A coding amount estimation device includes: a feature vector generation unit that generates a feature vector on the basis of a feature map generated by an estimation target image and at least one filter set in advance; and a coding amount evaluation unit that evaluates a coding amount of the estimation target image on the basis of the feature vector.

Methods, and systems, including computer programs encoded on computer storage media for generating compressed representations from a co-occurrence matrix. A method includes obtaining a set of sub matrices of a co-occurrence matrix, where each row of the co-occurrence matrix corresponds to a feature from a first feature vocabulary and each column of the co-occurrence matrix corresponds to a feature from a second feature vocabulary; selecting a sub matrix, wherein the sub matrix is associated with a particular row block and column block of the co-occurrence matrix; assigning respective d-dimensional initial row and column embedding vectors to each row and column from the particular row and column blocks, respectively; and determining a final row embedding vector and a final column embedding vector by iteratively adjusting the initial row embedding vectors and the initial column embedding vectors using the co-occurrence matrix.

The present disclosure provides a visual perception method and apparatus, a perception network training method and apparatus, a device and a storage medium. The visual perception method recognizes the acquired image to be perceived with a perception network to determine a perceived target and a pose of the perceived target, and finally determines a control command according to a preset control algorithm and the pose, so as to enable an object to be controlled to determine a processing strategy for the perceived target according to the control command. According to the perception network training method, acquire image data and model data, then generate an edited image with a preset editing algorithm according to a 2D image and a 3D model, and finally train the perception network to be trained according to the edited image and the label.

A computer device extracts local features of sample images based on a first part of a convolutional neural network (CNN) model. The sample images comprise a plurality of images taken at the same place. The device; aggregates the local features into feature vectors having a first dimensionality based on a second part of the CNN model. The device obtains compressed representation vectors of the feature vectors based on a third part of the CNN model. The compressed representation vectors have a second dimensionality less than the first dimensionality. The device trains the CNN model, and obtains a trained CNN mode satisfying a preset condition in accordance with the training.

A signal analysis device includes a memory and processing circuitry coupled to the memory and configured to obtain, for a spatial covariance matrix R.sub.j (j is an integral number equal to or larger than 1 and equal to or smaller than J) for modeling spatial characteristics of J (J is an integral number equal to or larger than 2) source signals that are present in a mixed manner, a simultaneous decorrelation matrix P as a matrix in which all P.sup.HR.sub.jP are diagonal matrices, or/and Hermitian transposition P.sup.H thereof, as a parameter for decorrelating components corresponding to the J source signals for observation signal vectors based on observation signals acquired at I (I is an integral number equal to or larger than 2) different positions.

A coding amount estimation device includes: a feature vector generation unit that generates a feature vector on the basis of a feature map generated by an estimation target image and at least one filter set in advance; and a coding amount evaluation unit that evaluates a coding amount of the estimation target image on the basis of the feature vector.

A method includes acquiring a first data column output from a plurality of sensors, generating a model for estimating data from the plurality of sensors on the basis of the first data column, acquiring a second data column output from the plurality of sensors, obtaining an estimated data column corresponding to the second data column based on the model by using regularization for making an error between the second data column and the estimated data column sparse, and identifying a sensor in which a change occurred between the first data column and the second data column on the basis of the error between the second data column and the estimated data column. A corresponding computer program product and apparatus are also disclosed herein.

An acceleration and compression method for a deep convolutional neural network based on quantization of a parameter provided by the present application comprises: quantizing the parameter of the deep convolutional neural network to obtain a plurality of subcode books and respective corresponding index values of the plurality of subcode books; acquiring an output feature map of the deep convolutional neural network according to the plurality of subcode books and respective corresponding index values of the plurality of subcode books. The present application may implement the acceleration and compression for a deep convolutional neural network.

An apparatus for pattern recognition includes a generator which transforms noisy feature vectors into denoised feature vectors, a discriminator which takes the denoised feature vectors and the original clean feature vectors corresponding to the denoised feature vectors as input and predicts probability for both of the input features of being an original clean feature, classifies the input feature vectors into its corresponding classes, an objective function calculator which calculates generator and discriminator losses using the denoised feature vectors, the clean feature vectors from which the noisy feature vectors have been made, the estimated classes and their true classes, and a Parameter updater which updates parameters of the generator and the discriminator according to loss minimization.

A method for processing electronic data includes the steps of transforming the electronic data to a matrix representation including a plurality of matrices; decomposing the matrix representation into a series of matrix approximations; and processing, with an approximation process, the plurality of matrices thereby obtaining a low-rank approximation of the plurality of matrices.