740results about How to "Improve interpretability" patented technology

A system and method are provided for diagnosing diseases or conditions from digital images taken by a remote user with a smart phone or a digital camera and transmitted to an image analysis server in communication with a distributed network. The image analysis server includes a trained learning machine for classification of the images. The user-provided image is pre-processed to extract dimensional, shape and color features then is processed using the trained learning machine to classify the image. The classification result is postprocessed to generate a risk score that is transmitted to the remote user. A database associated with the server may include referral information for geographically matching the remote user with a local physician. An optional operation includes collection of financial information to secure payment for analysis services.

The methods, systems and devices of the present invention comprise use of Support Vector Machines and RFE (Recursive Feature Elimination) for the identification of patterns that are useful for medical diagnosis, prognosis and treatment. SVM-RFE can be used with varied data sets.

This invention describes methods and algorithms for processing a plurality of relevant signals/data intrinsic to a patient and/or derived from external diagnostic equipment for management of atrial arrhythmias. The intrinsic signals are acquired from intracardiac leads/sensors and analogous extrinsic data obtained from imaging equipment and patient demographics. These data are input into software algorithms that use digital signal processing to output informational data of clinical and technical relevance after comparisons are made to patients with access to this technology whose outcome under varying treatments is known. These combined data are used to define prognosis, make treatment suggestions, direct programming of cardiac devices and digitally convert intrinsically and extrinsically derived indices into a common metric. In a preferred embodiment, the intrinsically and extrinsically acquired data is utilized in the design of catheters and software algorithms for performing intracardiac procedures such as ablation of atrial arrhythmias.

The invention discloses an implementation method of a fusion network question and answer system based on a multi-attention mechanism, which comprises the following steps of constructing a question andanswer system network model, preprocessing an original data set to obtain a standby data set, and performing text length distribution analysis; subjecting text in standby data set to one-hot vector representation, using a CBOW model to train one-hot word vector and forming a word2vec word list; adjusting the sequence length of each sentence in the text, and adding a sentence end mark; training the word2vec vector by using an ELMO language model to obtain an ELMO word vector; encoding the ELMO vector to obtain a sentence vector; performing coarse-fine granularity attention on the sentence vectors respectively to obtain memory vectors and attention vectors based on each word; carrying out vector splicing to obtain expression vectors based on words and sentences; and decoding an answer representing the vector generation question sentence. According to the method, the representation ability of sentences is improved through an ELMO language model; and various attention mechanisms are fused, so that the decision making accuracy of the system is improved, and the interpretability of the system is enhanced.

The invention discloses a sparse dimension reduction-based spectral hash indexing method, which comprises the following steps: 1) extracting image low-level features of an original image by using an SIFT method; 2) clustering the image low-level features by using a K-means method, and using each cluster center as a sight word; 3) reducing the dimensions of the vectors the sight words by using a sparse component analysis method directly and making the vectors sparse; 4) resolving an Euclidean-to-Hamming space mapping function by using the characteristic equation and characteristic roots of a weighted Laplace-Beltrami operator so as to obtain a low-dimension Hamming space vector; and 5) for an image to be searched, the Hamming distance between the image to be searched and the original image in the low-dimensional Hamming space and using the Hamming distance as the image similarity computation result. In the invention, the sparse dimension reduction mode instead of a spectral has principle component analysis dimension reduction mode is adopted, so the interpretability of the result is improved; and the searching problem of the Euclidean space is mapped into the Hamming space, and the search efficiency is improved.

A computer-aided detection system for evaluating tissue based on a series of timed images acquired both before and after a contrast agent is administered performs a temporal analysis of the tissue and then a spatial analysis in which the tissue is categorized. After the temporal and spatial analysis is performed, the results can be displayed. The displayed results can include both tissue characterization results, underlying curves used to determine the characterization, and confidence data regarding how good the characterization is expected to be. The confidence data can be provided, for example, by using variations in color schemes, displaying numerical confidence levels, or providing graphical features such as piecewise linear models.

