Linear and Non-Linear Visual Feature Learning in Rat and Humans.

Christophe Bossens,Hans P. Op de Beeck

doi:10.3389/fnbeh.2016.00235

Abstract

The visual system processes visual input in a hierarchical manner in order to extract relevant features that can be used in tasks such as invariant object recognition. Although typically investigated in primates, recent work has shown that rats can be trained in a variety of visual object and shape recognition tasks. These studies did not pinpoint the complexity of the features used by these animals. Many tasks might be solved by using a combination of relatively simple features which tend to be correlated. Alternatively, rats might extract complex features or feature combinations which are nonlinear with respect to those simple features. In the present study, we address this question by starting from a small stimulus set for which one stimulus-response mapping involves a simple linear feature to solve the task while another mapping needs a well-defined nonlinear combination of simpler features related to shape symmetry. We verified computationally that the nonlinear task cannot be trivially solved by a simple V1-model. We show how rats are able to solve the linear feature task but are unable to acquire the nonlinear feature. In contrast, humans are able to use the nonlinear feature and are even faster in uncovering this solution as compared to the linear feature. The implications for the computational capabilities of the rat visual system are discussed.

Highlights

Starting with the discovery of simple cells by Hubel and Wiesel (1959), decades of neurophysiological research have revealed the coding of a multitude of visual features in the mammalian visual system (Grill-Spector and Malach, 2004) The extraction of these features seems to follow a general principle where simple visual features are coded at the beginning of the visual information processing pathway and tend to be very sensitive to viewing conditions (Rust and Dicarlo, 2010)
Knowing that a specific feature is detected by the visual system does not necessarily mean that this feature will be used in an object recognition task, even if the object contains that feature
After training animals to distinguish between squares and triangles, Visual Feature Learning we demonstrated how rats are capable of applying a flexible recognition template that is invariant to position and size

Summary

Introduction

Starting with the discovery of simple cells by Hubel and Wiesel (1959), decades of neurophysiological research have revealed the coding of a multitude of visual features in the mammalian visual system (Grill-Spector and Malach, 2004) The extraction of these features seems to follow a general principle where simple visual features are coded at the beginning of the visual information processing pathway and tend to be very sensitive to viewing conditions (Rust and Dicarlo, 2010). More complex features appear more upstream and tend to be more robust to different viewing conditions. These complex features are necessary for the visual system to perform tasks such as object recognition (Palmeri and Gauthier, 2004). The main mechanism behind this technique is to cover the stimulus with a mask so that only parts of it are visible. By allowing the location of the mask to vary from trial to trial and keeping track of the behavioral performance as a function of visible locations, we can determine which object regions are critical in making correct or incorrect identifications

Methods

Results

Conclusion