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.     

A large-scale model of the locust antennal lobe

The antennal lobe (AL) is the primary structure within the locust’s brain that receives information from olfactory receptor neurons (ORNs) within the antennae. Different odors activate distinct subsets of ORNs, implying that neuronal signals at the level of the antennae encode odors combinatorially. Within the AL, however, different odors produce signals with long-lasting dynamic transients carried by overlapping neural ensembles, suggesting a more complex coding scheme. In this work we use a large-scale point neuron model of the locust AL to investigate this shift in stimulus encoding and potential consequences for odor discrimination. Consistent with experiment, our model produces stimulus-sensitive, dynamically evolving populations of active AL neurons. Our model relies critically on the persistence time-scale associated with ORN input to the AL, sparse connectivity among projection neurons, and a synaptic slow inhibitory mechanism. Collectively, these architectural features can generate network odor representations of considerably higher dimension than would be generated by a direct feed-forward representation of stimulus space.     

How Lateral Connections and Spiking Dynamics May Separate Multiple Objects Moving Together

Over successive stages, the ventral visual system of the primate brain develops neurons that respond selectively to particular objects or faces with translation, size and view invariance. The powerful neural representations found in Inferotemporal cortex form a remarkably rapid and robust basis for object recognition which belies the difficulties faced by the system when learning in natural visual environments. A central issue in understanding the process of biological object recognition is how these neurons learn to form separate representations of objects from complex visual scenes composed of multiple objects. We show how a one-layer competitive network comprised of ‘spiking’ neurons is able to learn separate transformation-invariant representations (exemplified by one-dimensional translations) of visual objects that are always seen together moving in lock-step, but separated in space. This is achieved by combining ‘Mexican hat’ functional lateral connectivity with cell firing-rate adaptation to temporally segment input representations of competing stimuli through anti-phase oscillations (perceptual cycles). These spiking dynamics are quickly and reliably generated, enabling selective modification of the feed-forward connections to neurons in the next layer through Spike-Time-Dependent Plasticity (STDP), resulting in separate translation-invariant representations of each stimulus. Variations in key properties of the model are investigated with respect to the network’s ability to develop appropriate input representations and subsequently output representations through STDP. Contrary to earlier rate-coded models of this learning process, this work shows how spiking neural networks may learn about more than one stimulus together without suffering from the ‘superposition catastrophe’. We take these results to suggest that spiking dynamics are key to understanding biological visual object recognition.     

Invariant recognition of feature combinations in the visual system

 The operation of a hierarchical competitive network model (VisNet) of invariance learning in the visual system is investigated to determine how this class of architecture can solve problems that require the spatial binding of features. First, we show that VisNet neurons can be trained to provide transform-invariant discriminative responses to stimuli which are composed of the same basic alphabet of features, where no single stimulus contains a unique feature not shared by any other stimulus. The investigation shows that the network can discriminate stimuli consisting of sets of features which are subsets or supersets of each other. Second, a key feature-binding issue we address is how invariant representations of low-order combinations of features in the early layers of the visual system are able to uniquely specify the correct spatial arrangement of features in the overall stimulus and ensure correct stimulus identification in the output layer. We show that output layer neurons can learn new stimuli if the lower layers are trained solely through exposure to simpler feature combinations from which the new stimuli are composed. Moreover, we show that after training on the low-order feature combinations which are common to many objects, this architecture can – after training with a whole stimulus in some locations – generalise correctly to the same stimulus when it is shown in a new location. We conclude that this type of hierarchical model can solve feature-binding problems to produce correct invariant identification of whole stimuli. Received: 4 August 1999 / Accepted in revised form: 11 October 2000     

Shape Invariant Coding of Motion Direction in Somatosensory Cortex

Invariant representations of stimulus features are thought to play an important role in producing stable percepts of objects. In the present study, we assess the invariance of neural representations of tactile motion direction with respect to other stimulus properties. To this end, we record the responses evoked in individual neurons in somatosensory cortex of primates, including areas 3b, 1, and 2, by three types of motion stimuli, namely scanned bars and dot patterns, and random dot displays, presented to the fingertips of macaque monkeys. We identify a population of neurons in area 1 that is highly sensitive to the direction of stimulus motion and whose motion signals are invariant across stimulus types and conditions. The motion signals conveyed by individual neurons in area 1 can account for the ability of human observers to discriminate the direction of motion of these stimuli, as measured in paired psychophysical experiments. We conclude that area 1 contains a robust representation of motion and discuss similarities in the neural mechanisms of visual and tactile motion processing.     