The invention discloses an item recommendation method for combining user comment content and grades. A model for combining the user comment content with user grades is provided for solving the cold start problem and the poor interpretability problem in a recommendation system. By means of the rich information contained in comments, the prediction accuracy is greatly improved; particularly, when data is quite sparse, the cold start problem and the poor interpretability problem can be well solved. The method mainly considers that descriptions of item characteristics are contained in user comment information, and potential characteristics in numerical value grades correspond to item characteristics in comment information through a mapping function. A model can be well built for user preferences, and therefore prediction and recommendation can be well carried out even if data is quite sparse.

The invention provides a convolutional neural network power system intelligent fault detection method and system based on Spearman level correlation, and the method comprises the steps: setting a phasor measurement unit at a regional network node, and carrying out the measurement of data; performing Spearman correlation analysis on the acquired data, and proposing an image generation method basedon an analysis result; establishing an equivalent fault network, verifying the relation between the fault characteristics and the Spearman level correlation, and demonstrating the feasibility of the method; taking the generated image as an initial convolutional layer, and establishing a convolutional neural network architecture based on Spearman level correlation; and verifying the rationality andsuperiority of the method based on PSCAD/EMTDC according to the established architecture. A plurality of types of electric quantity data are comprehensively used for fault diagnosis, the position ofa fault in the power system can be quickly and accurately identified through the convolutional neural network, the problems that the power system has volatility and the traditional detection method isinaccurate due to addition of a distributed power supply and the like are solved, and the robustness and the self-adaptability of the power system are higher.

A method for stratigraphic analysis of seismic data, known as GeoLeaf, comprises: selecting a seismic data volume comprising a plurality of samples; selecting a number of horizons (TH, BH1 TPH, BPH, SMH) within the seismic data volume; selecting a scale of interest within the seismic data volume, which ranges from 500 m up to the size of the entire seismic data volume; applying an attribute function to at least one of the selected number of horizons (TH, BH, TPH, BPH, SMH) at the selected scale of interest; and displaying a stratigraphic attribute, which represents a geologic feature, such as a seismic fades attribute (FM), an unconformity attribute (Z_max/Z) and/or a thinning attribute (Z₁/Z₂).

To follow and inspect a common edge between two sheet metal blanks which are to be welded together, a line pattern is projected across the edge by a projector and is imaged by a camera. The image captured is analyzed in order to derive from the trace of the lines information on the gap between the blanks, or on the track of the edge. The pattern projected has several lines with different light intensities. When the photographed image is processed, it is then possible—even if there is variation in the reflection characteristics of the sheet metal blanks—to select a line which is neither too faint nor too bright in the image obtained and which can therefore be used for meaningful analysis.

The invention discloses an electroencephalogram signal recognition fuzzy system and method with the transfer learning ability. In a traditional intelligent recognition method, a training set and a testing set of a model are assumed to conform to the same data distribution, and therefore the good classification performance can be obtained only when data in a training domain and data in a testing domain conform to the same distribution. The electroencephalogram signal recognition fuzzy method helps epileptic electroencephalogram signal recognition under the transfer learning environment by means of the transfer learning strategy. The 0-order TSK type fuzzy system modeling technology with the direct-pushing type transfer learning ability is built based on the fuzzy system. The technology has the transfer learning ability and is not confined to the assumption of the uniform data distribution of the training domain and the testing domain, a certain difference is allowed to exist between the data in the training domain and the data in the testing domain, the good performance is kept on the condition of the same data distribution of the training domain and the testing domain, and the recognition effect of the finally-obtained model under the diversified electroencephalogram signal recognition problems is greatly improved.

The embodiment of the invention provides a product recommendation method and device and a storage medium. In some exemplary embodiments of the present application, a meta-path set associated with therecommended service scene is selected from the insurance knowledge graph in combination with the recommended service scene, and for the target client, a similar client list of the target client and asimilar product list of the insurance products purchased by the target client are generated according to the entity characteristics contained in the plurality of meta-paths and the service relationship existing between the entity characteristics; collaborative filtering recommendation is performed according to the client similarity list and the similar product list to obtain a first product set capable of being recommended to the target client, a recommendation strength score of each product in the first product set, a second product set and a product recommendation strength score of each product in the second product set; according to the recommendation strength score of each product in the first product set and the product recommendation strength score of each product in the second product set, the target product recommended to the target user is determined. According to the product recommendation method, the insurance knowledge graph is utilized to fully express various semantic information, and the accuracy, the diversity and the interpretability of the product recommendation result are improved.

