Open Access

Orientation hypercolumns in the visual cortex are delimited by the repeating pinwheel patterns of orientation selective neurons. We design a generative model for visual cortex maps that reproduces such orientation hypercolumns as well as ocular dominance maps while preserving retinotopy. The model uses a neural placement method based on t–distributed stochastic neighbour embedding (t–SNE) to create maps that order common features in the connectivity matrix of the circuit. We find that, in our model, hypercolumns generally appear with fixed cell numbers independently of the overall network size. These results would suggest that existing differences in absolute pinwheel densities are a consequence of variations in neuronal density. Indeed, available measurements in the visual cortex indicate that pinwheels consist of a constant number of ∼30, 000 neurons. Our model is able to reproduce a large number of characteristic properties known for visual cortex maps. We provide the corresponding software in our MAPStoolbox for Matlab.

The connectivity of cortical microcircuits is a major determinant of brain function; defining how activity propagates between different cell types is key to scaling our understanding of individual neuronal behaviour to encompass functional networks. Furthermore, the integration of synaptic currents within a dendrite depends on the spatial organisation of inputs, both excitatory and inhibitory. We identify a simple equation to estimate the number of potential anatomical contacts between neurons; finding a linear increase in potential connectivity with cable length and maximum spine length, and a decrease with overlapping volume. This enables us to predict the mean number of candidate synapses for reconstructed cells, including those realistically arranged. We identify an excess of putative connections in cortical data, with densities of neurite higher than is necessary to reliably ensure the possible implementation of any given connection. We show that potential contacts allow the particular implementation of connectivity at a subcellular level.

Abstract Trial-to-trial variability and spontaneous activity of cortical recordings have been suggested to reflect intrinsic noise. This view is currently challenged by mounting evidence for structure in these phenomena: Trial-to-trial variability decreases following stimulus onset and can be predicted by previous spontaneous activity. This spontaneous activity is similar in magnitude and structure to evoked activity and can predict decisions. Allof the observed neuronal properties described above can be accounted for, at an abstract computational level, by the sampling-hypothesis, according to which response variability reflects stimulus uncertainty. However, a mechanistic explanation at the level of neural circuit dynamics is still missing. In this study, we demonstrate that all of these phenomena can be accounted for by a noise-free self-organizing recurrent neural network model (SORN). It combines spike-timing dependent plasticity (STDP) and homeostatic mechanisms in a deterministic network of excitatory and inhibitory McCulloch-Pitts neurons. The network self-organizes to spatio-temporally varying input sequences. We find that the key properties of neural variability mentioned above develop in this model as the network learns to perform sampling-like inference. Importantly, the model shows high trial-to-trial variability although it is fully deterministic. This suggests that the trial-to-trial variability in neural recordings may not reflect intrinsic noise. Rather, it may reflect a deterministic approximation of sampling-like learning and inference. The simplicity of the model suggests that these correlates of the sampling theory are canonical properties of recurrent networks that learn with a combination of STDP and homeostatic plasticity mechanisms. Author Summary Neural recordings seem very noisy. If the exact same stimulus is shown to an animal multiple times, the neural response will vary. In fact, the activity of a single neuron shows many features of a stochastic process. Furthermore, in the absence of a sensory stimulus, cortical spontaneous activity has a magnitude comparable to the activity observed during stimulus presentation. These findings have led to a widespread belief that neural activity is indeed very noisy. However, recent evidence indicates that individual neurons can operate very reliably and that the spontaneous activity in the brain is highly structured, suggesting that much of the noise may in fact be signal. One hypothesis regarding this putative signal is that it reflects a form of probabilistic inference through sampling. Here we show that the key features of neural variability can be accounted for in a completely deterministic network model through self-organization. As the network learns a model of its sensory inputs, the deterministic dynamics give rise to sampling-like inference. Our findings show that the notorious variability in neural recordings does not need to be seen as evidence for a noisy brain. Instead it may reflect sampling-like inference emerging from a self-organized learning process.

