195 results about "Multilayer perceptron" patented technology

A method of producing noise-free images is disclosed. The method includes using machine learning incorporating a filter to output filter parameters using the training images. The machine learning may include training a neural network. The filter parameters are applied to Monte Carlo rendered training images that have noise to generate noise-free images. The training may include determining, computing and extracting features of the training images; computing filter parameters; applying an error metric; and applying backpropgation. The neural network may be a multilayer perceptron. The machine learning model is applied to new noisy Monte Carlo rendered images to create noise-free images. This may include applying the filter to the noisy Monte Carlo rendered images using the filter parameters to create the noise-free images.

The invention relates to the technical field of traffic control and artificial intelligence and provides a traffic signal self-adaptive control method based on deep reinforcement learning. The method includes the following steps that 1, a traffic signal control agent, a state space S, a motion space A and a return function r are defined; 2, a deep neutral network is pre-trained; 3, the neutral network is trained through a deep reinforcement learning method; 4, traffic signal control is carried out according to the trained deep neutral network. By preprocessing traffic data acquired by magnetic induction, video, an RFID, vehicle internet and the like, low-layer expression of the traffic state containing vehicle position information is obtained; then the traffic state is perceived through a multilayer perceptron of deep learning, and high-layer abstract features of the current traffic state are obtained; on the basis, a proper timing plan is selected according to the high-layer abstract features of the current traffic state through the decision making capacity of reinforcement learning, self-adaptive control of traffic signals is achieved, the vehicle travel time is shortened accordingly, and safe, smooth, orderly and efficient operation of traffic is guaranteed.

The invention discloses an image subtitle generation method and system fusing visual attention and semantic attention. The method comprises the steps of extracting an image feature from each image tobe subjected to subtitle generation through a convolutional neural network to obtain an image feature set; building an LSTM model, and transmitting a previously labeled text description correspondingto each image to be subjected to subtitle generation into the LSTM model to obtain time sequence information; in combination with the image feature set and the time sequence information, generating avisual attention model; in combination with the image feature set, the time sequence information and words of a previous time sequence, generating a semantic attention model; according to the visual attention model and the semantic attention model, generating an automatic balance policy model; according to the image feature set and a text corresponding to the image to be subjected to subtitle generation, building a gLSTM model; according to the gLSTM model and the automatic balance policy model, generating words corresponding to the image to be subjected to subtitle generation by utilizing anMLP (multilayer perceptron) model; and performing serial combination on all the obtained words to generate a subtitle.

The invention belongs to the technical field of computers, and discloses a machine translation method and system based on a generative adversarial neural network. The method comprises the following steps that: on the basis of an original machine translation generation network, a discrimination network which generates network countermeasure with the original machine translation generation network is imported; a translation used for judging a target language is from a training parallel corpus and is a network machine translation result of the original machine translation generation network; and the discrimination network adopts a multi-layer sensor feedforward neural network model to realize binary classification. The system comprises the discrimination network, a generation network, a mono-lingual corpus and a parallel corpus. While manually annotated bilingual parallel corpus resources are fully utilized, and mono-lingual corpus resources also can be fully utilized to carry out semi-supervised learning; and the mono-lingual corpus resources are very rich and can be easily obtained, and the problem that required training corpora required by the neural network machine translation model are not sufficient is solved.

The invention provides a sentiment classification method for generating a model deep-convinced-degree network on the basis of the probability of depth study. According to the technical scheme of the method, a plurality of Boltzmann machine layers are stacked, namely, output of this layer is used as input of the next layer. By the adoption of the mode, input information can be expressed in a grading mode, and abstraction can be conducted layer by layer. A multi-layer sensor containing a plurality of hidden layers is the basic study structure of the method. More abstract high layers are formed through combining the characteristics of lower layers and are used for expressing attribute categories or characteristics, so that the distribution type character presentation of data can be discovered. The method belongs to monitoring-free study, and a mainly-used model is the deep-convinced-degree network. The method enables a machine to conduct characteristic abstract better so as to improve the accuracy of sentiment classifications.

The invention discloses a deep convolutional neural network-based human face occlusion detection method. The method comprises the steps of performing block segmentation on an input image to obtain a target pre-selected region; constructing a first deep convolutional neural network, training the first deep convolutional neural network comprising a first deep convolutional network and a first multilayer perceptron connected with the first deep convolutional neural network to obtain required parameters, extracting features of the target pre-selected region, and performing classification; predicting the position of a human head through a second multilayer perceptron according to the extracted features; filtering the credibility of a classification type which is the human head and the predicted position of the human head through non-maximum suppression to remove an overlapped duplicate detection box; and obtaining a human head block in combination with original image segmentation, constructing a multi-task learning policy-based second deep convolutional neural network, and judging whether the left eye, the right eye, the nose and the mouth of the human head block are occluded or not. According to the method, the occluded human face can be accurately detected and the specific occluded part of the human face can be judged; and the method is mainly used for crime pre-warning of videos of a camera in front of an automatic teller machine.

