86 results about "Pattern learning" patented technology

A learning device includes: a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data; model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and sharing strength updating means for updating sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.

The invention discloses a detection method for an abnormal behavior of a vehicle based on spectrum clustering. The detection method comprises the following steps: obtaining a space-time track of a moving target through video tracking; removing abnormality and preprocessing, thereby obtaining a normal track; constructing an image for the track, thereby obtaining an undirected image corresponding to a track sequence; calculating similarity among tracks, thereby obtaining a similarity matrix; performing Laplace transformation on the similarity matrix, thereby obtaining a Laplace matrix; clustering the feature vector matrix of the front k maximal feature values; after performing mode learning on a motion track, obtaining motion modes of the target under a normal state; if a new track meets one of the motion modes, i.e. a normal motion mode, confirming that the traffic is normal; and if not, confirming that the vehicle abnormally runs, namely, the traffic abnormality occurs. According to the detection method, through the clustering learning for the vehicle track, the monitoring for the abnormal behavior of the vehicle is realized, the abnormal lane change is detected and the basis for automation of traffic management is supplied.

A method for transparently optimizing data access. The method includes gathering information related to data usage when a system is processing and determining a usage pattern of the system using gathered information.

The invention relates to determination of a presentation pattern of a teaching material in a computer-assisted education system. An information processing apparatus includes first processor means for analyzing an answer from a user to determine a trait of the user related to a personality, and for determining a teaching material presentation pattern for the user in accordance with the trait of the user. Second processor means is provided for selecting and editing ones of a plurality of teaching material elements of a specific subject in accordance with the presentation pattern, generating a teaching material module, and presenting the teaching material module to the user in accordance with the presentation pattern. Third processor means is provided for analyzing learning behavior of the user during learning of the user using the presented teaching material module in accordance with the presentation pattern, and modifying the presentation pattern in accordance with the analysis.

A system for fault prediction in a home network includes: a context generator for generating context information based on status data collected in real time about components of the home network; a specification interpreter for generating knowledge rules for fault detection by using specifications of the components of the home network; a context analyzer for analyzing if the context information meet the knowledge rules to classify the context information into normal situation contexts and abnormal situation contexts; a context pattern learner for generating new knowledge rules based on the abnormal situation contexts and fault rules corresponding to the abnormal situation contexts; a knowledge rule database for storing and managing the knowledge rules and the new knowledge rules; and a fault predictor for analyzing a correlation between the knowledge rules or the new knowledge rules and the generated context information, thereby predicting faults to be generated.

This invention is to use knowledge pattern learning and search system for selecting microorganisms to produce useful materials and to generate clean energy from wastes, wastewaters, biomass or from other inexpensive sources. The method starts with an in silico screening platform which involves multiple steps. First, the organisms' profiles are compiled by linking the massive genetic and chemical fingerprints in the metabolic and energy-generating biological pathways (e.g. codon usages, gene distributions in function categories, etc.) to the organisms' biological behaviors. Second, a machine learning and pattern recognition system is used to group the organism population into characteristic groups based on the profiles. Lastly, one or a group of microorganisms are selected based on profile match scores calculated from a defined metabolic efficiency measure, which, in term, is a prediction of a desired capability in real life based on an organism's profile. In the example of recovering clean energy from treating wastewaters from food process industries, domestic or municipal wastes, animal or meat-packing wastes, microorganisms' metabolic capabilities to digest organic matter and generate clean energy are assessed using the invention, and the most effective organisms in terms of waste reduction and energy generation are selected based on the content of a biowaste input and a desired clean energy output. By selecting a microorganism or consortia of multiply microorganisms using this method, one can clean the water and also directly generate electricity from Microbial Fuel Cells (MFC), or hydrogen, methane or other biogases from microorganism fermentation. In addition, using similar screening method, clean hydrogen can be recovered first from an anaerobic fermentation process accompanying the wastewater treatment, and the end products from the fermentation process can be fed into a Microbial Fuel Cell (MFC) process to generate clean electricity and at the same time treat the wastewater. The invention can be used to first select the hydrogenic microorganisms to efficiently generate hydrogen and to select electrogenic organisms to convert the wastes into electricity. This method can be used for converting wastes to one or more forms of renewable energies.

Method and continuously wearable noninvasive apparatus for automatically detecting a stroke's onset are invented. The method comprises measuring pulse transit time or a related hemodynamic parameter and using adaptive pattern learning and stroke identification algorithms to identify the start of a stroke attack. The algorithms are run on an embedded processor and electronics in the apparatus. The result of the identification and detection is used to generate an alarm to alert the patient, caregiver or medical professional to immediately seek further medical treatment. The method can also be applied to identify and detect other abnormal health conditions.

