Computational modelling of salamander retinal ganglion cells using machine learning approaches

Abstract

Artificial vision using computational models that can mimic biological vision is an area of ongoing research. One of the main themes within this research is the study of the retina and in particular, retinal ganglion cells which are responsible for encoding the visual stimuli. A common approach to modelling the internal processes of retinal ganglion cells is the use of a linear-non-linear cascade model, which models the cell's response using a linear filter followed by a static non-linearity. However, the resulting model is generally restrictive as it is often a poor estimator of the neuron's response. In this paper we present an alternative to the linear-non-linear model by modelling retinal ganglion cells using a number of machine learning techniques which have a proven track record for learning complex non-linearities in many different domains. A comparison of the model predicted spike rate shows that the machine learning models perform better than the standard linear-non-linear approach in the case of temporal white noise stimuli.

Deep Auto-encoder with Neural Response

“…Nonlinear methods provide a more general way to predict neural responses, for instance, energy models [14], the linear-nonlinear model (LN) and its extended version LN-LN [25]. Traditional machine learning methods, such as generalized linear models (GLMs) [40], the multi-layer percep- tron (MLP) and the support vector regression (SVRs) [5], have been brought into the computational neuroscience field to increase the neural similarity. More recently, the hierarchical structure has been found strikingly similar in ventral visual pathway [6,35,29] and in deep convolutional neural networks [10,20,19].…”

“…The loss function of DAE-NR explicitly considers both the image reconstruction task and the neural representation similarity task, as defined in Eq. (5). (5) where α and β are the hyperparameters to trade-off the image reconstruction task and the neural representation similarity task.…”

“…(5). (5) where α and β are the hyperparameters to trade-off the image reconstruction task and the neural representation similarity task. Intuitively, the larger α favors the reconstruction task, while the larger β biases toward the the neural representation similarity task.…”

Abstract and Image Analysis of High-Temperature Materials from Scientific Journals Using Deep Learning and Rule-Based Machine Learning Approaches

“…A concept-based extraction is introduced to improve text classification performance using a twolayer concept with better performance than the traditional one [4]. For reproducing biological vision in Artificial vision using machine learning computational models [5]. A NEXTWRAP project implemented low-level page segmentation techniques to extract and wrap pdf documents [6].…”

Simultaneous Neural Spike Encoding and Decoding Based on Cross-modal Dual Deep Generative Model

“…Till now, there have been lots of researches on RGC spike encoding. The existing methods contain the linear nonlinear model (LN) and its cascaded version LN-LN, the generalized linear model (GLM) taking spike history as feedback [10] and kinds of machine learning techniques [11]. However, the above methods only fit well on stimuli with simple artificial stimuli and are easy to overfitting with natural scenes which have more complicated distribution.…”