Slow Feature Analysis on Retinal Waves Leads to V1 Complex Cells

The developing visual system of many mammalian species is partially structured and organized even before the onset of vision. Spontaneous neural activity, which spreads in waves across the retina, has been suggested to play a major role in these prenatal structuring processes. Recently, it has been shown that when employing an efficient coding strategy, such as sparse coding, these retinal activity patterns lead to basis functions that resemble optimal stimuli of simple cells in primary visual cortex (V1). Here we present the results of applying a coding strategy that optimizes for temporal slowness, namely Slow Feature Analysis (SFA), to a biologically plausible model of retinal waves. Previously, SFA has been successfully applied to model parts of the visual system, most notably in reproducing a rich set of complex-cell features by training SFA with quasi-natural image sequences. In the present work, we obtain SFA units that share a number of properties with cortical complex-cells by training on simulated retinal waves. The emergence of two distinct properties of the SFA units (phase invariance and orientation tuning) is thoroughly investigated via control experiments and mathematical analysis of the input-output functions found by SFA. The results support the idea that retinal waves share relevant temporal and spatial properties with natural visual input. Hence, retinal waves seem suitable training stimuli to learn invariances and thereby shape the developing early visual system such that it is best prepared for coding input from the natural world.     

Acquisition of color opponent representation by a three-layered neural network model

This paper discusses color representation in the visual system by analysis of a three-layered neural network model. The model incorporates physiological knowledge of color representation at the sensor level (broad-band trichromatic representation by cones) and the higher level (narrow-band color representation by color-coded cells in V4). We trained the model to perform a mapping between these color representations by the back propagation algorithm and analyzed the acquired characteristics of the hidden units. It turned out that the hidden units learned characteristics similar to those of the color opponent cells found in the visual system. It was concluded that the R-G and Y-B color opponent representations reflect the efficiency of the color representation in the visual system from investigations on the efficiency of color representation in the hidden layer and on the capability of the color recognition task of the model.     

Multi-scale lines and edges in V1 and beyond: Brightness,object categorization and recognition,and consciousness

In this paper we present an improved model for line and edge detection in cortical area V1. This model is based on responses of simple and complex cells, and it is multi-scale with no free parameters. We illustrate the use of the multi-scale line/edge representation in different processes: visual reconstruction or brightness perception, automatic scale selection and object segregation. A two-level object categorization scenario is tested in which pre-categorization is based on coarse scales only and final categorization on coarse plus fine scales. We also present a multi-scale object and face recognition model. Processing schemes are discussed in the framework of a complete cortical architecture. The fact that brightness perception and object recognition may be based on the same symbolic image representation is an indication that the entire (visual) cortex is involved in consciousness.     

Changes in Expression of Fibroblast Surface Antigen (SFA) during Cytodifferentiation and Heterokaryon Formation

Fibroblast surface antigen (SF antigen, SFA) is a major glycoprotein antigen detected in connective tissue cells (primitive mesenchymal cells, fibroblasts, and astroglial cells). In this study the expression of SFA was followed during differentiation of the mesenchymal cells of the mouse metanephros and during heterokaryon formation produced by Sendai-virus induced fusion of human fibroblasts and chick red blood cells. It was demonstrated by immunofluorescence that SFA was lost from the kidney mesenchymal cells when they differentiate into epithelial cells of the secretory tubuli. During this process SFA became detectable in the basement membrane formed around the tubuli. In cell fusion experiments human SFA which was present as fibrillar network on the surface of cultured fibroblasts, was gradually lost from the heterokaryons when the incorporated chick nuclei became activated. These two sets of experiments indicate that SFA can be used as a phenotypic marker of Cytodifferentiation.     