The development of vision during the first months of life is an active process that comprises the learning of appropriate neural representations and the learning of accurate eye movements. While it has long been suspected that the two learning processes are coupled, there is still no widely accepted theoretical framework describing this joint development. Here we propose a computational model of the development of active binocular vision to fill this gap. The model is based on a new formulation of the Active Efficient Coding theory, which proposes that eye movements, as well as stimulus encoding, are jointly adapted to maximize the overall coding efficiency. Under healthy conditions, the model self-calibrates to perform accurate vergence and accommodation eye movements. It exploits disparity cues to deduce the direction of defocus, which leads to co-ordinated vergence and accommodation responses. In a simulated anisometropic case, where the refraction power of the two eyes differs, an amblyopia-like state develops, in which the foveal region of one eye is suppressed due to inputs from the other eye. After correcting for refractive errors, the model can only reach healthy performance levels if receptive fields are still plastic, in line with findings on a critical period for binocular vision development. Overall, our model offers a unifying conceptual framework for understanding the development of binocular vision.

In the human brain, the incoming light to the retina is transformed into meaningful representations that allow us to interact with the world. In a similar vein, the RGB pixel values are transformed by a deep neural network (DNN) into meaningful representations relevant to solving a computer vision task it was trained for. Therefore, in my research, I aim to reveal insights into the visual representations in the human visual cortex and DNNs solving vision tasks. In the previous decade, DNNs have emerged as the state-of-the-art models for predicting neural responses in the human and monkey visual cortex. Research has shown that training on a task related to a brain region’s function leads to better predictivity than a randomly initialized network. Based on this observation, we proposed that we can use DNNs trained on different computer vision tasks to identify functional mapping of the human visual cortex. To validate our proposed idea, we first investigate a brain region occipital place area (OPA) using DNNs trained on scene parsing task and scene classification task. From the previous investigations about OPA’s functions, we knew that it encodes navigational affordances that require spatial information about the scene. Therefore, we hypothesized that OPA’s representation should be closer to a scene parsing model than a scene classification model as the scene parsing task explicitly requires spatial information about the scene. Our results showed that scene parsing models had representation closer to OPA than scene classification models thus validating our approach. We then selected multiple DNNs performing a wide range of computer vision tasks ranging from low-level tasks such as edge detection, 3D tasks such as surface normals, and semantic tasks such as semantic segmentation. We compared the representations of these DNNs with all the regions in the visual cortex, thus revealing the functional representations of different regions of the visual cortex. Our results highly converged with previous investigations of these brain regions validating the feasibility of the proposed approach in finding functional representations of the human brain. Our results also provided new insights into underinvestigated brain regions that can serve as starting hypotheses and promote further investigation into those brain regions. We applied the same approach to find representational insights about the DNNs. A DNN usually consists of multiple layers with each layer performing a computation leading to the final layer that performs prediction for a given task. Training on different tasks could lead to very different representations. Therefore, we first investigate at which stage does the representation in DNNs trained on different tasks starts to differ. We further investigate if the DNNs trained on similar tasks lead to similar representations and on dissimilar tasks lead to more dissimilar representations. We selected the same set of DNNs used in the previous work that were trained on the Taskonomy dataset on a diverse range of 2D, 3D and semantic tasks. Then, given a DNN trained on a particular task, we compared the representation of multiple layers to corresponding layers in other DNNs. From this analysis, we aimed to reveal where in the network architecture task-specific representation is prominent. We found that task specificity increases as we go deeper into the DNN architecture and similar tasks start to cluster in groups. We found that the grouping we found using representational similarity was highly correlated with grouping based on transfer learning thus creating an interesting application of the approach to model selection in transfer learning. During previous works, several new measures were introduced to compare DNN representations. So, we identified the commonalities in different measures and unified different measures into a single framework referred to as duality diagram similarity. This work opens up new possibilities for similarity measures to understand DNN representations. While demonstrating a much higher correlation with transfer learning than previous state-of-the-art measures we extend it to understanding layer-wise representations of models trained on the Imagenet and Places dataset using different tasks and demonstrate its applicability to layer selection for transfer learning. In all the previous works, we used the task-specific DNN representations to understand the representations in the human visual cortex and other DNNs. We were able to interpret our findings in terms of computer vision tasks such as edge detection, semantic segmentation, depth estimation, etc. however we were not able to map the representations to human interpretable concepts. Therefore in our most recent work, we developed a new method that associates individual artificial neurons with human interpretable concepts. Overall, the works in this thesis revealed new insights into the representation of the visual cortex and DNNs...