55 results about "Negative weight" patented technology

Data carrying optical signals are subjected to equalization or pulse shaping of the optical signal waveform. A plurality of optical tap signals derived from the optical signal are used to control a modulator operating on an input signal to provide an output signal having the desired waveform. A Mach-Zehnder interferometer having semiconductor optical amplifiers in each arm provides modulation by the effect of cross modulation induced by propagating the respective tap signals through a selected one of the semiconductor optical amplifiers. Various forms of transversal filter are provided by selecting the number of optical taps, assigning positive or negative weights and appropriate delays. The effects of dispersion in optical signals can be mitigated by utilizing optical taps with negative weights to subtract tail portions from the leading and trailing edges of a signal pulse. The invention has application to systems with high bit rates where equalization or pulse shaping in the electrical domain is difficult to implement.

A memory-centric neural network system and operating method thereof includes: a processing unit; semiconductor memory devices coupled to the processing unit, the semiconductor memory devices contain instructions executed by the processing unit; weight matrixes including a positive weight matrix and a negative weight matrix constructed with rows and columns of memory cells, inputs of the memory cells of a same row are connected to one of Axons, outputs of the memory cells of a same column are connected to one of Neurons; timestamp registers registering timestamps of the Axons and the Neurons; and a lookup table containing adjusting values indexed in accordance with the timestamps, the processing unit updates the weight matrixes in accordance with the adjusting values.

PCT No. PCT / US96 / 16332 Sec. 371 Date Feb. 25, 1998 Sec. 102(e) Date Feb. 25, 1998 PCT Filed Oct. 11, 1996 PCT Pub. No. WO97 / 14102 PCT Pub. Date Apr. 17, 1997A synthetic creature (10) in which links (26), representing limbs, are loosely connected by joints (28) to independently interact and implement the physics of everyday life-like actions (18) in a character (16). Elements of mass, positive weight (46), and negative weight (48) are independently handled to mimic the closed-loop control characteristics of real animals. The synthetic creature (10) uses intelligent appearing behaviors (20) built up out of the actions (18) and recursively arranged in a hierarchical behavior tree (72) to cooperate and compete in a run-time computer (202) to give a feeling of autonomy and awareness. Further taught is an authoring environment (200) of tools used to author the character (16), actions (18), and behaviors (20) of the synthetic creature (10), which provide author feedback and simulation testing capability at the various layers (12) of the synthetic creature (10).

A method and system separates components in individual signals, such as time series data streams. A single sensor acquires concurrently multiple individual signals. Each individual signal is generated by a different source. An input non-negative matrix representing the individual signals is constructed. The columns of the input non-negative matrix represent features of the individual signals at different instances in time. The input non-negative matrix is factored into a set of non-negative bases matrices and a non-negative weight matrix. The set of bases matrices and the weight matrix represent the individual signals at the different instances of time.

A network equipment multiple port channel selection method includes evaluating every output port load status of network equipment according to port band width and flow rate load status, to obtain every output port load status evaluating result, in service retransmission, querying every output port load status evaluating result, comparing evaluating result, preference selecting large load-carrying capacity port to make service retransmission. Making weight value analyzing to evaluating result, defining band width factor and flow rate capacity factor value taking or value taking interval to different size and positive-negative weight value, then to make weighing to obtain port load weight value reflecting port load status evaluating result. The present invention makes multiple path outlet channel selection according to port load status weight value, raising channel selection rationality.

A method for estimating locations of devices in a network, the method including: identifying each of the devices; inputting an approximate location of at least one of the devices; constructing a constrain for each of the devices, the constrain comprising at least one of a type of connectivity of the device and a technology characteristic related to the device; deriving an optimization criterion on a location for each of the devices, the criterion assigning a positive weight to each constrain that is satisfied by at least one of an estimated location and the approximate location and assigning a negative weight to each constrain that is not satisfied by at least one of the estimated location and the approximate location; calculating a best fit location for each of the devices that maximizes the optimization criterion; presenting the best fit location for each of the devices to a user, the user either accepting or rejecting each best fit location, the user providing additional information for constructing each constrain associated with a rejected best fit location; and iterating at least one of the identifying, the inputting, the constructing, the deriving, the calculating, and the presenting using the additional information until the user accepts each best fit location.

