Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons (G06N3/06)

G   Physics(388756)
G06N3/06                     Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons(26)

ulti-input adder by module two // 2614370
FIELD: computer engineering.SUBSTANCE: invention relates to computer engineering and can be used in digital computing devices, as well as in end fields GF(2ν) elements formation devices. Technical result is achieved by using new activation function in hidden layer, use of synaptic weights ωi,j, equal to one, which enables to eliminate synaptic weights multipliers from formal neuron structure, as well as elimination from neuron structure of output layer of unit,implementing activation function calculations.EFFECT: technical result consists in reduction of system expenses required for multi-input adder realization by module two.1 cl, 1 dwg

Artificial neuron (versions) // 2604331
FIELD: computer engineering.SUBSTANCE: invention relates to bionics and computer engineering and can be used as structural and functional element of artificial neural networks for simulating biological neural networks, as well as for designing of parallel neural computers and other computer systems for various applications, including tasks of image recognition, data classification, image processing, mathematical operations and creation of artificial intelligence. Device comprises information and modulating inputs, information outputs, computing core consisting of processor device, application non-volatile memory, random access memory and non-volatile data memory interacting through system bus.EFFECT: technical result is providing most complete emulation of biological neuron functioning, as well as application versatility and multiplicity, which provides significant saving of hardware resources.5 cl, 3 dwg

Near-real impulse neuron // 2598298
FIELD: neurons simulation.SUBSTANCE: invention can be used on the neural computers, in the technical systems based on neural networks for pattern recognition, analysis and processing of the images. Device comprises inputs for impulse flows, interconnection links, cluster made of part of the input connection links; cluster is formed from the input connection links depending on the structure of periods and localisation of partial impulse flows at the neuron input, that implements existence for total impulse flow at the output of the cluster ε-almost-periods with maximum sum of amplitudes of all partial impulse flows; impulse flow from the output of the cluster is fed to the adder, which is associated with a liminal excitable element which generates an output impulse sequence or single impulse at exceeding of amplitude of impulse flows at the output of the adder liminal excitable element by the maximum sum; signal from the output of liminal excitable element is used as the output one.EFFECT: technical result is provision of an opportunity to achieve selective detection of the input objects without using a balance measurement of the input signals, possibility of encoding the input object of a certain type by the channel number or by the number of recording neuron, compression of input information, performance increase, objects recognition reliability increase.1 cl, 9 dwg, 2 tbl

Single-layer perceptron based on selective neurons // 2597497
FIELD: simulation of neural structures.SUBSTANCE: invention can be used on the neural computers, in the technical systems based on neural networks for pattern recognition, analysis and processing of the images. Device has several recording neurons, inputs for the signals from the objects, internal communication channels from the inputs, individual clusters in each perceptron recording neurons, which are formed from the part of the input of communication channels of separate neurons, formation of clusters and alignment of channels number in clusters is carried out in accordance with the code combinations of input signals, clusters from communication channels are connected to summers, after summers nonlinear threshold transformations of the signals are manufactured; converted signals are used as output signals.EFFECT: provision of an opportunity to achieve selective detection of input objects without using balance measurement of input signals, possibility of encoding the input object of a certain type by the channel number or by the number of recording neuron, compression of input information, performance increase, objects recognition reliability increase.1 cl, 6 dwg, 1 tbl

Single-layer perceptron, simulating real perceptron properties // 2597496
FIELD: simulation of neural structures.SUBSTANCE: invention can be used on the neural computers, in the technical systems based on neural networks for pattern recognition, analysis and processing of the images. Device comprises recording impulse neurons, inputs for input signals-impulse flows, internal communication channels from the inputs, individual clusters from part of communication channels in each of recording neurons, which pass the part of input signals; formation of clusters and adjustment of channels number in the clusters is carried out in accordance with the code combinations and structure of periods of partial components of input impulse flows; inertialess summation of all partial impulse flows at the outputs of clusters is performed; obtained signal is converted by liminal excitable elements, which generate single impulses or impulse sequences at exceeding of amplitude of impulse flows of liminal excitable elements which are used as the input signals.EFFECT: technical result is provision of an opportunity to achieve selective detection of the input objects without using a balance measurement of the input signals, possibility of encoding the input object of a certain type by the channel number or by the number of recording neuron, compression of input information, performance increase, objects recognition reliability increase.1 cl, 13 dwg, 2 tbl