Open Chinese entity relationship extraction refers to, on the premise of not limiting a corpus field and a relationship category, automatic extraction of relationship information between entities from a Chinese corpus to obtain an entity relationship tuple. The present invention discloses a pattern self-learning based Chinese opening relationship extraction method. The method comprises the following three main steps of: firstly, based on an existing knowledge library, acquiring a high-quality entity relationship tuple and a corresponding sentence as a training corpus, and obtaining a dependent path mode between an entity and a relationship word by a pattern learning method proposed by the present invention; secondly, performing pre-processing of word segmentation, part-of-speech tagging, dependency analysis and the like on a to-be-extracted text, and performing entity relationship extraction by means of a relationship mode obtained by previous learning; and finally, performing quality evaluation on an entity relationship extracted automatically from the Chinese corpus by using a machine learning method, and obtaining the high-quality entity relationship tuple.

An adaptive learning interface system for end-users for controlling one or more machines or robots to perform a given task, combining identification of gaze patterns, EEG channel's signal patterns, voice commands and / or touch commands. The output streams of these sensors are analysed by the processing unit in order to detect one or more patterns that are translated into one or more commands to the robot, to the processing unit or to other devices. A pattern learning mechanism is implemented by keeping immediate history of outputs collected from those sensors, analysing their individual behaviour and analysing time correlation between patterns recognized from each of the sensors. Prediction of patterns or combination of patterns is enabled by analysing partial history of sensors' outputs. A method for defining a common coordinate system between robots and sensors in a given environment, and therefore dynamically calibrating these sensors and devices, is used to share characteristics and positions of each object detected on the scene.

A learning device includes: a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data; model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and classification means for classifying the plurality of learning modules on the basis of the plurality of model parameters of each of the learning modules after the update learning.

The invention discloses a behavior pattern learning system for an augmentor, and relates to the technical field of a robot control system. An intelligent terminal comprises an action control input part, a camera I, a microphone I, a display screen I, a loudspeaker I, and the camera I; an augmentor interaction system comprises robot actions, a display screen II, a loudspeaker II, a camera II and a microphone II; a substituted person behavior sub system comprises an interaction person voice recognition part, an associated syntax database, a language habit library, an expression library, an action library, a substituted person voice recognition part, a language habit training part, a language expression matching training part, and a language action matching training part; a robot performs long-term deep learning on action habits, language habits, expression habits and the like which can express behavior habit characteristics of the substituted person through learning algorithms of a neural network and the like in the use process of the substituted person, so that when the substituted person is not online, the robot still can express behavior characteristics of the substituted person so as to improve the affinity of the augmentor in personification substituting.

The invention discloses an event extraction method, device and system and a storage medium. The method includes: utilizing triggering words, which are obtained by carrying out pattern learning on a corpus, to position event sentences in documents of the corpus, and determining event categories of the event sentences according to relationship classifications of the triggering words; using patterns,which are obtained by pattern learning, to carry out pattern matching on the event sentences to obtain event sentence sets matching the patterns; and extracting event contents in the event sentence sets according to preset event learning templates corresponding to the event categories. According to the event extraction method provided by the embodiment of the invention, efficiency and integrity of event content extraction can be improved.

The invention provides an on-vehicle situation detection apparatus and method. The on-vehicle situation detection apparatus may include a detection unit to identify a driver and acquire driver status data and data about vehicle driving information or vehicle surrounding obstacles, a driving pattern learning unit to learn and store a driving pattern of a driver, based on the data acquired by the detection unit, a weighted value determination unit to determine a weighted value assigned to the information data acquired by the detection unit, based on the driving pattern learned by the driving pattern learning unit, a determination unit to determine a safe driving state of the driver, based on the data to which the weighted value determined by the weighted value determination unit is assigned, and a warning unit to warn the driver when the driver is determined to be not in the safe driving state. The on-vehicle situation detection apparatus and method determine whether the driver is driving safely by knowing mental and physical states associated with vehicle driving or vehicle operation by the driver, and guide, if determining that the driver is not in a safe driving state, the diver to drive safely by displaying warning, giving warning sound, reminding the driver by vibration, forcibly controlling the vehicle or other methods so as to protect the driver.