The invention discloses a dipulse engine with an integrated propellent twining structure and a preparation method. The dipulse engine comprises a combustion chamber shell, a soft interlayer, a first pulse grain assembly, a second pulse grain assembly, an ignition device and a jet pipe, wherein a front opening and a rear opening are formed in the two ends of the combustion chamber shell separately;the soft interlayer comprises a first section and a second section and divides the cavity of the combustion chamber shell into a first cavity, a third cavity and a second cavity; the first cavity isclose to the rear opening; the third cavity and the second cavity are close to the front opening and are formed internally and externally; the first pulse grain assembly comprises a first grain and afirst heat insulation layer; a central groove is formed in the first grain; the second pulse grain assembly comprises a second grain and a second heat insulation layer; the ignition device is arrangedat the front opening, and the top cover body of the ignition device is provided with a second extension part; the two wall surfaces of the second extension part are spliced with the combustion chamber shell and the second heat insulation layer separately; and the jet pipe is arranged at the rear opening. According to the dipulse engine, the second extension part is spliced with the combustion chamber shell and the second heat insulation layer, so that the negative weight is greatly reduced.

A method and system separates components in individual signals, such as time series data streams. A single sensor acquires concurrently multiple individual signals. Each individual signal is generated by a different source. An input non-negative matrix representing the individual signals is constructed. The columns of the input non-negative matrix represent features of the individual signals at different instances in time. The input non-negative matrix is factored into a set of non-negative bases matrices and a non-negative weight matrix. The set of bases matrices and the weight matrix represent the individual signals at the different instances of time.

An input signal, in the form of a sequence of feature vectors, is transformed to an output signal by first storing parameters of a model of the input signal in a memory. Using the vectors and the parameters, a sequence of vectors of hidden variables is inferred. There is at least one vector hn of hidden variables hi,n for each feature vector xn, and each hidden variable is nonnegative. The output signal is generated using the feature vectors, the vectors of hidden variables, and the parameters. Each feature vector xn is dependent on at least one of the hidden variables hi,n for the same n. The hidden variables are related according tohi,n=∑j,lci,j,lɛl,nhj,n-1,where j and l are summation indices. The parameters include non-negative weights ci,j,l, and εl,n are independent non-negative random variables.

A method of performing node-based static timing analysis on a digital network and a program storage device for implementing the method, wherein the method comprises partitioning timing delays in the digital network into portions attributable to a factor of interest and portions attributable to other factors; multiplying the timing delays by different weights based on the factor of interest to produce weighted timing delays; and using the multiplied timing delays to determine a relative impact of the factor of interest on the various paths in the digital network. The method further comprises setting arrival times of timing signals at digital network path start points to zero and identifying digital network paths whose timing delays are dominated by a particular factor of interest. The different weights comprise any of a positive weight, a negative weight, and a zero weight.

The invention discloses a multi-user multi-task mobile edge computing energy-saving method based on heterogeneous energy acquisition. The method comprises the following steps: firstly constructing a problem of minimizing the average energy consumption under the guarantee of time delay on the premise of meeting the stability constraint of a data queue and the stability constraint of an energy queue; in order to solve the random optimization problem, the problem is converted into two deterministic sub-problems, and a convex optimization technology and a linear programming technology are respectively adopted to solve the two deterministic sub-problems. The method does not need priori knowledge of a random process, such as channel state, data arrival and green energy collection. And energy consumption-time delay balance of [O (1 / V), O (V)] is realized. And V is used as a non-negative weight, so that the energy consumption-time delay performance can be effectively controlled.

The invention requests to protect an electroencephalogram classification signal method based on a Gaussian Bernoulli convolution deep belief network. According to the method, a blind source separationalgorithm based on the maximum negative entropy is adopted in the preprocessing stage to remove signal interference of motor imagery electroencephalogram; Selecting frequency and electrode parametersbased on the mutual information; combining the unsupervised learning of the Gaussian Bernoulli restricted Boltzmann machine with the convolutional neural network to carry out feature extraction and classification; A new convolutional deep belief network model based on a Gaussian Bernoulli restricted Boltzmann machine can extract meaningful characteristics from a full-size image through a generated convolutional filter. Compared with the prior art, the method has the advantages that a large number of negative weights are reduced, spatial information can be learned from adjacent image patches more effectively, the accuracy of electroencephalogram category judgment is obviously improved, and the accuracy of electroencephalogram category judgment is greatly improved.