Cloning, sequencing and characterization of the biosynthetic gene cluster of sanglifehrin A, a potent cyclophilin inhibitor

Sanglifehrin A (SFA), a potent cyclophilin inhibitor produced by Streptomyces flaveolus DSM 9954, bears a unique [5.5] spirolactam moiety conjugated with a 22-membered, highly functionalized macrolide through a linear carbon chain. SFA displays a diverse range of biological activities and offers significant therapeutic potential. However, the structural complexity of SFA poses a tremendous challenge for new analogue development via chemical synthesis. Based on a rational prediction of its biosynthetic origin, herein we report the cloning, sequencing and characterization of the gene cluster responsible for SFA biosynthesis. Analysis of the 92 776 bp contiguous DNA region reveals a mixed polyketide synthase (PKS)/non-ribosomal peptide synthetase (NRPS) pathway which includes a variety of unique features for unusual PKS and NRPS building block formation. Our findings suggest that SFA biosynthesis requires a crotonyl-CoA reductase/carboxylase (CCR) for generation of the putative unusual PKS starter unit (2R)-2-ethylmalonamyl-CoA, an iterative type I PKS for the putative atypical extender unit (2S)-2-(2-oxo-butyl)malonyl-CoA and a phenylalanine hydroxylase for the NRPS extender unit (2S)-m-tyrosine. A spontaneous ketalization of significant note, may trigger spirolactam formation in a stereo-selective manner. This study provides a framework for the application of combinatorial biosynthesis methods in order to expand the structural diversity of SFA.     

Glial cell development and function in the Drosophila visual system

In the developing nervous system, building a functional neuronal network relies on coordinating the formation, specification and survival to diverse neuronal and glial cell subtypes. The establishment of neuronal connections further depends on sequential neuron-neuron and neuron-glia interactions that regulate cell-migration patterns and axon guidance. The visual system of Drosophila has a highly regular, retinotopic organization into reiterated interconnected synaptic circuits. It is therefore an excellent invertebrate model to investigate basic cellular strategies and molecular determinants regulating the different developmental processes that lead to network formation. Studies in the visual system have provided important insights into the mechanisms by which photoreceptor axons connect with their synaptic partners within the optic lobe. In this review, we highlight that this system is also well suited for uncovering general principles that underlie glial cell biology. We describe the glial cell subtypes in the visual system and discuss recent findings about their development and migration. Finally, we outline the pivotal roles of glial cells in mediating neural circuit assembly, boundary formation, neural proliferation and survival, as well as synaptic function.     

Perception, representation and recognition: a holistic view of recognition

It is clear that humans have mental representations of their spatial environments and that these representations are useful, if not essential, in a wide variety of cognitive tasks such as identification of landmarks and objects, guiding actions and navigation and in directing spatial awareness and attention. Determining the properties of mental representation has long been a contentious issue (see Pinker, 1984). One method of probing the nature of human representation is by studying the extent to which representation can surpass or go beyond the visual (or sensory) experience from which it derives. From a strictly empiricist standpoint what is not sensed cannot be represented; except as a combination of things that have been experienced. But perceptual experience is always limited by our view of the world and the properties of our visual system. It is therefore not surprising when human representation is found to be highly dependent on the initial viewpoint of the observer and on any shortcomings thereof. However, representation is not a static entity; it evolves with experience. The debate as to whether human representation of objects is view-dependent or view-invariant that has dominated research journals recently may simply be a discussion concerning how much information is available in the retinal image during experimental tests and whether this information is sufficient for the task at hand. Here we review an approach to the study of the development of human spatial representation under realistic problem solving scenarios. This is facilitated by the use of realistic virtual environments, exploratory learning and redundancy in visual detail.     

The novel cyclophilin-binding drug sanglifehrin A specifically affects antigen uptake receptor expression and endocytic capacity of human dendritic cells