The invention discloses an early warning method and device applied to an information system. The method comprises the steps that original monitoring data of a monitored node within a preset period of time is collected, original data description is constructed according to preset characteristic indexes; according to the original data description, comprehensive indexes are determined, wherein the comprehensive indexes include weight coefficients corresponding to index components; according to the original data description and the comprehensive indexes, training data is determined, according to the training data, a prediction model is trained, the prediction model is determined, and target prediction data of a target node is determined according to the prediction model; actually obtained data of the target node is determined and converted according to the comprehensive indexes, and the conversion result is compared with the target prediction data; when the comparison result conforms to preset characteristics, corresponding information system early warning information is sent out. According to the method, flexibility and adaptability of information system risk early warning and the early warning result accuracy can be improved.

Computer-implemented methods for learning decision trees to optimize classification accuracy, comprising inputting an initial decision tree and an initial data training set and, for nodes not descendants of each other, if the node is a leaf, assigning a label based on a majority label of training points that reach the leaf, and if the node is a decision node, updating the parameters of the node's decision function based on solution of a reduced problem, iterating over the all nodes of the tree until parameters change less than a set threshold, or a number of iterations reaches a set limit, pruning the resulting tree to remove dead branches and pure subtrees, and using the resulting tree to make predictions from target data. In some embodiments, the TAO algorithm employs a sparsity penalty to learn sparse oblique trees where each decision function is a hyperplane involving only a small subset of features.

The invention discloses an intelligent vehicle driving decision method based on a generating countermeasure network, which comprises establishing a driving decision model and driving decision control.The invention processes the driving image based on the generated antagonistic network, can process the vehicle driving path planning under the non-ideal road condition, and improves the executabilityof the end-to-end neural network. The invention extracts the most essential characteristics of driving images through the generated antagonistic network processing, maps the driving data of differentsources into a unified virtual domain, realizes the application of reinforcement learning to a real vehicle, improves the generalization of the network and adapts to the ability of different samples.For the input of the driving image, the input image used each time is the first several frames of the current time stamp video image. The predicted images obtained by this method can be used to judgethe driving decision-making planning to a greater extent. As the basis for predicting the optimal decision of the vehicle, the invention builds a bridge from the reinforcement learning to the real vehicle application.

The invention belongs to the field of deep learning, and discloses a multivariate time series multilayer space-time dependence modeling method based on deep learning. According to the method, a novelattention mechanism is introduced to carry out finer-grained processing on space-time dependence features extracted from different layers in the neural network; the model provided by the invention iscomposed of a stacked long-short-term neural network-convolutional neural network (LSTM-CNN). The network is composed of a spatial attention mechanism based on a CNN, a CNN-based channel attention mechanism, a time attention mechanism and an autoregressive assembly. The concept of multi-layer space-time dependence is introduced; a CNN-based channel attention mechanism and a CNN-based spatial attention mechanism are used to pay attention to space-time dependence characteristics of different layers respectively. According to the method, filtering of redundant information and effective extractionof features having greater influence on a prediction result are realized, the purpose of improving the prediction result is achieved, and the method is excellent in performance on multivariate time series data in different fields and can be extended to a task of unit time series prediction.

The invention discloses an ICD automatic coding method and system for diagnosis reason visualization. The method comprises the following steps: acquiring medical record data from a medical record document library, and constructing a multi-label classification data set; preprocessing the data set, and converting the multi-label classified data set into a plurality of single-label classified data sets; completing training of a hierarchical attention neural network model based on the data sets of the plurality of single-label classifications; inputting doctor writing diagnosis data and illness state description data into the trained hierarchical attention neural network model so as to obtain predicted ICD codes and names; and according to the predicted ICD name, extracting a corresponding sentence from the disease description data as a diagnosis reason for visualization. According to the method, doctor writing diagnosis and medical record description data are used as mode features at thesame time, doctor writing diagnosis is ingeniously used as supervision information in the classification process, traditional multi-label classification is converted into a simpler single-label classification problem, and therefore the accuracy of model coding is improved.