The invention relates to a semi-supervised extreme learning machine classification method based on graph balance regularization. According to the method, an adjacency graph based on label consistencyand an adjacency graph based on information structure consistency are balanced through a non-negative weight value, so that the graph balance is achieved, a Laplace regular term of an optimal graph can be obtained to constrain a model, it is considered that the weight of the label consistency graph is increased when the information consistency graph cannot well describe the structure information of the sample set, and otherwise, the corresponding proportion needs to be reduced. The method comprises the following steps of firstly, constructing an adjacent supervised graph between the training samples through the label consistency of the samples; and combining with a semi-supervised graph based on the sample information consistency to constrain the output of the model, changing the capability of describing data distribution by reasonably adjusting the proportion of the graph, and obtaining an optimal output weight vector after obtaining an optimal adjacent graph. The method has a wide application prospect in the electroencephalogram signal processing systems and the brain-computer interface systems.

The invention belongs to the technical field of terminal ballistic tests, and particularly relates to a subcaliber bullet shooting auxiliary device. A 30-40g cylindrical fragment 1100-1700m / s acceleration bullet support comprises a bullet support body and a bottom pusher (12) in threaded connection. Speed-free gap loading of a 30-40g large mass cylindrical fragment in a speed range between 1100-1700m / s by a ballistic cannon is realized, accuracy of a loading speed of the large mass fragment is improved, and a test cost is reduced; stable separation between the fragment and the bullet support is realized, the stability of the fragment fly is improved, and well target hitting gesture is realized; chamber pressure directly acts on the bottom pusher; an attack angle when the fragment flies is avoided and damage to a bore is reduced; negative weight of the bullet support is reduced by the material of the bullet support and structural strength of the bullet support is ensured; an inner diameter of a placing position loading the bullet support is changed, the device is suitable for the cylindrical fragments with different diameters in a mass range of 30-40g and well universality is realized.

A method of performing node-based static timing analysis on a digital network and a program storage device for implementing the method, wherein the method comprises partitioning timing delays in the digital network into portions attributable to a factor of interest and portions attributable to other factors; multiplying the timing delays by different weights based on the factor of interest to produce weighted timing delays; and using the multiplied timing delays to determine a relative impact of the factor of interest on the various paths in the digital network. The method further comprises setting arrival times of timing signals at digital network path start points to zero and identifying digital network paths whose timing delays are dominated by a particular factor of interest. The different weights comprise any of a positive weight, a negative weight, and a zero weight.

A shortest annular path navigation method for an automated guided vehicle and the automated guided vehicle are disclosed. The method includes generating an adjacent matrix graph including actual formation of a road segment in a workplace according to node coordinate information in the workplace; finding out a shortest one-way annular path required by a task in the adjacent matrix graph; arranginga millimeter-wave radar in the workplace; measuring, by a gyroscope, the velocity, direction and acceleration of the automated guided vehicle and feeding the velocity, direction and acceleration backto a terminal controller in real time; performing calculation, by the terminal controller, according to feedback information from the millimeter-wave radar; and controlling, by an industrial personalcomputer, a drive module to complete position correction. A one-way annular path planning manner is adopted and a Dijkstra algorithm is adopted to achieve shortest path searching between two points ina non-negative weight graph in a shortest path algorithm, and the path planning efficiency of the automated guided vehicle is increased, thus simplifying path planning. In addition, by utilizing real-time correction through the millimeter-wave radar and the gyroscope, that the automated guided vehicle is ready to reach a target position through the shortest path is ensured.

The invention discloses a shortest annular path navigation method for an automatic guide transport cart and the transport cart. A shortest one-way annular path required by a mission is found out in anadjacent matrix through a Dijkstra algorithm; the shortest one-way annular path determined in step 2 is input into a centrally-controlling machine, and an industrial personal computer controls a driving module to drive the automatic guide transport cart; a binocular vision sensor and a high-speed image collecting system collect information, and determine the position of the cart in a ground coordinate system; the centrally-controlling machine sends position correcting information to the industrial personal computer according to the actual position and the movement state. By means of the one-way annular path planning method, the Dijkstra algorithm is adopted for achieving the shortest path search between two points in a non-negative-weight map in a shortest path algorithm, the path planning efficiency for the automatic guide transport cart is improved, and thus the difficulty of path planning is reduced; moreover, through real-time correction of the binocular vision sensor and the high-speed image collecting system, it is guaranteed that the automatic guide transport cart is prepared to arrive at the target position through the shortest path.