The invention discloses a single-image three-dimensional reconstruction method based on a multi-stage neural network. The three-dimensional shapes in the existing three-dimensional shape set are rendered from multiple angles to obtain a training image set, and the training point cloud is obtained at the surface acquisition points thereof. A point cloud generation network is constructed, an image encoder is constructed by using depth residual network to extract image information, and a dual-branch primary decoder is constructed by using deconvolution network and full-connection network to generate initial point cloud. A point cloud refinement network is constructed, a point cloud encoder is constructed using a posture transformation network, a multilayer perceptron and a maximum pool function, an image encoder is constructed using a depth residual network, and an image is constructed using a full connection layer. A point cloud coupler and an advanced decoder generate a fine point cloud; Training the point cloud generation network and pre-training and fine-tuning the point cloud refinement network; The input image is reconstructed by using the trained model to obtain the 3D point cloud, and the surface mesh is reconstructed to generate the 3D shape represented by polygonal mesh.

Methods and apparatus for color correction of color image data obtained by an electronic camera determine a correction to data representative of the color image based upon an estimated illuminant using a neural network, multilayer perceptron models and / or coactive neuro-fuzzy inference system models, and apply the correction to the data representative of the color image. Data representative of the color corrected data may be recorded or transmitted. A method of recording image data obtained by an electronic camera captures a color image, generates data representative of the image, estimates an illuminant for the captured color image, generates data representative of the estimated illuminant and records the data representative of the image with the data representative of the estimated illuminant. A method of transmitting image data obtained by an electronic camera captures a color image, generates data representative of the image, estimates an illuminant for the captured color image, generates data representative of the estimated illuminant and transmits the data representative of the image with the data representative of the estimated illuminant.

The invention discloses a dependency tree and attention mechanism-based attribute sentiment classification method. The method comprises the steps of selecting a smallest sub-tree part comprising givenattributes based on a dependency tree analysis result of a whole text, and taking clauses of the part as representations of context information of the attributes; then performing modeling on contextsof sentences and contexts of the attributes by utilizing two bidirectional threshold circulation units to obtain two feature representation matrixes fixed in size; obtaining feature representations of the text and specific attributes by utilizing an attention mechanism; and finally performing sentiment polarity classification of the specific attributes by utilizing a multilayer perceptron. The classification method proposed by the invention can extract different attribute feature information for different attributes in the same text, and is high in classification accuracy.

An automatic speech recognition system for recognizing a user voice command in noisy environment, including: matching means for matching elements retrieved from speech units forming said command with templates in a template library; characterized by processing means including a MultiLayer Perceptron for computing posterior templates (P(Otemplate(q))) stored as said templates in said template library; means for retrieving posterior vectors (P(Otest(q))) from said speech units, said posterior vectors being used as said elements. The present invention relates also to a method for recognizing a user voice command in noisy environments.

The invention discloses a living body detection method, a computer device and a computer readable storage medium. In the living body detection method, a multi-layer perceptron is trained by using a preset training set to determine a multi-layer perceptron model, continuous N-frame human face images to be detected is obtained, an intermediate frame human face image of the consecutive N-frame humanface images is switched from a first color space to a second color space, textural features of the intermediate frame human face image and dynamic pattern features of the continuous N-frame human faceimages are extracted, fusion features are acquired by fusing the textural features and the dynamic pattern features, the multi-layer perceptron model is used to perform feature mapping on the fusionfeatures, mapping features are output and normalized, a predicted probability value of a living body tag and a predicted probability value of a non-living body tag are obtained, and then the continuous N-frame human face images are determined as living body or non-living body human face images. The fusion features include the textural features and the dynamic pattern features, and thus the recognition accuracy and safety of living body detection can be improved.