Neuron simulation of real neuron properties // 2597495
FIELD: information technology.SUBSTANCE: invention relates to simulation of neurons and can be applied to neuro-computers, and technical systems based on neural networks for pattern identification, analysis and processing images. Device comprises inputs for signals from objects, internal communication channels from inputs, cluster that is formed from part of internal communication channels in accordance with code combination of input signal; cluster is connected to adder, after that, nonlinear threshold signal conversion is performed, which is used as output signal.EFFECT: technical result is provision of capability to achieve selective identification of input objects without using balanced measurement of input signals, capability of encoding input object of certain type by channel number or number of recording neuron, compression of input information, faster operation, higher reliability of object identification.1 cl, 7 dwg, 1 tbl

Integral image pickup apparatus and method // 2585985
FIELD: physics.SUBSTANCE: invention based on use of photosensitive cells built or embedded in a transparent or semitransparent substrate or on a substrate, for example, glass or plastic. Substrate itself can work as an optical device which focuses incident photons associated with reflected image onto photosensitive elements. Digital memory neurons can be trained to recognise objects in accordance with incident photons. Photosensitive elements and elements of digital memory neurons can be combined with light-emitting elements controlled in accordance with recognised images. In one application intelligent light-emitting units emit light by monitoring surrounding area and (or) by adjusting light in accordance with surrounding space. In another application of intelligent display show image and (or) video, monitoring surrounding area and (or) change displayed images and (or) a video image in accordance with surrounding space.EFFECT: technical result is creating an image recognition device having a sensitive area, directly embedded in a transparent or translucent material forming an optical interface.68 cl, 29 dwg

Artificial neuron // 2579958
FIELD: information technology.SUBSTANCE: invention can be used in the construction of information processing systems in the neural network basis, including pattern recognition (classification). Apparatus comprises an input nodes multiplying the input value by a weighting factor, adder block activation function, each node multiplying the input value by a weighting factor includes a multiplier input by a weighting coefficient, and forming unit weight, comprising a comparator input, and the threshold values of the cell nonvolatile memory with data written therein thresholds switch memory permanent storage device to record therein the weights.EFFECT: technical result is the ability to ensure multi-parameter classification.1 cl, 2 dwg

ethod for prediction of clinical course of early postoperative period in patients with complicated rectal cancer and aid for implementing it // 2567038
FIELD: medicine.SUBSTANCE: individual's clinical and medical history data and laboratory findings are analysed including: sex, age, cancer diagnosis time, type of complication, degree of cancer spread according to TNM classification, stage of cancer, concomitant pathology, previous treatment, complete blood cell counts, clinical urine analysis results, biochemical blood analysis results, ultrasonic data, colonofibroscopy findings, type of surgery. The analysis is performed by means of an artificial neural network, which represents a feedforward network with three neuron layers: 4 neurons in the first layer, 2 neurons in the second layer, and one neuron in the third layer. The neurons in the layers are coupled each to each. Each neuron of the first layer has 28 synapses; neurons of the second layer have 5 synapses; the neuron of the output layer has 3 synapses. As an activation function, the neurons use a logistic function. The neural network divides an output signal space into two groups and relates the value below 0.5 to the unfavourable clinical course of the early postoperative period, and the value exceeding 0.5 - to the group with the favourable clinical course.EFFECT: presented method is simple, cost-effective in comparison with most of approaches existing; it is non-invasive and based on analysing the data obtained as a part of routine preoperative examination and requires no additional diagnostic techniques that in turn is expected to increase reliability, objectification and correct the treatment as may be necessary.2 ex, 2 dwg

Neurocomputer // 2553098
FIELD: physics, computer engineering.SUBSTANCE: invention relates to computer engineering and can be used in designing strapdown inertial reference systems which are part of automatic control systems for highly-manoeuvrable ships, aircraft, space rockets and spacecraft in particular, as well as mobile robotic systems which are characterised by operability in extreme conditions. The device includes a microprogramme control unit, two matrix neuroprocessor units, an operational device, matrix memory, a secondary power source, a communication unit, authorised access memory and an environment sensor.EFFECT: faster matrix computations.23 cl, 20 dwg

Neural network number-to-frequency converter // 2540823
FIELD: electricity.SUBSTANCE: device contains two adders, two "OR" elements, two delay elements, counter, decoder, code memory, four "AND" elements, weight coefficient memory unit, training unit, memory unit for weight coefficients, training unit, multiplier, activation function selection unit.EFFECT: implementation of various functional relations of output frequency and input code, and improvement of ability of the converter to adjust a multiplicative component of errors of sensors.2 tbl, 1 dwg