Neurophysiological mechanisms underlying face processing within and beyond the temporal cortical visual areas.

The ways in which information about faces is represented and stored in the temporal lobe visual areas of primates, as shown by recordings from single neurons in macaques, are considered. Some neurons that respond primarily to faces are found in the cortex in the anterior part of the superior temporal sulcus (in which neurons are especially likely to be tuned to facial expression and to face movement involved in gesture), and in the TE areas more ventrally forming the inferior temporal gyrus (in which neurons are more likely to have responses related to the identity of faces). Quantitative studies of the responses of the neurons that respond differently to the faces of different individuals show that information about the identity of the individual is represented by the responses of a population of neurons, that is, ensemble encoding rather than 'grandmother cell' encoding is used. It is argued that this type of tuning is a delicate compromise between very fine tuning, which has the advantage of low interference in neuronal network operations but the disadvantage of losing the useful properties (such as generalization, completion and graceful degradation) of storage in neuronal networks, and broad tuning, which has the advantage of allowing these properties of neuronal networks to be realized but the disadvantage of leading to interference between the different memories stored in an associative network. There is evidence that the responses of some of these neurons are altered by experience so that new stimuli become incorporated in the network. It is shown that the representation that is built in temporal cortical areas shows considerable invariance for size, contrast, spatial frequency and translation. Thus the representation is in a form which is particularly useful for storage and as an output from the visual system. It is also shown that one of the representations that is built is object based, which is suitable for recognition and as an input to associative memory, and that another is viewer centred, which is appropriate for conveying information about gesture. Ways are considered in which such cortical representations might be built by competitive self-organization aided by back projections in the multi-stage cortical processing hierarchy which has convergence from stage to stage.     

On the use of size functions for shape analysis

According to a recent mathematical theory a shape can be represented by size functions, which convey information on both the topological and metric properties of the viewed shape. In this paper the relevance of the theory of size functions to visual perception is investigated. An algorithm for the computation of the size functions is presented, and many theoretical properties of the theory are demonstrated on real images. It is shown that the representation of shape in terms of size functions (1) can be tailored to suit the invariance of the problem at hand and (2) is stable against small qualitative and quantitative changes of the viewed shape. A distance between size functions is used as a measure of similarity between the representations of two different shapes. The results obtained indicate that size functions are likely to be very useful for object recognition. In particular, they seem to be well suited for the recognition of natural and articulated objects.     

Functional implications of temporal structure in primate cortical information processing

Processing of information in the cerebral cortex of primates is characterized by distributed representations and processing in neuronal assemblies rather than by detector neurons, cardinal cells or command neurons. Responses of individual neurons in sensory cortical areas contain limited and ambiguous information on common features of the natural environment which is disambiguated by comparison with the responses of other, related neurons. Distributed representations are also capable to represent the enormous complexity and variability of the natural environment by the large number of possible combinations of neurons that can engage in the representation of a stimulus or other content. A critical problem of distributed representation and processing is the superposition of several assemblies activated at the same time since interpretation and processing of a population code requires that the responses related to a single representation can be identified and distinguished from other, related activity. A possible mechanism which tags related responses is the synchronization of neuronal responses of the same assembly with a precision in the millisecond range. This mechanism also supports the separate processing of distributed activity and dynamic assembly formation. Experimental evidence from electrophysiological investigations of non-human primates and human subjects shows that synchronous activity can be found in visual, auditory and motor areas of the cortex. Simultaneous recordings of neurons in the visual cortex indicate that individual neurons synchronize their activity with each other, if they respond to the same stimulus but not if they are part of different assemblies representing different contents. Furthermore, evidence for synchronous activity related to perception, expectation, memory, and attention has been observed.     

Spatial cognition and neuro-mimetic navigation: a model of hippocampal place cell activity

 A computational model of hippocampal activity during spatial cognition and navigation tasks is presented. The spatial representation in our model of the rat hippocampus is built on-line during exploration via two processing streams. An allothetic vision-based representation is built by unsupervised Hebbian learning extracting spatio-temporal properties of the environment from visual input. An idiothetic representation is learned based on internal movement-related information provided by path integration. On the level of the hippocampus, allothetic and idiothetic representations are integrated to yield a stable representation of the environment by a population of localized overlapping CA3-CA1 place fields. The hippocampal spatial representation is used as a basis for goal-oriented spatial behavior. We focus on the neural pathway connecting the hippocampus to the nucleus accumbens. Place cells drive a population of locomotor action neurons in the nucleus accumbens. Reward-based learning is applied to map place cell activity into action cell activity. The ensemble action cell activity provides navigational maps to support spatial behavior. We present experimental results obtained with a mobile Khepera robot. Received: 02 July 1999 / Accepted in revised form: 20 March 2000     