The invention discloses a method of modulating a variable pulse amplitude position and improving an error rate for a visible light communication system, belonging to the field of visible light wireless communication technologies. According to the method, modulation of the variable pulse amplitude position is utilized in a modulation module of a signal transmitting end, a modulation method is multi-system modulation capable of combining pulse amplitude position modulation and pulse width modulation, transmission of communication data is guaranteed while a system dimming function is provided, improvement of the communication quality mainly includes that the transmission rate can be improved under the condition of the same symbol rates, the advantages of high multi-system modulation bandwidth efficiency and high pulse position modulation power efficiency are combined, the spectral efficiency can be effectively improved, and compared with the other multi-system modulation manner, the complexity is lower. In order to further improve the error rate of the visible light communication system, the method utilizes a channel balancer based on a feed-forward back propagation multilayer perceptron on a receiving end, so that the error rate is effectively improved, and the communication quality is further improved.

Embodiments of the present invention provide a three-dimensional point cloud marking method and apparatus based on a fusion voxel. The method comprises the following steps: the data set of the three-dimensional point cloud is voxelized and voxel features in the voxels are extracted based on the processing results to form a first voxel feature matrix; the first voxel feature matrix is used as the input of the three-dimensional convolution neural network to calculate the multi-scale feature of the voxel, and the multi-scale feature is fused in series to obtain the second voxel feature matrix. The first voxel feature matrix is used as the input of the three-dimensional convolution neural network to calculate the multi-scale feature of the voxel. Based on the feature interpolation algorithm, the voxel features in the second voxel feature matrix are extended to the points in the three-dimensional point cloud data set to obtain the point cloud feature matrix. The feature matrix of point cloud is inputted into the multilayer perceptron to mark the attributes of three-dimensional point cloud. The invention can realize fine classification and recognition point by point, so as to further improve the performance of point cloud marking.

The invention is an add-on implementation of a stabilized association memory matrix system to an existing convolutional neural network framework. This invention emulates the intra-action and the inter-action of the cognitive processes of the (logical) left-brain and (intuitive) right-brain. The invention is a numerically stable soft-ware based implementation that (1) reduces the long training times, (2) reduces the execution time, and (3) produces intralayer and interlayer connections. The implementation of this joint processing architecture is designed to take an existing hierarchy of stepped based processes, add next to it a parallel hierarchy of associative memory processes, and then connect the two processes by another set of associative memory processes. Or, the stepped-based process may be replaced with additional associative memory processes to enhance the emulation of several bidirectional intralayer and interlayer cognitive process communication. In addition, the invention can be used as a neural network layer compression tool that takes in a multilayer perceptron, also known as a multilayer neural network, and outputs a single layer perceptron. The final construction can be visualized as two vertical rails connected with a set of horizontal rungs which motivates the name to this invention: J. Patrick's Ladder: A Machine Learning Enhancement Tool.

A method and apparatus of training a neural network. The method and apparatus include creating a model for a desired function as a multi-dimensional function, determining if the created model fits a simple finite geometry model, and generating a Radon transform to fit the simple finite geometry model. The desired function is fed through the Radon transform to generate weights. A multilayer perceptron of the neural network is trained using the weights.

A method for calculating a neuron layer of a multi-layer perceptron model that includes a permanently hardwired processor core configured in hardware for calculating a permanently predefined processing algorithm in coupled functional blocks, a neuron of a neuron layer of the perceptron model being calculated with the aid of an activation function, the activation function corresponding to a simplified sigmoid function and to a simplified tank function, the activation function being formed by zero-point mirroring of the negative definition range of the exponential function.

The invention discloses an intelligent auxiliary diagnosis method based on deep learning and ensemble classification. Through named entity identification and relation extraction, entities and attributes in the chief complaint and the current medical history are accurately extracted, and invalid information is removed. In the label topic model, the position weight of the feature words is added, andthe proportion of the feature words at the key position is increased. Adjusting parameters are added into a loss function of a multilayer perceptron model, so that the problem caused by uneven sampledistribution is solved. For the same sample, the classification boundaries obtained by different classification methods are different, so that the label topic model and the multi-layer perceptron model are integrated by adopting a stacking integration method, and the disease prediction accuracy is improved.

The invention relates to the technical field of hypertension prevalence forecasting preformed according to personal basic information and healthy information during the medical process, in particular to a method for using a neural network to forecast hypertension. The method for using the neural network to forecast the hypertension includes following steps: (1) finding out dangerous factors which influence the hypertension; (2) extracting health survey data which influences the hypertension; (3) confirming dangerous factors which really influences the hypertension; (4) collecting health information survey data of the dangerous factors which really influences the hypertension; (5) screening the data; (6) performing standardization processing on valid data; (7) building an MLP (multilayer perceptron) model of a BP (back propagation) neural network, and forecasting whether a person is suffered from the hypertension or not through the MLP model of the BP neural network; (8) comparing a forecasted result with an actual situation whether the person is suffered from the hypertension so as to obtain probability details in forecasting of the situation that the person is suffered from the hypertension and the other situation that the person is not suffered from the hypertension, performed through the method for using the neural network to forecast the hypertension. The method for using the neural network to forecast the hypertension provides a scientific basis for old people to prevent hypertension disease in advance, and enables the old people to early discover, early interfere and early cure the hypertension.

