Saccade-related remapping of target representations between topographic maps: a neural network study

The goal of this study was to explore how a neural network could solve the updating task associated with the double-saccade paradigm, where two targets are flashed in succession and the subject must make saccades to the remembered locations of both targets. Because of the eye rotation of the saccade to the first target, the remembered retinal position of the second target must be updated if an accurate saccade to that target is to be made. We trained a three-layer, feed-forward neural network to solve this updating task using back-propagation. The network's inputs were the initial retinal position of the second target represented by a hill of activation in a 2D topographic array of units, as well as the initial eye orientation and the motor error of the saccade to the first target, each represented as 3D vectors in brainstem coordinates. The output of the network was the updated retinal position of the second target, also represented in a 2D topographic array of units. The network was trained to perform this updating using the full 3D geometry of eye rotations, and was able to produce the updated second-target position to within a 1 degrees RMS accuracy for a set of test points that included saccades of up to 70 degrees . Emergent properties in the network's hidden layer included sigmoidal receptive fields whose orientations formed distinct clusters, and predictive remapping similar to that seen in brain areas associated with saccade generation. Networks with the larger numbers of hidden-layer units developed two distinct types of units with different transformation properties: units that preferentially performed the linear remapping of vector subtraction, and units that performed the nonlinear elements of remapping that arise from initial eye orientation.     

Self-Organizing Task Modules and Explicit Coordinate Systems in a Neural Network Model for 3-D Saccades

The goal of this study was to train an artificial neural network to generate accurate saccades in Listing's plane and then determine how the hidden units performed the visuomotor transformation. A three-layer neural network was successfully trained, using back-prop, to take in oculocentric retinal error vectors and three-dimensional eye orientation and to generate the correct head-centric motor error vector within Listing's plane. Analysis of the hidden layer of trained networks showed that explicit representations of desired target direction and eye orientation were not employed. Instead, the hidden-layer units consistently divided themselves into four parallel modules: a dominant "vector-propagation" class (approximately 50% of units) with similar visual and motor tuning but negligible position sensitivity and three classes with specific spatial relations between position, visual, and motor tuning. Surprisingly, the vector-propagation units, and only these, formed a highly precise and consistent orthogonal coordinate system aligned with Listing's plane. Selective "lesions" confirmed that the vector-propagation module provided the main drive for saccade magnitude and direction, whereas a balance between activity in the other modules was required for the correct eye-position modulation. Thus, contrary to popular expectation, error-driven learning in itself was sufficient to produce a "neural" algorithm with discrete functional modules and explicit coordinate systems, much like those observed in the real saccade generator.     

A neural network study of precollicular saccadic averaging

Saccadic averaging is the phenomenon that two simultaneously presented retinal inputs result in a saccade with an endpoint located on an intermediate position between the two stimuli. Recordings from neurons in the deeper layers of the superior colliculus have revealed neural correlates of saccade averaging, indicating that it takes place at this level or upstream. Recently, we proposed a neural network for internal feedback in saccades. This neural network model is different from other models in that it suggests the possibility that averaging takes place in a stage upstream of the colliculus. The network consists of output units representing the neural map of the deeper layers of the superior colliculus and hidden layers imitating areas in the posterior parietal cortex. The deeper layers of the superior colliculus represent the motor error of a desired saccade, e.g. an eye movement to a visual target. In this article we show that averaging is an emergent property of the proposed network. When two retinal targets with different intensities are simultaneously presented to the network, the activity in the output layer represents a single motor error with a weighted average value. Our goal is to understand the mechanism of weighted averaging in this neural network. It appears that averaging in the model is caused by the linear dependence of the net input, received by the hidden units, on retinal error, independent of its retinal coding format. For nonnormalized retinal error inputs, also the nonlinearity between the net input and the activity of the hidden units plays a role in the averaging process. The averaging properties of the model are in agreement with physiological experiments if the hypothetical retinal error input map is normalized. The neural network predicts that if this normalization is overruled by electrical stimulation, averaging still takes place. However, in this case – as a consequence of the feedback task – the location of the resulting saccade depends on the initial eye position and the total intensity/current applied at the two locations. This could be a way to verify the neural network model. If the assumptions for the model are valid, a physiological implication of this paper is that averaging of saccades takes place upstream of the superior colliculus. Received: 22 June 1997 / Accepted in revised form: 19 February 1998     