Temporal codes and sparse representations: A key to understanding rapid processing in the visual system

Where neural information processing is concerned, there is no debate about the fact that spikes are the basic currency for transmitting information between neurons. How the brain actually uses them to encode information remains more controversial. It is commonly assumed that neuronal firing rate is the key variable, but the speed with which images can be analysed by the visual system poses a major challenge for rate-based approaches. We will thus expose here the possibility that the brain makes use of the spatio-temporal structure of spike patterns to encode information. We then consider how such rapid selective neural responses can be generated rapidly through spike-timing-dependent plasticity (STDP) and how these selectivities can be used for visual representation and recognition. Finally, we show how temporal codes and sparse representations may very well arise one from another and explain some of the remarkable features of processing in the visual system.     

Neuronal representations of stimuli in the mouth: the primate insular taste cortex, orbitofrontal cortex and amygdala

The responses of 3687 neurons in the macaque primary taste cortex in the insula/frontal operculum, orbitofrontal cortex (OFC) and amygdala to oral sensory stimuli reveals principles of representation in these areas. Information about the taste, texture of what is in the mouth (viscosity, fat texture and grittiness, which reflect somatosensory inputs), temperature and capsaicin is represented in all three areas. In the primary taste cortex, taste and viscosity are more likely to activate different neurons, with more convergence onto single neurons particularly in the OFC and amygdala. The different responses of different OFC neurons to different combinations of these oral sensory stimuli potentially provides a basis for different behavioral responses. Consistently, the mean correlations between the representations of the different stimuli provided by the population of OFC neurons were lower (0.71) than for the insula (0.81) and amygdala (0.89). Further, the encoding was more sparse in the OFC (0.67) than in the insula (0.74) and amygdala (0.79). The insular neurons did not respond to olfactory and visual stimuli, with convergence occurring in the OFC and amygdala. Human psychophysics showed that the sensory spaces revealed by multidimensional scaling were similar to those provided by the neurons.     

Learning to Attend: Modeling the Shaping of Selectivity in Infero-temporal Cortex in a Categorization Task

Recent experiments on behaving monkeys have shown that learning a visual categorization task makes the neurons in infero-temporal cortex (ITC) more selective to the task-relevant features of the stimuli (Sigala and Logothetis in Nature 415 318–320, 2002). We hypothesize that such a selectivity modulation emerges from the interaction between ITC and other cortical area, presumably the prefrontal cortex (PFC), where the previously learned stimulus categories are encoded. We propose a biologically inspired model of excitatory and inhibitory spiking neurons with plastic synapses, modified according to a reward based Hebbian learning rule, to explain the experimental results and test the validity of our hypothesis. We assume that the ITC neurons, receiving feature selective inputs, form stronger connections with the category specific neurons to which they are consistently associated in rewarded trials. After learning, the top-down influence of PFC neurons enhances the selectivity of the ITC neurons encoding the behaviorally relevant features of the stimuli, as observed in the experiments. We conclude that the perceptual representation in visual areas like ITC can be strongly affected by the interaction with other areas which are devoted to higher cognitive functions. Electronic Supplementary Material: Supplementary material is available in the online: version of this article at http://dx.doi.org/10.007/s00422-006-0054-z     

From Sensory Signals to Modality-Independent Conceptual Representations: A Probabilistic Language of Thought Approach

People learn modality-independent, conceptual representations from modality-specific sensory signals. Here, we hypothesize that any system that accomplishes this feat will include three components: a representational language for characterizing modality-independent representations, a set of sensory-specific forward models for mapping from modality-independent representations to sensory signals, and an inference algorithm for inverting forward models—that is, an algorithm for using sensory signals to infer modality-independent representations. To evaluate this hypothesis, we instantiate it in the form of a computational model that learns object shape representations from visual and/or haptic signals. The model uses a probabilistic grammar to characterize modality-independent representations of object shape, uses a computer graphics toolkit and a human hand simulator to map from object representations to visual and haptic features, respectively, and uses a Bayesian inference algorithm to infer modality-independent object representations from visual and/or haptic signals. Simulation results show that the model infers identical object representations when an object is viewed, grasped, or both. That is, the model’s percepts are modality invariant. We also report the results of an experiment in which different subjects rated the similarity of pairs of objects in different sensory conditions, and show that the model provides a very accurate account of subjects’ ratings. Conceptually, this research significantly contributes to our understanding of modality invariance, an important type of perceptual constancy, by demonstrating how modality-independent representations can be acquired and used. Methodologically, it provides an important contribution to cognitive modeling, particularly an emerging probabilistic language-of-thought approach, by showing how symbolic and statistical approaches can be combined in order to understand aspects of human perception.     

