural information and coding accuracy

Wed, 01 Apr 2026 00:00:00 +0000

Abstract:

Stimulus intensity, as a physical quantity, can be equivalently expressed in different unit systems. Researchers implicitly expect that the inferred neural coding precision and the amount of stimulus-related information are independent of such a subjective choice. We show, however, that even under regular reparameterizations these two popular methodological lines are affected in distinct ways. First, Fisher information is not invariant and may yield incompatible conclusions about coding precision when the identical stimulation scenario is evaluated in transformed coordinates. Second, while Shannon’s mutual information is invariant, its stimulus-specific decompositions are highly non-unique and may change under coordinate transformations unless constrained. We argue that requiring coordinate invariance naturally removes the ambiguity, and we discuss the psychophysical reference frame as a candidate for meaningful coding accuracy comparisons.

Neuronal information transmission: from finite-size effects to source-channel coding

Mon, 09 Jun 2025 00:00:00 +0000

Abstract:

Shannon’s information theory provides a valuable framework for analyzing neural coding and information transmission. However, Shannon limits are only achievable asymptotically, as the complexity of encoding and decoding — as well as the associated delays — grow without bound, a scenario unlikely in biological systems. In this talk, we explore the finite-size effects and decoding errors that arise in realistically constrained neural populations. We then show that it may be possible to match the statistics of the input (stimulus) and neuronal noise in such a way that uncoded transmission becomes exactly optimal in the Shannon sense. Because uncoded transmission is entirely analog, it avoids both source discretization and block coding. We thus hypothesize that it may represent a viable strategy for information transmission in real neural systems.

Instantaneous firing rate and counting statistics of spike trains

Tue, 12 Sep 2023 00:00:00 +0000

Abstract:

The rate coding hypothesis is the oldest and still one of the most accepted and investigated scenarios in neuronal activity analyses. However, the actual neuronal firing rate, while informally understood, can be mathematically defined in several different ways. These definitions yield distinct results, even their average values may differ for the simplest neuronal models. Such an inconsistency motivates us to revisit the classical concept of the instantaneous firing rate. We show that different notions of firing rate can be made compatible, at least in terms of their averages, by carefully discerning the time instant at which the neuronal activity is observed. In addition, we show that it is possible to estimate a quantity equivalent to the Fano factor and the squared coefficient of variation based on the intervals from only one specific (random) time.

Neuronal population size and reliable information transmission

Thu, 13 Apr 2023 00:00:00 +0000

Abstract:

The concept of mutual information has become indispensable in the analysis of information flow in various stochastic physical or biological systems. However, the amount of information that can actually be transmitted, and reliably decoded, is related to the mutual information only in the thermodynamic limit of infinitely many codewords, implying delays and untractable decoder complexity. We demonstrate the importance of the finite-size effects that occur in restricted systems, on an example of a layer of independent Hodgkin-Huxley type neurons (though the analysis is applicable to a more general situation as well).

On the efficiency of neuronal information processing

Thu, 28 Feb 2019 00:00:00 +0000

Abstract:

The research in computational neuroscience has a tradition of more than 100 years, marked, e.g., by the now-classical Lapicque, McCulloch-Pitts or Hodgkin-Huxley neuronal models. During the last three decades the field has experienced a dramatic increase, attracting a number of scientists from different disciplines. New topics have emerged alongside the traditional neuronal modeling approaches and the long-standing problem of neuronal coding is recently receiving substantial attention. The approach to the problem relies on the applications of information theory, signal detection and estimation theory and theory of stochastic processes to different aspects of neuronal information processing, including coding and decoding in individual neurons and populations, or analysis of beneficial role of the noise in the system. Understanding the principles of information processing in neurons may help to introduce, e.g., new algorithms or new generation of hardware which could enhance artificial sensors.

On the metabolically efficient neuronal information transmission

Fri, 06 Oct 2017 00:00:00 +0000

Abstract:

Variability and randomness in neuronal firing patterns

Mon, 18 Jul 2016 00:00:00 +0000

Abstract:

We propose and discuss two information-based measures of statistical dispersion of positive continuous random variables: the entropy-based dispersion and Fisher information-based dispersion. Although standard deviation is the most frequently employed dispersion measure, we show, that it is not well suited to quantify some aspects that are often expected intuitively, such as the degree of randomness. The proposed dispersion measures are not entirely independent, though each describes the quality of probability distribution from a different point of view. We discuss relationships between the measures, describe their extremal values and illustrate their properties on the Pareto, the lognormal and the lognormal mixture distributions. Application possibilities are also mentioned.