The invention relates to a non-model self-adaptive controller. The controller consists of two forward neural networks which are connected in series, wherein the front neural network is a network controller MLPc; and the back neural controller is a network simulator MLPo. The output of the network controller is taken as the input of the network simulator and a controlled object; a closed loop reverse feedback is arranged between the output of the controlled object and the input of the network controller; and the relation between the output of the network simulator and the output of the controlled object is set as the subtraction relation. The network simulator consists of a time lag multilayer perceptron (TDMLP) and is used for simulating the industrial process; and the network controller which is also a TDMLP is used for exerting the optimal control on the simulator in the precondition of considering model estimation errors. The non-model self-adaptive controller has the advantages of high control accuracy, high system response speed, and simplified network internal mechanism.

The present invention discloses a method for elevator fault identification by employing multilayer perceptrons. The method comprises the steps of: collecting related data of elevator faults, and establishing a database; establishing multilayer perceptrons for features of elevator fault identification; reading elevator fault data for training of a neural network from the database to perform training and learning of the neural network, and after the training of the neural network is completed, using the neural network to perform fault identification of the actual data of the elevator; and finally, outputting one or more than one fault codes with large a probability for maintenance of field staff according to a total quantity control principle. Improvement of safety can be achieved through remote update of a weight when part of the fault identification algorithm needs upgrading. The method provided by the invention can remote grade the fault identification algorithm through adoption of elevator state parameters and analysis of fault reasons so as to maintain rapidly determined faults and have a great meaning for guaranteeing the masses' life safety.

The invention discloses a text classification method and system based on label information and text characteristics and a medium, and the method comprises the steps: a characteristic extraction step:extracting a first text characteristic from a given text, and extracting a first label characteristic from a label corresponding to the given text; a feature fusion step: performing feature fusion onthe first text feature and the first tag feature to obtain a fused third feature; a model training step: inputting the fused third features into a multi-layer perceptron to train the multi-layer perceptron so as to obtain a trained multi-layer perceptron; text classification step, extracting a second text feature from the to-be-classified text, and extracting a second label feature from the labelcorresponding to the text to be classified, carrying out feature fusion on the second text feature and the second label feature to obtain a fused fourth feature, inputting the fused fourth feature into the trained multilayer perceptron for classification, and outputting a classification result.

The invention discloses a deep calculation model for aero-engine gas circuit fault diagnosis, which belongs to the technical field of computers and comprises the following steps: (1) carrying out segmentation preprocessing on continuous aero-engine performance data; (2) extracting spatial features of the data object blocks by using a convolutional neural network; (3) learning local nonlinear features in the data object block by using a multilayer perceptron; (4) extracting long-term dependence features in the data object blocks by using a recurrent neural network; and (5) training the deep hybrid model by using an end-to-end method, i.e., combining an error back propagation algorithm. According to the embodiment of the invention, characteristics of an aero-engine model are considered, spatial characteristics, nonlinear characteristics and long-term dependence characteristics of gas path performance data generated by the aero-engine model are fused, an aero-engine gas path fault diagnosis model is constructed, and a joint error back propagation algorithm is designed for end-to-end training of a deep hybrid model by taking a classification task as a target.

The invention discloses a short video recommendation method based on video content understanding and user dynamic interest, and the method comprises the steps of firstly, extracting deep visual features of a video through a deep learning technology, extracting an audio file from the video, and extracting auditory features; fusing the video features, the social features and the user features by using PCA dimension reduction, data standardization and other technologies to obtain deep fusion features for feature representation of historical behaviors of the user; secondly, extracting the influence of historical behaviors on the current interest by using a self-attention mechanism, and learning an interest evolution path on the candidate video by using a recurrent neural network to obtain accurate dynamic interest of the user; and finally, utilizing a multilayer perceptron to carry out click probability prediction and recommendation on the video candidate set. The method is applied to thepersonalized recommendation of the short video, and the recommendation accuracy can be effectively improved by adopting the technical scheme of the invention.