Simulator for self-forming networks of informal neurons // 2484527
FIELD: medicine.SUBSTANCE: simulator comprises a number of neuron-like elements each of which contains synaptic weight changing units, a summation unit, a comparator, a converter, a random number generator, a multiplier unit, a pulse strobing unit and a prolongation unit controlling the comparator, as well as a summation unit with a regenerative loop.EFFECT: improving the accuracy of the neural network simulation taking into account the real properties of the human and animal cerebral neurons, better understanding of the processes taking place in the prototype neuron, the use of the reproduced signal processing processes in the nerve cells to re-create artificial intelligence systems.4 dwg

Adaptive control device, neuron-like base element and method for operation of said device // 2475843
FIELD: information technology.SUBSTANCE: device includes a plurality of neuron-like base elements (BE) which are grouped into modular units such that outputs of like BE are orthogonal to inputs of other BE and are separated by a dielectric film with a semiconductor layer. The neuron-like BE has a multipoint input for reception and summation of electrical signals, a working output and a signal processing unit, having a threshold voltage former, a comparator and two normalised voltage formers.EFFECT: simple design of the device, broader functional capabilities thereof and a more complex category of problems solved.10 cl, 7 dwg

Neuroprocessor // 2473126
FIELD: information technology.SUBSTANCE: invention can be used in designing computing means for systems for controlling highly manoeuvrable aviation and space-rocket objects where there is need for fast computation of functions, for example, trigonometric functions, used in matrix transformations when solving tasks of forming an inertial coordinate system based on information from angular velocity sensors, and when solving tasks for maintaining operating capacity of computers during changes in parameters of LSI elements due to the effect of natural or artificial ionising radiations. The device has a unit for communication with an on-board computer or which controls a higher level subsystem, a control device, a buffer register, a memory device, multipliers, an adder and an output register.EFFECT: fewer faults in microchips.5 cl, 5 dwg

Predicting properties of underground formation // 2462755
FIELD: physics.SUBSTANCE: method comprises steps for: obtaining seismic data for an area of interest; obtaining an initial seismic cube using said seismic data, wherein the initial seismic cube is a three-dimensional representation of the seismic data; generating a plurality of shifted seismic cubes within the area of interest using said seismic data and a shifting parameter, wherein each of the plurality of shifted seismic cubes is shifted from the initial seismic cube; and wherein the shifting parameter defines a direction and a range that the initial seismic cube should be shifted; generating a neural network using the initial seismic cube, the plurality of shifted seismic cubes, and well log data; and applying the neural network to said seismic data to obtain a model for the area of interest, the model being configured for use in adjusting an operation of the wellsite.EFFECT: high accuracy.20 cl, 19 dwg

Ophthalmic-microsurgical computer local area network for vitreoretinal operations // 2420803
FIELD: information technology.SUBSTANCE: in an ophthalmic-microsurgical computer local area network for vitreoretinal operations, formatting devices are in form of a radial-annular structure consisting of a single set of automated workstations (AWS), which synchronously or asynchronously functioning, processing, converting, transmitting, analysing, synthesising hierarchical structures of an artificial neural network: diagnosis AWS (DAWS), ophthalmic-microsurgical AWS (OMAWS), subsequent operation stages AWS (SOSAWS), component AWS (CAWS), surgeon's operating unit (SOUAWS), with opposite forward and reverse flow of information in between, where each AWS has at least one neural circuit, interconnected identification units (IU), an interpolation unit (INU), an extrapolation unit (EU), which are the neural network converting and transmitting elements (NNCTE), a decision unit (DU), which is the neural network analysis and synthesis element (NNASE).EFFECT: simultaneous improvement of accuracy of determination and quality of identifying diagnoses, determining indications for conducting operations, high selectivity when conducting operations, accuracy in determining the sequence of operations, simulating operations, accuracy in choosing the anaesthetic method, accuracy of providing implants and expendable materials, optimisation of flow of information and necessities during vitreoretinal ophthalmic-microsurgical operations.1 dwg

Neuron simulation method // 2402813
FIELD: physics.SUBSTANCE: neuron simulation method is based on calculation of squares of Euclidean distance from the input vector to each of 2n vertices of a unit n-dimensional cube in weighting units, and multiplication of values inverse to these distance values with components of the target vector respectively, and then summation in an adder and conversion in the activation unit through an activation function.EFFECT: possibility of simulating a neuron of any given Boolean function from a complete set of from n variables.6 dwg, 1 tbl

ultilayer modular computer system // 2398281
FIELD: information technology.SUBSTANCE: multilayer modular computer system has several layers, including a neural network layer, a transport layer and a processor layer, wherein the transport layer contains network controller-router modules, the processor layer contains processor modules, and all the said modules have multiple inputs and outputs connected to each other and connected to the inputs and outputs of the system. The processor modules train neural network domain modules.EFFECT: high decision speed, possibility of grafting layers and modules in each layer during operation of the system with a complex task, high reliability of the computer system.3 cl, 1 dwg

