WO2020122287A1 - 미세 분포 변화를 이용한 비정상 데이터 구분 장치 및 방법 - Google Patents
미세 분포 변화를 이용한 비정상 데이터 구분 장치 및 방법 Download PDFInfo
- Publication number
- WO2020122287A1 WO2020122287A1 PCT/KR2018/015874 KR2018015874W WO2020122287A1 WO 2020122287 A1 WO2020122287 A1 WO 2020122287A1 KR 2018015874 W KR2018015874 W KR 2018015874W WO 2020122287 A1 WO2020122287 A1 WO 2020122287A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- module
- vector
- classification
- change
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Definitions
- the present invention relates to an apparatus and method for classifying abnormal data using a fine distribution change.
- Convlutional Neural Network may be used to develop a model that classifies the posture of a user seated in a seat based on time series pressure distribution data in the form of a Time series matrix generated from a smart seat composed of a mxn pressure sensor matrix.
- . 1 is a distribution diagram showing pressure distribution data
- FIG. 2 is a graph showing a change in pressure magnitude over time.
- pressure distribution data as shown in FIG. 1 may be generated from a plurality of pressure sensors configured in a matrix form, and each pressure sensor changes time series pressure magnitude as shown in FIG. 2 with respect to a time dimension. It can be represented to sense.
- 3 is an exemplary view of time series pressure distribution data.
- the time-series pressure distribution data may mean a plurality of pressure distribution data continuous in a time dimension generated based on a change in time series pressure magnitude generated by each pressure sensor as shown in FIG. 2.
- the corresponding ConvNet may be supervised learning based on pressure distribution data labeled or tagged with at least one posture.
- the accuracy of ConvNet is improved as the quality of data is collected as much as possible to train ConvNet. Therefore, it is useful to use the actual pressure distribution data of users generated in the course of operating the service to learn ConvNet. need.
- an object of the present invention is to provide an apparatus and method for classifying abnormal data using a fine distribution change in order to classify abnormal data input by a user with high accuracy and use only normal data for learning ConvNet.
- An object of the present invention a memory module for storing the program code of the abnormal data classification module for receiving the time series distribution data and outputting an abnormal data score for distinguishing whether it is abnormal data; And a processing module for processing the program code of the abnormal data classification module, wherein the program code of the abnormal data classification module learns to generate a normal vector that is a multidimensional vector of the normal time series distribution data based on potential variables.
- the abnormal data classification module if the abnormal data score lowered by the control of the latent variable in the latent variable adjustment step is more than a specific value, the data to be classified as the abnormal data It can be achieved by providing an abnormal data classification device using a fine distribution change, characterized in that the classification.
- the memory module further includes a fine change module that outputs fine change data representing fine change in the time series distribution data or the target data
- the processing module further processes the program code of the fine change module
- the program code of the fine change module may include: a receiving step of receiving the time series distribution data or the classification target data; A change amount data generation step of generating change amount data which is data on a change amount of the received time series distribution data or the distribution of the classification target data; A change amount distribution data generation step of converting the change amount data into a change amount distribution data in a matrix form; And a fine change data generation step of generating fine change data based on the change amount distribution data over time; and configured to be performed on a computer, wherein the normal vector is configured to include the fine change data of the time series distribution data.
- the vector to be classified may be configured to include the fine change data of the data to be classified.
- the memory module further includes a spatial data module that outputs spatial data indicating a distribution change due to the dynamic movement of the time series distribution data or the classification target data
- the processing module is a program code of the spatial data module Further processing, the program code of the spatial data module, the receiving step of receiving the time-series distribution data or the classification target data; A spatial feature extraction step of inputting the time series distribution data or the classification target data into an embedding network composed of ConvNet and extracting spatial features of the distribution using a feature map; And a spatial data generation step of embedding a sequence of temporal features and generating spatial data by inputting the spatial features into a long-term memory (LSTM); and wherein the normal vector is distributed over the time series.
- LSTM long-term memory
- Another object of the present invention is a generation vector step in which the generation vector module generates a generation vector by a generation module, which is a component of an abnormal data classification module that receives time series distribution data and outputs an abnormal data score for distinguishing whether it is abnormal data.
