Search

Many species are able to recognize objects, but it has been proven difficult to pinpoint and compare how different species solve this task. Recent research suggested to combine computational and animal modelling in order to obtain a more systematic understanding of task complexity and compare strategies between species. In the present study, we created a large multidimensional stimulus set and designed a visual categorization task partially based upon modelling with a convolutional deep neural network (cDNN). Experiments included rats (N = 11; 1115 daily sessions in total for all rats together) and humans (N = 50). Each species was able to master the task and generalize to a variety of new images. Nevertheless, rats and humans showed very little convergence in terms of which object pairs were associated with high and low performance, suggesting the use of different strategies. There was an interaction between species and whether stimulus pairs favoured early or late processing in a cDNN. A direct comparison with cDNN representations revealed that rat performance was best captured by late convolutional layers while human performance related more to the higher-up fully connected layers. These findings highlight the additional value of using a computational approach for the design of object recognition tasks. Overall, this computationally informed investigation of object recognition behaviour reveals a strong discrepancy in strategies between rodent and human vision.

A remarkable characteristic of overparameterized deep neural networks (DNNs) is that their accuracy does not degrade when the network's width is increased. Recent evidence suggests that developing compressible representations is key for adjusting the complexity of large networks to the learning task at hand. However, these compressible representations are poorly understood. A promising strand of research inspired from biology is understanding representations at the unit level as it offers a more granular and intuitive interpretation of the neural mechanisms. In order to better understand what facilitates increases in width without decreases in accuracy, we ask: Are there mechanisms at the unit level by which networks control their effective complexity as their width is increased? If so, how do these depend on the architecture, dataset, and training parameters? We identify two distinct types of "frivolous" units that proliferate when the network's width is increased: prunable units which can be dropped out of the network without significant change to the output and redundant units whose activities can be expressed as a linear combination of others. These units imply complexity constraints as the function the network represents could be expressed by a network without them. We also identify how the development of these units can be influenced by architecture and a number of training factors. Together, these results help to explain why the accuracy of DNNs does not degrade when width is increased and highlight the importance of frivolous units toward understanding implicit regularization in DNNs.

This repository contains the data and stimuli for https://doi.org/10.1093/cercor/bhw111

This PhD project deals with two topics set within the broader context of visual and cognitive neuroscience. First of all, we investigate whether certain properties of the primate visual system might be found in rats. The second topic deals with how representations of visual input are not fixed but are influenced by recent sensory input or even cognitive processes. In a first set of studies in rats, we investigate whether and to what extent some general principles of the primate visual system (e.g. object representations) are also present in that of the rat. This includes comparing complex stimulus representations based on action potential activity measured at different levels of the visual system. In addition, we investigate how previous sensory input affects neural responses. A robust manifestation of the influence of recent input is neural adaptation or repetition suppression, which refers to a reduced neural activity when stimuli are repeated. Additionally neural responses could also be enhanced, for example when a stimulus is rare. These mechanisms have an important role in the predictive coding theory, which postulates that cortical responses signal violations of prior expectations. For the final part of the project, we moved to macaques for experiments that could not be performed in less cognitively developed animals. Here we investigate claims that perceptual expectations can modulate the strength of adaptation of action potential activity in visual cortex. status: published