A learning device includes: a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data; model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; module creating means for creating a new learning module corresponding to new learning data for learning the pattern when the new learning data are supplied as the input data; similarity evaluation means for evaluating similarities among the learning modules after the update learning is performed over all the learning modules including the new learning module; and module integrating means for determining whether to integrate the learning modules on the basis of the similarities among the learning modules and integrating the learning modules.

The invention discloses a judgment method of ultra-high voltage equipment local discharge detection data. The judgment method comprises the following steps: sampling a continuous ultrasonic frequencysignal to reduce to the continuous sound wave frequency signal capable of being heard by the human ear; continuously intercepting a frame sound wave frequency signal with a set time length; extractinga Mayer frequency cepstrum coefficient of the frame sound wave frequency signal as a to-be-identified fault discharge feature; sending the extracted to-be-identified fault discharge feature into a CNN convolution neural network, enabling the to-be-identified fault discharge feature to enter a fault classifier of a CNN convolution neural network output classification layer through CNN convolutionneural network analysis, wherein the CNN convolution neural network identifies the to-be-identified fault discharge feature and outputs the to-be-identified fault discharge feature according to the fault classifier formed by learning the known fault discharge feature in advance. The mode learning and identification are performed on the fault type by directly using the convolution neural network CNN, the identification accuracy rate is improved, and the manual intervention is reduced or avoided.

A method for implementing automated information technology (IT) system failure recommendation and mitigation includes performing log pattern learning to automatically generate sparse time series for each log pattern for a set of classification logs corresponding to a failure, performing multivariate log time series extraction based on the log pattern learning to generate a failure signature for the set of classification logs, including representing the sparse time series as a run-length encoded sequence for efficient storage and computation, calculating a similarity distance between the failure signature for the set of classification logs and each failure signature from a failure signature model file, determining a failure label for the failure corresponding to a most similar known failure based on the similarity distance, and initiating failure mitigation based on the failure label.

The invention discloses a near-surface air temperature inversion method comprising the following steps: establishing an original data record set of an unmanned weather station; constructing a first sub-pattern learning set and a first sub-pattern validation set; and acquiring a second sub-pattern to a fth sub-pattern, performing near-surface air temperature inversion to acquire a near-surface air temperature inversion image map of a target zone, and performing error correction to acquire a corrected near-surface air temperature inversion image map. According to the near-surface air temperature inversion method disclosed by the invention, the near-surface air temperature inversion is performed by collecting actually-measured air temperature of the unmanned weather station, collecting meteorological satellite data, DEM data and astronomy and calendar rules and also adopting a super nonlinear algorithm, and the near-surface air temperature inversion image map is then calculated by using a high-performance computer. Results show that the near-surface air temperature inversion method disclosed by the invention is relatively high in pattern accuracy, high in result reliability and strong in generalization ability, and ensures that the interferences of clouds, terrains and the like can be overcome; and a constructed CPU+GPU heterogeneously-cooperative parallel computer ensures that the computation speed can be increased by more than 1000 times, so that the near-surface air temperature inversion method is convenient for large-area application and computing capacity expansion.

A word classification system is provided with an inter-word pattern learning section for learning at least either the context information or the layout information between classification-known words which co-appear and creating an inter-word pattern for determining whether data relating to a word pair which is a combination of words is data relating to a same-classification word pair which is the combination of words in the same classification or data relating to a different-classification word pair which is a combination of words in different classifications on the basis of the relationship between the classification-known words which co-appear in a document.

A robot controller capable of easily speeding up the motion of a robot by learning, without using a special device and teaching know-how. The robot controller has a storing part which stores a reference motion pattern of the robot; an inputting part which designates at least one of a working start region where a motion of the robot based on the reference motion pattern is initiated and a working end region where the motion of the robot is terminated; an automatic generating part which automatically generates a plurality of motion patterns of the robot based on the reference motion pattern and divided regions formed by dividing a working region at a predetermined resolution; and a learning control part which learns a motion speed-up ratio for speeding up the motion by changing a velocity or acceleration in relation to each of the automatically generated motion patterns.

The invention discloses an iterative extraction method of Chinese synonyms based on pattern learning: taking unstructured data of encyclopedia entries as a corpus, matching seed synonyms pair obtainedby redirecting with corpus text to obtain text between word pairs as a candidate pattern; Through candidate pattern matching, extracting the entity pairs before and after the pattern as candidate synonym pairs in the text statement. Adopting Word2vec to calculate the semantic similarity between entity pairs and evaluate the similarity of word pairs. Counting The number of seeds supported by candidate patterns, and evulating the quality of the candidate patterns according to the words extracted from the candidate patterns. Then, scoring the candidate synonyms by pattern score, entity confidence and word pair similarity, and the effective synonym entity pairs are selected. Using the extracted high quality synonyms as new seeds and obtaining more Chinese synonym pairs. The method provided bythe invention successfully extracts a large number of Chinese synonym entities with high accuracy from ten million encyclopedic entry texts, and has great application significance for extracting synonym information by using massive unstructured texts.