Neuron network for detection of errors in symmetrical system of residual classes // 2374678
FIELD: information technologies.SUBSTANCE: invention may be used for building of modular neural computers, which function in symmetrical system of residual classes. Stated neuron network comprises unit of neuron network of end ring of senior coefficient generation for generalised positional system of numeration, unit of polarity shift, unit of error detection, buses "with errors" and "without errors".EFFECT: reduced hardware complexity.3 dwg

Neurocomputer and neural information processing method // 2351011
FIELD: physics; computer engineering.SUBSTANCE: present invention pertains to neurocomputers. The device has a unit for storing a binary input signal, a logic AND-OR circuit, internal memory unit, unit for generating the output string of codes, a generator of synchronising pulses, control unit, a unit for selecting duration and extracting information, analysis block and a corrector unit.EFFECT: increased rate of operation, providing for the possibility of distinguishing change in state of processed signals, increased noise immunity, possibility of making super-complex neural networks, and simplification of design.9 cl, 1 dwg

Neuron network for dividing numbers which are represented in a system of residual classes // 2318239
FIELD: modular neuro-computing systems.SUBSTANCE: neuron network contains input layer of neurons, at inputs of which residuals of number being divided are received through system of modules, (n-1) neuron networks of finite ring for addition, (n-1) neuron networks of finite ring for multiplication, neuron network for expanding a tuple of numerical system of residues, and as output of neuron network for dividing numbers represented in system of residual classes are outputs of neuron network of finite ring for multiplication and output of neuron network for expansion of tuple of numerical system of residues.EFFECT: expanded functional capabilities, increased speed of division, reduced volume of equipment.1 dwg

Neuron network model // 2309457
FIELD: neuron-like computing structures, possible use as processor for high speed computer systems.SUBSTANCE: device contains artificial neuron network composed of analog neurons, at least one controllable voltage block, a group of long neuron-like nonlinear communication units, each one of which contains serially connected circuit for synchronization and selection of radio impulse envelope, auto-generator with self-suppression circuit, a length of coaxial line, realizing functions of antenna, additional circuit for synchronization and selection of radio-impulse envelope.EFFECT: increased information processing speed due to increased paralleling degree of computing processes.2 dwg

ethod for realizing logical nonequivalence function by a neuron // 2308758
FIELD: neuro-cybernetics, possible use in artificial neuron networks for solving various problems of logical processing of binary data.SUBSTANCE: method for realization of logical nonequivalence function by neuron with two inputs is based on multiplication of input signals with corresponding weight coefficients and summing them, after that the total is transformed in activation block firstly by quadratic transfer function, and then by threshold function at neuron output.EFFECT: realization by one neuron of first order of logical nonequivalence function of two variables.5 dwg, 1 tbl

Neuron network for finding, localizing and correcting errors in residual classes system // 2301442
FIELD: computer engineering, possible use in modular neuro-computer systems.SUBSTANCE: in accordance to invention, neuron network contains input layer, neuron nets of finite ring for determining errors syndrome, memory block for storing constants, neuron nets for computing correct result and OR element for determining whether an error is present.EFFECT: increased error correction speed, decreased amount of equipment, expanded functional capabilities.1 dwg, 3 tbl

Neuron-like element // 2295769
FIELD: cybernetics, possible use as a cell for neuron networks.SUBSTANCE: neuron-like element may be used for realization on its basis of neuron network for solving problems of estimation of functioning of complicated open systems, estimation of degree of optimality of various solutions by ensuring possible construction of model of researched system, both hierarchical and recurrent, with consideration of varying original and working condition of its elements and variants of their functioning, during modeling taking into consideration the level of self-sufficiency of neuron-like elements, susceptibility to effect of external signals, type and errors of setting of their parameters and parameters of input signals, and also provision of given precision of self-teaching of neuron network. Device contains input block, block for setting and normalizing weight coefficients, block for computing parameters of input signals, adder, signals share limiter, block for computing input part of condition, block for setting internal state, block for computing internal part of distance, block for counting distance, memory block, analyzer of state change value, block for determining precision of self-teaching of neuron network, block of determined dependencies, switch, output block, control block, random numbers generator.EFFECT: creation of neuron-like element.2 cl, 1 dwg

The method of signal processing // 2189078
The invention relates to the field of computer engineering and can be used in neural networks

Device optical neural network // 2165644
The invention relates to the field of computer engineering and can be used in neural computers
 
2551091.
Up!