The invention discloses an image retrieval method based on multi-view local reconstruction preserving embedding. The method comprises: extracting features of an image, to obtain feature matrixes of multiple viewing angles, using a k-nearest neighbor sample of samples in an original space to reconstruct the sample, through a reconstruction weight matrix to describe similarity among the samples andneighboring samples, mapping similar structure features in the original space to a low-dimensional subspace, constructing a similarity matrix according to the reconstruction weight matrix of each viewing angle, at that same time, constructing a set of non-negative weight coefficients to represent different weight contribution of the same sample from different viewing angles, introducing auxiliarycoefficients to fuse multi-view feature information, and through an iterative optimal solution method, obtaining common low-dimensional subspace embedding representation of multiple viewing angles, finally, performing similarity measurement sorting on a to-be-retrieved image and all images, to obtain a retrieved result. The method gives full consideration to consistency and complementarity information of multiple viewing angles, optimally solves embedding representation of low-dimensional subspace of an image, and improves performance of image retrieval.

The invention relates to a text processing method and system based on meaning group division, and the method comprises the steps: obtaining an article to be analyzed in semantic tendency, wherein thearticle comprises paragraphs, the paragraphs comprise sentences, the sentences are divided into continuous language segments expressing a single meaning, ethe continuous language segments serve as a semantic meaning group, and the word segmentation of the semantic meaning group is carried out, and candidate words are obtained; obtaining a sentiment word library, allocating a tendency weight to each sentiment word in the word library, constructing a sentiment word list, retrieving candidate words in the sentiment word list, and extracting sentiment words corresponding to the candidate words astendency words of sentences; analyzing degree adverbs and negative words in front of the tendency words respectively; endowing the tendency words with degree weights and negative weights, and multiplying the negative weights, the degree weights and the tendency weights of the tendency words to obtain meaning group tendency components of the semantic meaning groups; and collecting the tendency component of each meaning group in the sentence to serve as a sentence tendency component, and obtaining a semantic tendency component of the article according to the sentence tendency component to serveas a semantic tendency analysis result of the article.

The invention relates to the field of intelligent meter reading and wireless communication, in particular to a wireless network routing algorithm for an intelligent meter reading system. According to the technology, a redundant routing distribution algorithm is provided through research on a wireless communication network and analysis on the importance of the routing algorithm to the network communication performance; the algorithm combines a Dijkstra algorithm, and an R-Dijkstra algorithm suitable for redundant routing distribution is designed. The R-Dijkstra algorithm defines a wireless network as a non-negative-weight connected graph, redundancy parameters and the path optimization target are introduced, and the transmission path can be selected; the algorithm steps and implementation processes are further provided. The application example shows that the algorithm can meet the requirement of wireless intelligent meter reading routing. The method has the advantages of being high in reliability, small in calculated amount, easy to program and implement and the like.

An input signal, in the form of a sequence of feature vectors, is transformed to an output signal by first storing parameters of a model of the input signal in a memory. Using the vectors and the parameters, a sequence of vectors of hidden variables is inferred. There is at least one vector hn of hidden variables hi,n for each feature vector xn, and each hidden variable is nonnegative. The output signal is generated using the feature vectors, the vectors of hidden variables, and the parameters. Each feature vector xn is dependent on at least one of the hidden variables hi,n for the same n. The hidden variables are related according to h i , n = ∑ j , l   c i , j , l  ɛ l , n  h j , n - 1 , where j and l are summation indices. The parameters include non-negative weights ci,j,l, and εl,n are independent non-negative random variables.

A robust human face recognition method based on weighted mixing norm regression relates to the field of pattern recognition, aiming to improving the recognition precision of the face image loss. The method comprises the following steps: a) selecting ni samples as a training set for each target object, determining a dictionary matrix X, and inputting a test face image Y; b) establishing a WMNR model, optimizing the model, and obtaining a corresponding characteristic weight matrix W, a coefficient vector a and a non-negative weight vector s; c) and substituting the W, a and s obtained in the step b into a classification reconstruction error minimization model, and finally realizing the face image recognition. According to the invention, the model is optimized by using a value self-adaptationmechanism and an effective weighted fast iterative algorithm, has the advantages of being high in operation efficiency, high in accuracy, and is very suitable for face detection and face classification under the condition of image damage.