- the classification target vector module the classification target vector step of receiving a classification target vector that is a multidimensional vector of the classification target data that is the target of the classification of abnormal data;
- a latent variable adjustment module wherein a latent variable adjustment step of adjusting the latent variable in a direction in which the abnormal data score is lowered is configured to be performed on a computer, and wherein the generation module is normal to the time series based on the latent variable.
- the abnormality data classification module if the abnormality data score lowered by the control of the latent variable in the latent variable adjustment step is equal to or greater than a specific value, the target data for classification is the It can be achieved by providing a method for classifying abnormal data using a fine distribution change, characterized in that it is classified as abnormal data.
- an effect of being able to classify and collect unexpected abnormal data when collecting data is generated. Since abnormal data is often generated in unexpected situations, it is very difficult to build a deep learning classification model through labeling using the existing general statistical approach or Manual Feature Engineering.
- the effect of being able to use the characteristics of the minute change that is processed as noise by the general deep learning system by the minute change data module and vanishing can be used for posture classification and abnormal data classification.
- 1 is a distribution diagram showing pressure distribution data
- FIG. 4 is a schematic diagram showing an apparatus for classifying abnormal data according to an embodiment of the present invention.
- Figure 5 is a schematic diagram showing an example of pre-treatment according to an embodiment of the present invention.
- FIG. 6 is a flowchart illustrating an example of pre-processing flow according to an embodiment of the present invention
- FIG. 7 is a flowchart illustrating a method for generating fine change data in the fine change data module 11 according to an embodiment of the present invention
- FIG. 8 is a flowchart illustrating a method for generating spatial data in the spatial data module 12 according to an embodiment of the present invention
- FIG. 9 is a flowchart illustrating a posture category classification of the posture classification module 13 according to an embodiment of the present invention.
- FIG. 10 is a schematic diagram showing a learning process of the abnormal data classification module 14 according to an embodiment of the present invention.
- 11 is a schematic diagram showing an abnormal data classification process of the abnormal data classification module 14 according to an embodiment of the present invention.
- the invention is described based on a module that classifies the user's posture based on pressure distribution data, but the scope of the invention is not limited thereto, and an apparatus for classifying a specific category based on time series distribution data B.
- a device that classifies a specific category based on time series distribution data it may include a range including features for classifying abnormal data.
- the abnormal data classification apparatus 1 using the fine distribution change includes a pre-processing module 10, a fine change data module 11, a spatial data module 12, The posture classification module 13 and the abnormal data classification module 14 may be included.
- the pre-processing module 10 receives a plurality of pressure sensor data and performs time window setting, noise removal, normalization, and sensor deflection removal on the received pressure sensor data to generate time series pressure distribution data as shown in FIG. 3. to be.
- the pressure sensor data pre-processed by the pre-processing module 10 is processed as time-series pressure distribution data embedded as time-series data, thereby improving the accuracy of classification by a classification module composed of artificial neural networks.
- 5 is a schematic diagram showing an example of preprocessing according to an embodiment of the present invention
- FIG. 6 is a flowchart showing an example of preprocessing flow according to an embodiment of the present invention.
- the pre-processing module 10 can output a multi-dimensional vector (pre-processing data) based on the received pressure sensor data by performing time window setting, noise removal, normalization, and sensor deflection removal. And may be configured to generate time series pressure distribution data using a plurality of multidimensional vectors (preprocessing data) having a sequence.
- the fine change data module 11 is a module that generates fine change data based on time series pressure distribution data generated in the pre-processing module 10.
- the fine change data module 11 according to an embodiment of the present invention generates change amount data, which is data for a change amount of the pressure distribution, based on time series pressure distribution data, and converts the change amount data to change amount distribution data in a matrix form. Fine change data is generated based on the distribution data of the change amount over time.
- FIG. 7 is a flowchart illustrating a method for generating fine change data in the fine change data module 11 according to an embodiment of the present invention. As shown in FIG.
- the change amount distribution data P which is the distribution of the pressure distribution change, is generated through the histogram, and in this process, the spatial information about the change disappears and only the distribution information for the change remains.