A novel interpretation for the collicular role in saccade generation

Recently, we found evidence that the activity of neurons in the deep layers of the monkey superior colliculus (SC) is modulated by initial eye position (gain fields). In this paper, we propose a quantitative model of the motor SC which incorporates these new findings. Inputs to the motor map represent the desired eye displacement vector (motor error), as well as initial eye position. A unit's activity in the motor map is described by multiplying a weak linear eye position sensitivity with a gaussian tuning to motor error. The motor map projects to several sets of output neurons, representing the coordinates of the desired eye displacement vector, the desired eye position in the head, and the three-dimensional ocular rotation axis for saccades in Listing's plane, respectively. All these signals have been hypothesized in the literature to drive the saccade burst generator. We show that these signals can be extracted from the motor map by a linear weighting of the population activity. The saccadic system may employ all coding strategies in parallel to ensure high spatial accuracy in many complex sensorimotor tasks, such as orienting to multimodal stimuli.     

A two dimensional model for saccade generation

A model for the generation of oblique saccades is constructed by extending and modifying the one dimensional local feedback model. It is proposed that the visual system stores target location in inertial coordinates, but that the feedback loop which guides saccades works in retinotopic coordinates. To achieve straight trajectories for centripetal and centrifugal saccades in all meridians, a comparator computes motor error as a vector and uses the vectorial error signal to drive two orthogonally-acting burst generators. The generation of straight saccade trajectories when the extraocular muscles are of unequal strengths requires the introduction of a burst-tonic cell input to motor neurons. The model accounts for the results of two-site stimulation of the superior colliculus and frontal eye fields by allowing simultaneous activation of more than one comparator. The postulated existence of multiple comparators suggests that motor error may be computed topographically.     

Biologically inspired calibration-free adaptive saccade control of a binocular camera-head

This paper describes fast and accurate calibration-free adaptive saccade control of a four-degrees-of-freedom binocular camera-head by means of Dynamic Cell Structures (DCS). The approach has been inspired by biology because primates face a similar problem and there is strong evidence that they have solved it in a similar way, i.e., by error feedback learning of an inverse model. Yet the emphasis of this article is not on detailed biological modeling but on how incremental growth of our artificial neural network model up to a prespecified precision results in very small networks suitable for real-time saccade control. Error-feedback-based training of this network proceeds in two phases. In the first phase we use a crude model of the cameras and the kinematics of the head to learn the topology of the input manifold together with a rough approximation of the control function off-line. In contrast to, for example, Kohonen-type adaptation rules, the distribution of neural units minimizes the control error and does not merely mimic the input probability density. In the second phase, the operating phase, the linear output units of the network continue to adapt on-line. Besides our TRC binocular camera-head we use a Datacube image processing system and a St?ubli R90 robot arm for automated training in the second phase. It will be demonstrated that the controller successfully corrects errors in the model and rapidly adapts to changing parameters. Received: 27 August 1996 / Accepted in revised form: 22 July 1997     

Spatiotopic visual maps revealed by saccadic adaptation in humans

Saccadic adaptation [1] is a powerful experimental paradigm to probe the mechanisms of eye movement control and spatial vision, in which saccadic amplitudes change in response to false visual feedback. The adaptation occurs primarily in the motor system [2, 3], but there is also evidence for visual adaptation, depending on the size and the permanence of the postsaccadic error [4-7]. Here we confirm that adaptation has a strong visual component and show that the visual component of the adaptation is spatially selective in external, not retinal coordinates. Subjects performed?a memory-guided, double-saccade, outward-adaptation task designed to maximize visual adaptation and to dissociate the visual and motor corrections. When the memorized saccadic target was in the same position (in external space) as that used in the adaptation training, saccade targeting was strongly influenced by adaptation (even if not matched in retinal or cranial position), but when in the same retinal or cranial but different external spatial position, targeting was unaffected by adaptation, demonstrating unequivocal spatiotopic selectivity. These results point to the existence of a spatiotopic neural representation for eye movement control that adapts in response to saccade error signals.     

Craniotopic updating of visual space across saccades in the human posterior parietal cortex

The neural mechanisms underlying the craniotopic updating of visual space across saccadic eye movements are poorly understood. Previous single-unit recording studies in primates and clinical studies in brain-damaged patients have shown that the posterior parietal cortex (PPC) has a key role in this process. In the present study, we used single-pulse transcranial magnetic stimulation (TMS) to disrupt the processing within the PPC during a task that requires craniotopic updating: double saccades. In this task, two targets are presented in quick succession and the subject is required to make a saccade to each location as accurately as possible. We show here that TMS delivered to the PPC just prior to the second saccade effectively disrupts the craniotopic coding normally observed in this task. This causes subjects to revert to saccades more consistent with a representation of the targets based on their positions relative to one another. By contrast, stimulation at earlier times between the two saccades did not disrupt performance. These results suggest that extraretinal information generated during the first perisaccadic period is not put into functional use until just prior to the second saccade.     