Sanglifehrin A (SFA) is a recently developed immunosuppressant that belongs to the family of immunophilin-binding ligands. SFA is a cyclophilin A-binding immunosuppressive drug with a novel, but unidentified, mechanism of action. Several reports exist about the effect of SFA on T cells, but its effect on the initiators of the immune response, i.e., dendritic cells (DCs), is relatively unknown. Therefore, we examined the effect of SFA on the differentiation and function of human monocyte-derived DCs. Unlike the well-known cyclophilin A-binding immunosuppressant cyclosporin A, which did not affect DC phenotype, differentiation of DCs in the presence of SFA resulted in CD14-CD1a DCs with normal DC morphology, viability, and a proper capacity to activate allogeneic T cells. However, DCs generated in the presence of SFA demonstrated reduced macropinocytosis and lectin-mediated endocytosis, which was in line with a decreased expression of C-type lectins, including mannose receptor, C1qRP, DC-ASGPR, and especially, DC-SIGN. In contrast, FcalphaRI (CD89) and FcgammaRII (CD32) were increased by SFA. The explicit effect of SFA on the expression of Ag uptake receptors and Ag capture by DCs makes SFA unique among immunophilin-binding immunosuppressive drugs.     

Differences in cell morphology, lipid and apo B secretory capacity in caco-2 cells following long term treatment with saturated and monounsaturated fatty acids

The suitability of the caco-2 cell line as a model for studying the long term impact of dietary fatty acids on intestinal lipid handling and chylomicron production was examined. Chronic supplementation of caco-2 cells with palmitic acid (PA) resulted in a lower triacylglycerol secretion than oleic acid (OA). This was coupled with a detrimental effect of PA, but not OA, on transepithelial electrical resistance (TER) measurements, suggesting a loss of structural integrity across the cell monolayer. Addition of OA reversed the adverse effects of PA and stearic acid on TER and increased the ability of cells to synthesise and accumulate lipid, but did not normalise the secretion of lipids by caco-2 cells. Increasing amounts of OA and decreasing amounts of PA in the incubation media markedly improved the ability of cells to synthesise apolipoprotein B and secrete lipids. Real time RT-PCR revealed a down regulation of genes involved in lipoprotein synthesis following PA than OA. Electron microscopy showed adverse effects of PA on cellular morphology consistent with immature enterocytes such as stunted microvilli and poor tight junction formation. In conclusion, previously reported differences in lipoprotein secretion by caco-2 cells supplemented with saturated fatty acids (SFA) and OA may partly reflect early cytotoxic effects of SFA on cellular integrity and function.     

Motor-maps, navigation and implicit space representation in the hippocampus

Multiple sensory-motor maps located in the brainstem and the cortex are involved in spatial orientation. Guiding movements of eyes, head, neck and arms they provide an approximately linear relation between target distance and motor response. This involves especially the superior colliculus in the brainstem and the parietal cortex. There, the natural frame of reference follows from the retinal representation of the environment. A model of navigation is presented that is based on the modulation of activity in those sensory-motor maps. The actual mechanism chosen was gain-field modulation, a process of multimodal integration that has been demonstrated in the parietal cortex and superior colliculus, and was implemented as attraction to visual cues (colour). Dependent on the metric of the sensory-motor map, the relative attraction to these cues implemented as gain field modulation and their position define a fixed point attractor on the plane for locomotive behaviour. The actual implementation used Kohonen-networks in a variant of reinforcement learning that are well suited to generate such topographically organized sensory-motor maps with roughly linear visuo-motor response characteristics. In the following, it was investigated how such an implicit coding of target positions by gain-field parameters might be represented in the hippocampus formation and under what conditions a direction-invariant space representation can arise from such retinotopic representations of multiple cues. Information about the orientation in the plane—as could be provided by head direction cells—appeared to be necessary for unambiguous space representation in our model in agreement with physiological experiments. With this information, Gauss-shaped “place-cells” could be generated, however, the representation of the spatial environment was repetitive and clustered and single cells were always tuned to the gain-field parameters as well     

Continuous Attractor Network Model for Conjunctive Position-by-Velocity Tuning of Grid Cells

The spatial responses of many of the cells recorded in layer II of rodent medial entorhinal cortex (MEC) show a triangular grid pattern, which appears to provide an accurate population code for animal spatial position. In layer III, V and VI of the rat MEC, grid cells are also selective to head-direction and are modulated by the speed of the animal. Several putative mechanisms of grid-like maps were proposed, including attractor network dynamics, interactions with theta oscillations or single-unit mechanisms such as firing rate adaptation. In this paper, we present a new attractor network model that accounts for the conjunctive position-by-velocity selectivity of grid cells. Our network model is able to perform robust path integration even when the recurrent connections are subject to random perturbations.     

