Encoding and Decoding Neural Population Signals for Two-Dimensional Stimulus

Stimulus is encoded by neuronal populations and then the brain can decide what happens in the practical situations from the population patterns of neural spiking. It is meaningful to implement computations by making use of the response of population of neurons to determine a certain stimulus or to obtain the values of related parameters. Neural populations can not only encode a single attribute of a stimulus but also can encode multi-dimensional stimulus (or various properties of a stimulus) simultaneously. For example, by looking at a moving object we can estimate the direction and speed of it, which shows that the response of a population of optic neurons contain information about both the direction and speed. However, we do not find any models in literature for encoding and decoding neuronal population signals for multi-dimensional stimulus. In this paper we present a simple model for reading neural population signals for the two attributes of a stimulus (or two-dimensional stimulus). We use Poisson distribution to describe the encoding process of neural populations and then to extract the values of the stimulus by Bayesian methods. We demonstrate that the nervous population can encode two properties of a stimulus and extract the two kinds of estimated values of the stimulus at the same time. The results show that we can obtain perfect estimates of two attributes of a stimulus from our simple model. Finally, we use Fisher information matrix to examine the influence of the tuning widths on the encoding efficiency.