The real time computation of motion from real images using a single chip with integrated sensors is a hard problem. We present two analog VLSI schemes that use pulse domain neuromorphic circuits to compute motion. Pul...
ISBN:
(纸本)9781558602748
The real time computation of motion from real images using a single chip with integrated sensors is a hard problem. We present two analog VLSI schemes that use pulse domain neuromorphic circuits to compute motion. Pulses of variable width, rather than graded potentials, represent a natural medium for evaluating temporal relationships. Both algorithms measure speed by timing a moving edge in the image. Our first model is inspired by Reichardt's algorithm in the fly and yields a non-monotonic response vs. velocity curve. We present data from a chip that implements this model. Our second algorithm yields a monotonic response vs. velocity curve and is currently being translated into silicon.
Simplified models of the lateral geniculate nucles (LGN) and striate cortex illustrate the possibility that feedback to the LGN may be used for robust, low-level pattern analysis. The information fed back to the LGN i...
ISBN:
(纸本)9781558602748
Simplified models of the lateral geniculate nucles (LGN) and striate cortex illustrate the possibility that feedback to the LGN may be used for robust, low-level pattern analysis. The information fed back to the LGN is rebroadcast to cortex using the LGN's full fan-out, so the cortex→LGN→cortex pathway mediates extensive cortico-cortical communication while keeping the number of necessary connections small.
We analyze the dynamic behavior of large two-dimensional systems of limit-cycle oscillators with random intrinsic frequencies that interact via time-delayed nearest-neighbor coupling. We find that even small delay tim...
We analyze the dynamic behavior of large two-dimensional systems of limit-cycle oscillators with random intrinsic frequencies that interact via time-delayed nearest-neighbor coupling. We find that even small delay times lead to a novel form of frequency depression where the system decays to stable states which oscillate at a delay and interaction-dependent reduced collective frequency. For greater delay or tighter coupling between oscillators we find metastable synchronized states that we describe analytically and numerically.
We investigate the dynamics of large arrays of coupled phase oscillators driven by random intrinsic frequencies under a variety of coupling schemes, by computing the time-dependent cross-correlation function numerical...
We investigate the dynamics of large arrays of coupled phase oscillators driven by random intrinsic frequencies under a variety of coupling schemes, by computing the time-dependent cross-correlation function numerically for a two-dimensional array consisting of 128×128 oscillators as well as analytically for a simpler model. Our analysis shows that for overall equal interaction strength, a sparse-coupling scheme in which each oscillator is coupled to a small, randomly selected subset of its neighbors leads to a more rapid and robust phase locking than nearest-neighbor coupling or locally dense connection schemes.
Biological retinas extract spatial and temporal features in an attempt to reduce the complexity of performing visual tasks. We have built and tested a silicon retina which encodes several useful temporal features foun...
ISBN:
(纸本)9781558602229
Biological retinas extract spatial and temporal features in an attempt to reduce the complexity of performing visual tasks. We have built and tested a silicon retina which encodes several useful temporal features found in vertebrate retinas. The cells in our silicon retina are selective to direction, highly sensitive to positive contrast changes around an ambient light level, and tuned to a particular velocity. Inhibitory connections in the null direction perform the direction selectivity we desire. This silicon retina is on a 4.6 × 6.8mm die and consists of a 47 × 41 array of photoreceptors.
Single nerve cells with static properties have traditionally been viewed as the building blocks for networks that show emergent phenomena. In contrast to this approach, we study here how the overall network activity c...
ISBN:
(纸本)9781558602229
Single nerve cells with static properties have traditionally been viewed as the building blocks for networks that show emergent phenomena. In contrast to this approach, we study here how the overall network activity can control single cell parameters such as input resistance, as well as time and space constants, parameters that are crucial for excitability and spatio-temporal integration. Using detailed computer simulations of neocortical pyramidal cells, we show that the spontaneous background firing of the network provides a means for setting these parameters. The mechanism for this control is through the large conductance change of the membrane that is induced by both non-NMDA and NMDA excitatory and inhibitory synapses activated by the spontaneous background activity.
The authors have designed, built and tested a number of analog CMOS VLSI circuits for computing 1D motion from the time-varying intensity values provided by an array of on-chip phototransistors. The authors present ex...
详细信息
The authors have designed, built and tested a number of analog CMOS VLSI circuits for computing 1D motion from the time-varying intensity values provided by an array of on-chip phototransistors. The authors present experimental data for three such circuits and discuss their relative performance. One circuit approximates the correlation model, one the gradient model, while a third chip uses resistive grids to compute zerocrossings to be tracked over time by a separate digital processor. All circuits integrate image acquisition with image processing functions and compute velocity in real time. Finally, for comparison, the authors also describe the performance of a simple motion algorithm using off-the-shelf components.< >
A method for transforming performance evaluation signals distal both in space and time into proximal signals usable by supervised learning algorithms, presented in [Jordan & Jacobs 90], is examined. A simple obser...
ISBN:
(纸本)9781558602229
A method for transforming performance evaluation signals distal both in space and time into proximal signals usable by supervised learning algorithms, presented in [Jordan & Jacobs 90], is examined. A simple observation concerning differentiation through models trained with redundant inputs (as one of their networks is) explains a weakness in the original architecture and suggests a modification: an internal world model that encodes action-space exploration and, crucially, cancels input redundancy to the forward model is added. Learning time on an example task, cartpole balancing, is thereby reduced about 50 to 100 times.
The accuracy of optical flow estimation depends on the spatio-temporal discretization used in the computation. The authors propose an adaptive multiscale method, where the discretization scale is chosen locally accord...
详细信息
The accuracy of optical flow estimation depends on the spatio-temporal discretization used in the computation. The authors propose an adaptive multiscale method, where the discretization scale is chosen locally according to an estimate of the relative error in the velocity measurements. They show that their coarse-to-fine method provides substantially better results of optical flow than conventional algorithms. The authors map this multiscale strategy onto their model of motion computation in primate area MT. The model consists of two stages: (1) local velocities are measured across multiple spatio-temporal channels, while (2) the optical flow field is computed by a network of direction-selective neurons at multiple spatial resolutions. Their model neurons show the same nonclassical receptive field properties as Allman's type I MT neurons and lead to a novel interpretation of some aspect of the motion capture illusion.< >
暂无评论