Saccadic motor planning by integrating visual information and pre-information on neural dynamic fields

A functional model of target selection in the saccadic system is presented, incorporating elements of visual processing, motor planning, and motor control. We address the integration of visual information with pre-information. which is provided by manipulating the probability that a target appears at a certain location. This integration is achieved within a dynamic representation of planned eye movement which is modeled through distributions of activation on a topographic field. Visual input evokes activation, which is also constrained by lateral interaction within the field and by preshaping input representing pre-information. The model describes target selection observable in paradigms in which visual goals are presented at more than one location. Specifically, we model the transition from averaging, where endpoints of first saccades fall between two visual target locations, to decision making, where endpoints of first saccades fall accurately onto one of two simultaneously presented visual targets. We make predictions about how metrical biases of first saccades are induced by pre-information about target locations acquired by learning. When coupled to a motor control stage, activation dynamics on the planning level contribute to stabilizing gaze under fixation conditions. The neurophysiological relevance of our functional model is discussed.     

Biomimetic Construction of Category Mental Imagery Based on Recognition Mechanism of Visual Cortex of Human Brain

The principle of homology-continuity in Multi-Dimensional Biomimetic Informatics Space is applied to construct the identifying mechanism of category of deep representation of mental imagery. The model of each cerebral region involved in recognizing is established respectively and a feedforward method for establishing category mental imagery is proposed. First, the model of feature acquisition is developed based on Hubel-Wiesel model, and Gaussian function is used to simulate the simple cell receptive field to satisfy the specific function of visual cortex. Second, multiple input aggregation operation is employed to simulate the feature output of complex cells to get the invariance representation in feature space. Then, imagery basis is extracted by unsupervised learning algorithm based on the primary feature and category mental imagery is obtained by building Radial Basis Function (RBF) network. Finally, the system model is tested by training set and test set composed of real images. Experimental results show that the proposed method can establish valid deep representation of these samples, based on which the biomimetic construction of category mental imagery can be achieved. This method provides a new idea for solving imagery problem and studying imagery thinking.     

Computational model for perception of objects and motions

Perception of objects and motions in the visual scene is one of the basic problems in the visual system. There exist 'What' and 'Where' pathways in the superior visual cortex, starting from the simple cells in the primary visual cortex. The former is able to perceive objects such as forms, color, and texture, and the latter perceives 'where', for example, velocity and direction of spatial movement of objects. This paper explores brain-like computational architectures of visual information processing. We propose a visual perceptual model and computational mechanism for training the perceptual model. The compu- tational model is a three-layer network. The first layer is the input layer which is used to receive the stimuli from natural environments. The second layer is designed for representing the internal neural information. The connections between the first layer and the second layer, called the receptive fields of neurons, are self-adaptively learned based on principle of sparse neural representation. To this end, we introduce Kullback-Leibler divergence as the measure of independence between neural responses and derive the learning algorithm based on minimizing the cost function. The proposed algorithm is applied to train the basis functions, namely receptive fields, which are localized, oriented, and bandpassed. The resultant receptive fields of neurons in the second layer have the characteristics resembling that of simple cells in the primary visual cortex. Based on these basis functions, we further construct the third layer for perception of what and where in the superior visual cortex. The proposed model is able to perceive objects and their motions with a high accuracy and strong robustness against additive noise. Computer simulation results in the final section show the feasibility of the proposed perceptual model and high efficiency of the learning algorithm.     