The invention discloses a cancer medical image data processing method, system, device and a storage medium. The method comprises the following steps: inputting a to-be-processed cancer medical image into a pre-established focus area segmentation model, obtaining a focus area image identification result output by the focus area segmentation model, filtering processing of different standards is carried out on the focus area image cut from the cancer medical image for multiple times, channel stacking is carried out, a channel stacking result is input into a pre-established feature extraction model, and a feature screening algorithm is used for screening out target features from the obtained pathological feature vectors; inputting the target characteristics into a multilayer perceptron to obtain cancer biomarker state index values output by the multilayer perceptron, and the like. According to the method, the focus area on the cancer medical image can be automatically identified, and the focus area image is cut out, so that the identification and cutting efficiency and accuracy are greatly improved. The method is widely applied to the technical field of image processing.

A method and apparatus of training a neural network. The method and apparatus include creating a model for a desired function as a multi-dimensional function, determining if the created model fits a simple finite geometry model, and generating a Radon transform to fit the simple finite geometry model. The desired function is fed through the Radon transform to generate weights. A multilayer perceptron of the neural network is trained using the weights.

The invention provides an article recommendation method based on generalized nerve attention, and relates to the technical field of information processing. According to the method, a generalized matrix factorization model GMF and a nerve attention similarity model NAIS are combined to establish a generalized nerve attention recommendation model GNAS; the model is optimized by using an attention mechanism integrated by GMF and an MLP (Multilayer Perceptron) in the model; and after the model is optimized, the preference degree of the user to the target object is predicted through the optimized generalized neural attention recommendation model, and a personalized recommendation list is generated for the user. Potential hobbies and interests of users are mined, and interpretability and diversity of a recommendation system are improved; secondly, an attention mechanism combining a GMF model and an MLP model is adopted to estimate the weight of each historical article when the favorite degree of the target article is predicted, the recommendation accuracy is greatly improved with low time cost, and articles more conforming to the interest of the user are recommended to the user.

The invention relates to the construction of a convolution neural network, and aims at solving problems that a conventional neural network algorithm exerts higher requirements for the processing speed of hardware and cannot be applied in electric household appliances. The method comprises the following steps: determining an interconnection structure among nerve cell nodes, and determining the mutual relation between numerical calculation feedback and forwarding calculation transfer among the nerve cell nodes; building a multilayer perceptron model through employing a reverse propagation neural network, wherein the structure of the reverse propagation neural network comprises a forwarding propagation algorithm; operating a reverse propagation algorithm and the forwarding propagation algorithm at the same time, and carrying out training and calculation of the convolution neural network. The method is suitable for mode recognition of household electric appliances.

The invention discloses a state space probabilistic multi-time sequence prediction method based on a graph neural network, and the method comprises the steps: (1) obtaining a multi-time sequence, carrying out the preprocessing of the time sequence to construct a training set, and constructing a graph structure; (2) constructing a generation model for generating prior distribution and a time sequence of the hidden state according to a graph neural network and a multilayer perceptron, and constructing an inference network for generating approximate posteriori distribution of the hidden state according to the graph neural network and a recurrent neural network; (3) constructing a loss function according to the prior distribution and the approximate posteriori distribution of the implicit state, and optimizing a generation model and deducing parameters of the network by taking maximization of the loss function as a target; and (4) during application, obtaining the hidden state estimation of the to-be-predicted sequence at the latest moment by utilizing the inference network, then calculating to obtain the prior distribution of the hidden state by utilizing the generation model according to the hidden state estimation at the latest moment, and then calculating to obtain a time sequence observation estimation value according to the prior distribution of the hidden state.

The invention discloses a solenoid valve fault diagnosis device based on feature extraction and a multilayer perceptron and method. The method comprises the following steps: starting current and working voltage are acquired in view of solenoid valve training samples, the response time of the starting current, the stabilization time, the local maximum, the local maximum integral, the local minimumand the local minimum integral are extracted, and together with the working voltage, eigenvectors are formed; with the eigenvectors of the solenoid valve training samples as input and with a fault type as output, the multilayer perceptron is trained, and a solenoid valve fault diagnosis model is obtained; and the eigenvector of a to-be-detected solenoid valve is acquired based on the same method of acquiring the eigenvector of the solenoid valve training sample, the eigenvector is inputted to the solenoid valve fault diagnosis model, and the solenoid valve fault diagnosis model performs faultdetection on the to-be-detected solenoid valve. The solenoid valve eigenvector can better explain the starting current waveform of the electromagnetic valve and enhance the accuracy of the diagnosis of the electromagnetic valve fault, and can be widely applied to the fault diagnosis of the electromagnetic valve.