Related Experiment Videos
Quantification of sensory information transmission using timeseries decorrelation techniques
Marcus Eger1, Reinhard Eckhorn
1Department of Physics, Neurophysics Group, Philipps University, Renthof 7, Marburg 35032, Germany. marcus.eger@physik.uni-marburg.de
Bio Systems
|December 3, 2002
Summary
Estimating information transmission in neuronal systems requires accounting for serial dependence. New decorrelation techniques improve transinformation estimates by representing stimuli and responses as stochastic processes.
Area of Science:
- Neuroscience
- Information Theory
- Stochastic Processes
Background:
- Estimating information transmission in neuronal systems is complicated by serial dependence in stimulus and response signals.
- Existing methods often require extensive data or are limited to Gaussian stimuli.
Purpose of the Study:
- To develop novel methods for accurately estimating transinformation in neuronal sensory systems.
- To address the overestimation of transinformation caused by independent analysis of stimulus-response pairs.
Main Methods:
- Stimulus and response signals are modeled as stochastic processes (sequences of random variables).
- Two coordinate transformation-based decorrelation techniques are introduced to handle linear serial dependence.
- These methods create representations with uncorrelated random variables.
Main Results:
- The proposed decorrelation techniques yield more precise estimates of transinformation.
- The methods effectively account for the linear component of serial dependence.
- This approach improves the accuracy of information transmission calculations.
Conclusions:
- Accurate transinformation estimation in neuronal systems necessitates addressing serial dependence.
- Decorrelation techniques based on coordinate transformation offer a more precise approach.
- These methods enhance our understanding of information processing in sensory systems.