Human fast negative EEG potentials before express saccades

Fast negative EEG potentials preceding fast regular saccades and express saccades were studied by the method of backward averaging under conditions of monocular stimulation of the right and left eye. "Step" and "gap" experimental paradigms were used for visual stimulation. Analysis of parameters of potentials and their spatiotemporal dynamics suggests that, under conditions of the increased attention and optimal readiness of the neural structures, express saccades appear when the previously chosen program of the future eye movement coincides with the actual target coordinates. We assumed that the saccade latency decreases at the expense of the involvement of the main oculomotor areas of motor and saccadic planning in its initiation; an express saccade can be initiated also by means of direct transmission of the signal from the cortex to the brainstem saccadic generator passing by the superior colliculus. Moreover, anticipating release from the central fixation and attention distraction are necessary for the successful initiation of an express saccade.     

Neural network simulations of the primate oculomotor system

The performance of a neural network that simulates the vertical saccade-generating portion of the primate brain is evaluated. Consistent with presently available anatomical evidence, the model makes use of an eye displacement signal for its feedback. Its major features include a simple mechanism for resetting its integrator at the end of each saccade, the ability to generate staircases of saccades in response to stimulation of the superior colliculus, and the ability to account for the monotonic relation between motor error and the instantaneous discharge of presaccadic neurons of the superior colliculus without placing the latter within the local feedback loop. Several experimentally testable predictions about the effects of stimulation or lesion of saccaderelated areas of the primate brain are made on the basis of model output in response to “stimulation” or “lesion” of model elements.     

Coordinates transformation and learning control for visually-guided voluntary movement with iteration: A Newton-like method in a function space

In order to control visually-guided voluntary movements, the central nervous system (CNS) must solve the following three computational problems at different levels: (1) determination of a desired trajectory in the visual coordinates, (2) transformation of the coordinates of the desired trajectory to the body coordinates and (3) generation of motor command. In this paper, the second and the third problems are treated at computational, representational and hardware levels of Marr. We first study the problems at the computational level, and then propose an iterative learning scheme as a possible algorithm. This is a trial and error type learning such as repetitive training of golf swing. The amount of motor command needed to coordinate activities of many muscles is not determined at once, but in a step-wise, trial and error fashion in the course of a set of repetitions. Actually, the motor command in the (n+1)-th iteration is a sum of the motor command in then-th iteration plus two modification terms which are, respectively, proportional to acceleration and speed errors between the desired trajectory and the realized trajectory in then-th iteration. We mathematically formulate this iterative learning control as a Newton-like method in functional spaces and prove its convergence under appropriate mathematical conditions with use of dynamical system theory and functional analysis. Computer simulations of this iterative learning control of a robotic manipulator in the body or visual coordinates are shown. Finally, we propose that areas 2, 5, and 7 of the sensory association cortex are possible sites of this learning control. Further we propose neural network model which acquires transformation matrices from acceleration or velocity to motor command, which are used in these schemes.     

Motion Control in Saccade and Smooth Pursuit for Bionic Eye Based on Three-dimensional Coordinates

Saccade and smooth pursuit are two important functions of human eye.In order to enable bionic eye to imitate the two functions,a control method that implements saccade and smooth pursuit based on the three-dimensional coordinates of target is proposed.An optimal observation position is defined for bionic eye based on three-dimensional coordinates.A kind of motion planning method with high accuracy is developed.The motion parameters of stepper motor consisting of angle acceleration and turning time are computed according to the position deviation,the target's angular velocity and the stepper motor's current angular velocity in motion planning.The motors are controlled with the motion parameters moving to given position with desired angular velocity in schedule time.The experimental results show that the bionic eye can move to optimal observation positions in 0.6 s from initial location and the accuracy of 3D coordinates is improved.In addition,the bionic eye can track a target within the error of less than 20 pixels based on three-dimensional coordinates.It is verified that saccade and smooth pursuit of bionic eye based on three-dimensional coordinates are feasible.     