A system for fault prediction in a home network includes: a context generator for generating context information based on status data collected in real time about components of the home network; a specification interpreter for generating knowledge rules for fault detection by using specifications of the components of the home network; a context analyzer for analyzing if the context information meet the knowledge rules to classify the context information into normal situation contexts and abnormal situation contexts; a context pattern learner for generating new knowledge rules based on the abnormal situation contexts and fault rules corresponding to the abnormal situation contexts; a knowledge rule database for storing and managing the knowledge rules and the new knowledge rules; and a fault predictor for analyzing a correlation between the knowledge rules or the new knowledge rules and the generated context information, thereby predicting faults to be generated.

According to one embodiment, a pattern learning system includes a pattern learning tool that receives event messages in a sequential manner from multiple sensors and forms multiple sub-sequences that each includes a trigger event message, a consequence event message, and one or more intermediary event messages. The pattern learning tool then generates multiple graphs that each represents a sub-sequence of the plurality of event messages. The pattern learning tool then combines the graphs into a combined graph according to a type of each event message, and determines a causal sequence from the combined graph according to a heaviest weighted directed path from the trigger event message to the consequence event message.

The invention discloses a speech learning machine and player with a bilingual model. The speech learning machine includes a controller, as well as a memory, a loudspeaker and control buttons which are connected with the controller, wherein the memory includes a Chinese speech database and a foreign language database; the controller is provided with three learning modules, i.e. a Chinese model learning module, a foreign language model learning module and a bilingual model learning module; and the control buttons includes model selection keys for selecting a Chinese model, a foreign language model and the bilingual model. Through the independent selection function key for the bilingual learning model, learners can select the bilingual learning model (Chinese and a foreign language) at the initial learning stage, and select the independent foreign language or Chinese model when familiar with the learning. The speech learning machine further adopts the principle of the Ebbinghaus Forgetting Curve for an additional repeat play rote learning mode, so that users can select single repeat play, double repeat plays or triple repeat plays according to personal requirements, which can greatly improve the learning effect of learners.

A learning device, learning method, and program for learning a pattern are disclosed. A learning device includes: a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of a pattern learning model that learns a pattern using input data; model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and sharing strength updating means for updating sharing strengths between the learning modules so as to minimize learning errors when the plurality of model parameters are updated by the update learning.

The invention discloses a street lamp intelligent fault diagnosis method. A narrow band Internet of Things (IoT) module embedded with an illuminance sensor is installed at a light source of each street lamp, and each narrow band IoT module is numbered. Each illuminance sensor measures the illuminance at the corresponding narrow band IoT module, and uploads the collected illuminance data to a server to construct an illuminance database. A normal illuminance data set is selected from the illuminance database as a training data set, and a machine learning method is used for pattern learning of an illuminance sequence to construct a street lamp turn-on model. On the basis of the street lamp turn-on model, the real-time illuminance sampling data sequence is processed to realize the fault diagnosis and fault type diagnosis of a single street lamp and a lamp group. Also, the machine learning method is used for updating the street lamp turn-on model. According to the invention, the faults of the single lamp and the lamp group can be automatically detected in time without breaking the line, the fault types can be determined and fed back to working staff, and the intelligent diagnosis of the street lamp faults is realized.

A system and method for providing surveillance data are provided. The system includes: a pattern learner configured to learn a time-based data pattern by analyzing at least one of image data of one or more images and sound data of sound obtained from a surveillance zone at a predetermined time or time period, and to generate an event model based on the time-based data pattern; and an event detector configured to detect at least one event by comparing the event model with a time-based data pattern of at least one of first image data of one or more first images and first sound data of first sound obtained from the surveillance zone.

A device for implementing a pattern learning model, the device including a plurality of learning modules, each of which performs update learning to update a plurality of model parameters of the pattern learning model that learns a pattern using input data. The device further including a model parameter sharing means for causing two or more learning modules from among the plurality of learning modules to share the model parameters; and a classification means for classifying the plurality of learning modules on the basis of the plurality of model parameters of each of the learning modules after the update learning.