Quantification of phase synchronization phenomena and their importance for verbal memory processes

The tangential neurons in the lobula plate region of the flies are known to respond to visual motion across broad receptive fields in visual space.When intracellular recordings are made from tangential neurons while the intact animal is stimulated visually with moving natural imagery,we find that neural response depends upon speed of motion but is nearly invariant with respect to variations in natural scenery. We refer to this invariance as velocity constancy. It is remarkable because natural scenes, in spite of similarities in spatial structure, vary considerably in contrast, and contrast dependence is a feature of neurons in the early visual pathway as well as of most models for the elementary operations of visual motion detection. Thus, we expect that operations must be present in the processing pathway that reduce contrast dependence in order to approximate velocity constancy.We consider models for such operations, including spatial filtering, motion adaptation, saturating nonlinearities, and nonlinear spatial integration by the tangential neurons themselves, and evaluate their effects in simulations of a tangential neuron and precursor processing in response to animated natural imagery. We conclude that all such features reduce interscene variance in response, but that the model system does not approach velocity constancy as closely as the biological tangential cell.     

Visual filling-in for computing perceptual surface properties

 The visual system is constantly confronted with the problem of integrating local signals into more global arrangements. This arises from the nature of early cell responses, whether they signal localized measures of luminance, motion, retinal position differences, or discontinuities. Consequently, from sparse, local measurements, the visual system must somehow generate the most likely hypothesis that is consistent with them. In this paper, we study the problem of determining achromatic surface properties, namely brightness. Mechanisms of brightness filling-in have been described by qualitative as well as quantitative models, such as by the one proposed by Cohen and Grossberg [Cohen and Grossberg (1984) Percept Psychophys 36: 428–456]. We demonstrate that filling-in from contrast estimates leads to a regularized solution for the computational problem of generating brightness representations from sparse estimates. This provides deeper insights into the nature of filling-in processes and the underlying objective function one wishes to compute. This particularly guided the proposal of a new modified version of filling-in, namely confidence-based filling-in which generates more robust brightness representations. Our investigation relates the modeling of perceptual data for biological vision to the mathematical frameworks of regularization theory and linear spatially variant diffusion. It therefore unifies different research directions that have so far coexisted in different scientific communities. Received: 11 March 1999 / Accepted in revised form: 14 March 2001     

A strategy for mapping biophysical to abstract neuronal network models applied to primary visual cortex

A fundamental challenge for the theoretical study of neuronal networks is to make the link between complex biophysical models based directly on experimental data, to progressively simpler mathematical models that allow the derivation of general operating principles. We present a strategy that successively maps a relatively detailed biophysical population model, comprising conductance-based Hodgkin-Huxley type neuron models with connectivity rules derived from anatomical data, to various representations with fewer parameters, finishing with a firing rate network model that permits analysis. We apply this methodology to primary visual cortex of higher mammals, focusing on the functional property of stimulus orientation selectivity of receptive fields of individual neurons. The mapping produces compact expressions for the parameters of the abstract model that clearly identify the impact of specific electrophysiological and anatomical parameters on the analytical results, in particular as manifested by specific functional signatures of visual cortex, including input-output sharpening, conductance invariance, virtual rotation and the tilt after effect. Importantly, qualitative differences between model behaviours point out consequences of various simplifications. The strategy may be applied to other neuronal systems with appropriate modifications.     

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.     

Algorithm independent properties of RNA secondary structure predictions

Algorithms predicting RNA secondary structures based on different folding criteria – minimum free energies (mfe), kinetic folding (kin), maximum matching (mm) – and different parameter sets are studied systematically. Two base pairing alphabets were used: the binary GC and the natural four-letter AUGC alphabet. Computed structures and free energies depend strongly on both the algorithm and the parameter set. Statistical properties, such as mean number of base pairs, mean numbers of stacks, mean loop sizes, etc., are much less sensitive to the choice of parameter set and even of algorithm. Some features of RNA secondary structures, such as structure correlation functions, shape space covering and neutral networks, seem to depend only on the base pairing logic (GC or AUGC alphabet). Received: 16 May 1996 / Accepted: 10 July 1996