Sensory processing of motor inaccuracy depends on previously performed movement and on subsequent motor corrections: a study of the saccadic system

When goal-directed movements are inaccurate, two responses are generated by the brain: a fast motor correction toward the target and an adaptive motor recalibration developing progressively across subsequent trials. For the saccadic system, there is a clear dissociation between the fast motor correction (corrective saccade production) and the adaptive motor recalibration (primary saccade modification). Error signals used to trigger corrective saccades and to induce adaptation are based on post-saccadic visual feedback. The goal of this study was to determine if similar or different error signals are involved in saccadic adaptation and in corrective saccade generation. Saccadic accuracy was experimentally altered by systematically displacing the visual target during motor execution. Post-saccadic error signals were studied by manipulating visual information in two ways. First, the duration of the displaced target after primary saccade termination was set at 15, 50, 100 or 800 ms in different adaptation sessions. Second, in some sessions, the displaced target was followed by a visual mask that interfered with visual processing. Because they rely on different mechanisms, the adaptation of reactive saccades and the adaptation of voluntary saccades were both evaluated. We found that saccadic adaptation and corrective saccade production were both affected by the manipulations of post-saccadic visual information, but in different ways. This first finding suggests that different types of error signal processing are involved in the induction of these two motor corrections. Interestingly, voluntary saccades required a longer duration of post-saccadic target presentation to reach the same amount of adaptation as reactive saccades. Finally, the visual mask interfered with the production of corrective saccades only during the voluntary saccades adaptation task. These last observations suggest that post-saccadic perception depends on the previously performed action and that the differences between saccade categories of motor correction and adaptation occur at an early level of visual processing.     

Effects of stimulus-response compatibility on neural selection in frontal eye field

We investigated the neural basis of visual and saccade selection in the frontal eye field of macaque monkeys using a singleton search task with prosaccade or antisaccade responses. Two types of neurons were distinguished. The first initially selected the singleton even in antisaccade trials, although most subsequently selected the endpoint of the saccade. The time the singleton was located was not affected by stimulus-response compatibility and did not vary with reaction time across trials. The second type of neuron selected only the endpoint of the saccade. The time of endpoint selection by these neurons accounted for most of the effect of stimulus-response compatibility on reaction time. These results indicate that visual selection and saccade selection are different processes.     

Image Processing and Artificial Neural Network-Based Models to Measure and Predict Physical Properties of Embryogenic Callus and Number of Somatic Embryos in Ajowan (<Emphasis Type="Italic">Trachyspermum ammi</Emphasis> (L.) Sprague)

Trachyspermum ammi (L.) Sprague (Ajowan) is an endangered medicinal plant with useful pharmaceutical properties. Ex situ conservation of this medicinal plant needs the development of an in vitro regeneration protocol using somatic embryogenesis. In the present study, a high-precision image-processing approach was successfully applied to measure physical properties of embryogenic callus. Explant age and the concentrations of 2,4-dichlorophenoxyacetic acid (2,4-D), kinetin (Kin), and sucrose were used as inputs, and an artificial intelligence technique was applied to predict physical properties of embryogenic callus, and the number of somatic embryos produced. Artificial neural network (ANN) models were tested to find the best combinations of input variables that affected output variables. The lower values of root mean square error, and mean absolute error, and the highest values of determination coefficient, were achieved when all four input variables were applied to predict the number of somatic embryos, the area of the callus, the perimeter of the callus, the Feret diameter of the callus, the roundness of the callus, and the true density of the callus in ANN models. The highest measured and predicted number of somatic embryos were achieved from the interaction of 15-d-old explants?×?1.5 mg L^?1 2,4-D?×?0.5 mg L^?1 Kin?×?2.5% (w/v) sucrose. Based on sensitivity analysis, the 2,4-D concentration was the most important component in the culture medium that affected the number of somatic embryos and physical properties of the embryogenic callus tissue.     

Learning with slight forgetting optimizes sensorimotor transformation in redundant motor systems

Recent theoretical studies have proposed that the redundant motor system in humans achieves well-organized stereotypical movements by minimizing motor effort cost and motor error. However, it is unclear how this optimization process is implemented in the brain, presumably because conventional schemes have assumed a priori that the brain somehow constructs the optimal motor command, and largely ignored the underlying trial-by-trial learning process. In contrast, recent studies focusing on the trial-by-trial modification of motor commands based on error information suggested that forgetting (i.e., memory decay), which is usually considered as an inconvenient factor in motor learning, plays an important role in minimizing the motor effort cost. Here, we examine whether trial-by-trial error-feedback learning with slight forgetting could minimize the motor effort and error in a highly redundant neural network for sensorimotor transformation and whether it could predict the stereotypical activation patterns observed in primary motor cortex (M1) neurons. First, using a simple linear neural network model, we theoretically demonstrated that: 1) this algorithm consistently leads the neural network to converge at a unique optimal state; 2) the biomechanical properties of the musculoskeletal system necessarily determine the distribution of the preferred directions (PD; the direction in which the neuron is maximally active) of M1 neurons; and 3) the bias of the PDs is steadily formed during the minimization of the motor effort. Furthermore, using a non-linear network model with realistic musculoskeletal data, we demonstrated numerically that this algorithm could consistently reproduce the PD distribution observed in various motor tasks, including two-dimensional isometric torque production, two-dimensional reaching, and even three-dimensional reaching tasks. These results may suggest that slight forgetting in the sensorimotor transformation network is responsible for solving the redundancy problem in motor control.     