Computational model for perception of objects and motions

Perception of objects and motions in the visual scene is one of the basic problems in the visual system. There exist ‘What’ and ‘Where’ pathways in the superior visual cortex, starting from the simple cells in the primary visual cortex. The former is able to perceive objects such as forms, color, and texture, and the latter perceives ‘where’, for example, velocity and direction of spatial movement of objects. This paper explores brain-like computational architectures of visual information processing. We propose a visual perceptual model and computational mechanism for training the perceptual model. The computational model is a three-layer network. The first layer is the input layer which is used to receive the stimuli from natural environments. The second layer is designed for representing the internal neural information. The connections between the first layer and the second layer, called the receptive fields of neurons, are self-adaptively learned based on principle of sparse neural representation. To this end, we introduce Kullback-Leibler divergence as the measure of independence between neural responses and derive the learning algorithm based on minimizing the cost function. The proposed algorithm is applied to train the basis functions, namely receptive fields, which are localized, oriented, and bandpassed. The resultant receptive fields of neurons in the second layer have the characteristics resembling that of simple cells in the primary visual cortex. Based on these basis functions, we further construct the third layer for perception of what and where in the superior visual cortex. The proposed model is able to perceive objects and their motions with a high accuracy and strong robustness against additive noise. Computer simulation results in the final section show the feasibility of the proposed perceptual model and high efficiency of the learning algorithm.     

Visual map development: bidirectional signaling, bifunctional guidance molecules, and competition

Topographic maps are a two-dimensional representation of one neural structure within another and serve as the main strategy to organize sensory information. The retina's projection via axons of retinal ganglion cells to midbrain visual centers, the optic tectum/superior colliculus, is the leading model to elucidate mechanisms of topographic map formation. Each axis of the retina is mapped independently using different mechanisms and sets of axon guidance molecules expressed in gradients to achieve the goal of representing a point in the retina onto a point within the target. An axon's termination along the temporal-nasal mapping axis is determined by opposing gradients of EphAs and ephrin-As that act through their forward and reverse signaling, respectively, within the projecting axons, each of which inhibits interstitial branching, cooperating with a branch-promoting activity, to generate topographic specific branching along the shaft of the parent axons that overshoot their correct termination zone along the anterior-posterior axis of the target. The dorsal-ventral termination position is then determined using a gradient of ephrin-B that can act as a repellent or attractant depending on the ephrin-B concentration relative to EphB levels on the interstitial branches to guide them along the medial-lateral axis of the target to their correct termination zone, where they arborize. In both cases, axon-axon competition results in axon mapping based on relative rather than absolute levels of repellent or attractant activity. The map is subsequently refined through large-scale pruning driven in large part by patterned retinal activity.     

视觉皮层复杂细胞时空编码特性

针对输入在视皮层的编码表达,在地空滤波窗口基础上构建了一个复杂细胞时空编码模型,对几种特殊的输入函数进行了编码仿真实验,结果说明了视皮层复杂细胞时空整合编码序列的精细时间结构进行视觉输入的神经表象。     

Reinforcement Learning on Slow Features of High-Dimensional Input Streams

Humans and animals are able to learn complex behaviors based on a massive stream of sensory information from different modalities. Early animal studies have identified learning mechanisms that are based on reward and punishment such that animals tend to avoid actions that lead to punishment whereas rewarded actions are reinforced. However, most algorithms for reward-based learning are only applicable if the dimensionality of the state-space is sufficiently small or its structure is sufficiently simple. Therefore, the question arises how the problem of learning on high-dimensional data is solved in the brain. In this article, we propose a biologically plausible generic two-stage learning system that can directly be applied to raw high-dimensional input streams. The system is composed of a hierarchical slow feature analysis (SFA) network for preprocessing and a simple neural network on top that is trained based on rewards. We demonstrate by computer simulations that this generic architecture is able to learn quite demanding reinforcement learning tasks on high-dimensional visual input streams in a time that is comparable to the time needed when an explicit highly informative low-dimensional state-space representation is given instead of the high-dimensional visual input. The learning speed of the proposed architecture in a task similar to the Morris water maze task is comparable to that found in experimental studies with rats. This study thus supports the hypothesis that slowness learning is one important unsupervised learning principle utilized in the brain to form efficient state representations for behavioral learning.