A method for the determination of the reputational rating of natural or legal persons, including the following steps: populating a database with certain pieces of information relating to the natural or legal persons, the certain pieces of information being pertinent to prove evidence or facts object of the evaluation of the reputational rating, assigning a positive or negative weighted value to each of the certain pieces of information, proportional to a suitability factor of the evidence or facts to contribute to the determination; determining the reputational rating of any of the natural or legal persons by summing up all the contributions derived from the weighted values relating to the natural or legal persons.

The present invention provides a Bayesian model-based pseudo-base station identification method. The method at least comprises the steps of acquiring, by using a mobile communication terminal, a C1 criterion value and a C2 criterion value of each base station; searching a Bayesian model for feature values that are equal to the C1 criterion value and the C2 criterion value, acquiring a forward weight value and a negative weight value of the feature values, and using the forward weight value and the negative weight value as a forward weight value and a negative weight value of the C1 criterion value and the C2 criterion value; calculating a first probability integration value of the base station according to the forward weight value; calculating a second probability integration value of the base station according to the negative weight value; and determining, according to the first probability integration value and the second probability integration value, whether the base station is a pseudo-base station. The pseudo-base station is identified by using the Bayesian model, so that the identification accuracy rate of the pseudo-base station is improved, and the method lays a reliable foundation for further security processing operation on the pseudo-base station.

A system configured for calculating at least a parameter relating to the determination of the reputational rating of a natural or legal person includes a number of electronic modules capable of populating a database with data derived from pieces of information relating to the natural or legal person, the data being adapted to prove evidence or facts object of the evaluation of the parameter of reputational rating, assigning a positive or negative weighted value to each of the data, proportional to a suitability factor of the evidence or facts to contribute to the determination; determining the parameter of reputational rating of the natural or legal person on the basis of a sum of all the contributions derived from the weighted values relating to the natural or legal person.

The invention discloses a Lagrange function based least-squares multi-objective optimization method which includes the steps of step 1, converting multi-objective programming into single-objective programming; step 2, fitting established models to obtain a fitting error; step 3, obtaining an optimal solution model; step 4, interposing a weighted factor beta into a target function Q(n); step 5, obtaining an extended function; step 6, removing non-negative weights from the optimal solution model; and step 7, calculating an optimal weighting coefficient by least-squares criterion. By the method, a system can evolve to the equilibrium state no matter how the initial state of the system is, a neural network can converge a variable into a unique global optimal solution no matter how an initial value of an optimized variable is.

A method and system for spectrum data analysis. The method comprises the steps of collecting a spectrum of an unknown material; providing a set of element data templates; calculate optical least squares weights for the element data templates in approximating the spectrum; removing one or more of the templates having negative weights in approximating the spectrum; and re-calculating an approximation of the spectrum with said one or more templates removed.

The invention discloses a manifold learning and gradient lifting model-based picture multi-label classification method. Constructing a weighted graph from the training data set, obtaining a non-negative weight matrix by solving the first minimization model, establishing a second minimization model according to the weighted graph, solving to obtain a reconstructed label matrix, constructing the training data set according to the reconstructed label matrix, training a binary correlation model, and predicting to obtain a label matrix; and establishing a regression device minimization solution forthe feature vector matrix of the picture, enhancing the feature vector matrix by using an iterative prediction result matrix, constructing a data set by combining a negative gradient matrix, trainingand learning to obtain weak regression devices, summing all the weak regression devices to obtain a final regression device, and processing and judging a pre-to-be-tested picture. According to the method, the multi-label classification prediction performance of the picture can be improved by fully utilizing the correlation between the partial multi-label data of the picture, the disambiguation ofthe partial label data can be realized, the accuracy and the robustness are improved, and the performance of the method is superior to that of the existing partial multi-label method of the picture.

The invention discloses a method for weighing samples on automatic industrial analysis instruments. The steps are as follows: S1: placing all sample loading containers; S2: lowering the sample plate, and clearing the weighing balance; S3: filling the sample loading containers Add the sample into the sample, weigh it directly with a weighing balance to get the sample weight, and record it as the sample weight M 1 , the sample tray rises; S4: Take the sample container out of the weighing balance, and the absolute value of the negative weight value obtained on the weighing balance at this time is the weight of the sample container, which is recorded as M 0 ;S5: M obtained by weighing 0 and M 1 , to participate in the calculation of indicators. The invention has the advantages of simplifying steps, reducing working time, improving overall efficiency and the like.