A hierarchical method for finding optimal architecture and weights using evolutionary least square based learning

In this paper, we present a novel approach of implementing a combination methodology to find appropriate neural network architecture and weights using an evolutionary least square based algorithm (GALS).1 This paper focuses on aspects such as the heuristics of updating weights using an evolutionary least square based algorithm, finding the number of hidden neurons for a two layer feed forward neural network, the stopping criterion for the algorithm and finally some comparisons of the results with other existing methods for searching optimal or near optimal solution in the multidimensional complex search space comprising the architecture and the weight variables. We explain how the weight updating algorithm using evolutionary least square based approach can be combined with the growing architecture model to find the optimum number of hidden neurons. We also discuss the issues of finding a probabilistic solution space as a starting point for the least square method and address the problems involving fitness breaking. We apply the proposed approach to XOR problem, 10 bit odd parity problem and many real-world benchmark data sets such as handwriting data set from CEDAR, breast cancer and heart disease data sets from UCI ML repository. The comparative results based on classification accuracy and the time complexity are discussed.     

A hierarchical neural-network model for control and learning of voluntary movement

In order to control voluntary movements, the central nervous system (CNS) must solve the following three computational problems at different levels: the determination of a desired trajectory in the visual coordinates, the transformation of its coordinates to the body coordinates and the generation of motor command. Based on physiological knowledge and previous models, we propose a hierarchical neural network model which accounts for the generation of motor command. In our model the association cortex provides the motor cortex with the desired trajectory in the body coordinates, where the motor command is then calculated by means of long-loop sensory feedback. Within the spinocerebellum — magnocellular red nucleus system, an internal neural model of the dynamics of the musculoskeletal system is acquired with practice, because of the heterosynaptic plasticity, while monitoring the motor command and the results of movement. Internal feedback control with this dynamical model updates the motor command by predicting a possible error of movement. Within the cerebrocerebellum — parvocellular red nucleus system, an internal neural model of the inverse-dynamics of the musculo-skeletal system is acquired while monitoring the desired trajectory and the motor command. The inverse-dynamics model substitutes for other brain regions in the complex computation of the motor command. The dynamics and the inverse-dynamics models are realized by a parallel distributed neural network, which comprises many sub-systems computing various nonlinear transformations of input signals and a neuron with heterosynaptic plasticity (that is, changes of synaptic weights are assumed proportional to a product of two kinds of synaptic inputs). Control and learning performance of the model was investigated by computer simulation, in which a robotic manipulator was used as a controlled system, with the following results: (1) Both the dynamics and the inverse-dynamics models were acquired during control of movements. (2) As motor learning proceeded, the inverse-dynamics model gradually took the place of external feedback as the main controller. Concomitantly, overall control performance became much better. (3) Once the neural network model learned to control some movement, it could control quite different and faster movements. (4) The neural netowrk model worked well even when only very limited information about the fundamental dynamical structure of the controlled system was available. Consequently, the model not only accounts for the learning and control capability of the CNS, but also provides a promising parallel-distributed control scheme for a large-scale complex object whose dynamics are only partially known.     

A biologically inspired neural net for trajectory formation and obstacle avoidance

In this paper we present a biologically inspired two-layered neural network for trajectory formation and obstacle avoidance. The two topographically ordered neural maps consist of analog neurons having continuous dynamics. The first layer, the sensory map, receives sensory information and builds up an activity pattern which contains the optimal solution (i.e. shortest path without collisions) for any given set of current position, target positions and obstacle positions. Targets and obstacles are allowed to move, in which case the activity pattern in the sensory map will change accordingly. The time evolution of the neural activity in the second layer, the motor map, results in a moving cluster of activity, which can be interpreted as a population vector. Through the feedforward connections between the two layers, input of the sensory map directs the movement of the cluster along the optimal path from the current position of the cluster to the target position. The smooth trajectory is the result of the intrinsic dynamics of the network only. No supervisor is required. The output of the motor map can be used for direct control of an autonomous system in a cluttered environment or for control of the actuators of a biological limb or robot manipulator. The system is able to reach a target even in the presence of an external perturbation. Computer simulations of a point robot and a multi-joint manipulator illustrate the theory.