- the distribution of the pressure distribution change is calculated, the spatial information disappears, and the overall distribution information of the fine change remains as the fine change data.
- the effect of preventing the expression of ambiguous information that can be propagated later by the mixing of the fine change feature and the spatial feature is generated by the fine change data module 11.
- the spatial data module 12 is a module that generates spatial data indicating a change in attitude due to a dynamic movement of a user based on time-series pressure distribution data generated in the pre-processing module 10. In relation to the generation of spatial data, the spatial data module 12 extracts the spatial characteristics of the pressure distribution and embeds the spatial characteristics through processing for multiple sequences to generate spatial data.
- FIG. 8 is a flowchart illustrating a method for generating spatial data in the spatial data module 12 according to an embodiment of the present invention. As shown in FIG.
- the spatial data module 12 inputs pre-processed time series pressure distribution data composed of a plurality of pre-processed data into an embedding network composed of ConvNet, extracts spatial characteristics of the pressure distribution using a feature map, and , It embeds a sequence characterized by time through LSTM (Long-Short Term Memory) and outputs it as spatial data.
- LSTM Long-Short Term Memory
- the spatial data module 12 may be configured to output an average distribution of preprocessed data by using the output embedding vector spatial data as an input of a verification network composed of ConvNet. According to this, it is possible to check whether the spatial data, which is the embedded vector, contains the spatial characteristics of the pressure distribution well, and the spatial data output by the spatial data module 12 is generated in the posture classification module 13. The effect of not only having a spatial characteristic for category classification but also for identifying abnormal data is generated.
- the posture classification module 13 is a module that generates posture classification data by receiving the fine change data generated in the fine change data module 11 and the spatial data generated in the spatial data module 12 and performing posture category classification.
- the posture classification module 13 according to an embodiment of the present invention is configured to embed features through nonlinear operation based on fine change data and spatial data, and perform posture category classification through a linear algorithm.
- Can be. 9 is a flowchart illustrating a posture category classification of the posture classification module 13 according to an embodiment of the present invention. As illustrated in FIG. 9, the posture classification module 13 according to an embodiment of the present invention may include Convlutional Neural Network (CNN) and Feed-Foward Neural Network (FFNN), and finely change the CNN.
- CNN Convlutional Neural Network
- FFNN Feed-Foward Neural Network
- Data and spatial data are input, non-linear features are extracted through CNN, and embedded vectors are output in the form required for posture category classification based on spatio-temporal features, and vectors embedded through CNN are input to FFNN.
- Posture category classification is performed through FFNN.
- the abnormal data classification module 14 is a multidimensional integrating fine change data generated in the fine change data module 11, spatial data generated in the spatial data module 12, and posture classification data generated in the pose classification module 13. It is a module that classifies whether or not abnormal data of a plurality of pressure sensor data input to the abnormal data classification device 1 by receiving a vector.
- a generation module and a classification module may be included, and the generation module receives random noise (Z) by using the classification module, thereby allowing normal fine change data.
- the generation module of the abnormal data classification module 14 may be configured to generate a normal vector consisting of an encoder and a decoder, and the encoder of the generation module may classify normal fine change data, spatial data, and posture. It can be composed of a plurality of consecutive ConvNets that receive a standardized multidimensional vector of mxnx 3 incorporating data and encode it into a latent variable of 1 x 1 xk, and the decoder of the pose transition module 4 has a potential of 1 x 1 xk. It can be composed of a plurality of consecutive networks that decode variables to be output as multi-dimensional vectors of mxnx 3.
- the generation module may be trained to output a multidimensional vector close to the normal vector by inputting the normal vector multidimensional vector, and may be trained by a classification module that distinguishes whether the multidimensional vector is a normal vector of the multidimensional vector output by the generation module. have.
- the classification module of the abnormal data classification module 14 may be configured to distinguish whether it is a normal vector of a multidimensional vector output by a generation module through a CONCAT function and a plurality of encoders.
- FIG. 10 is a schematic diagram showing a learning process of the abnormal data classification module 14 according to an embodiment of the present invention.
- the generation module may be configured with a loss function to configure a division module and a MinMax game, and may be simultaneously learned. Equation 1 below is a loss function of the generation module and the classification module.
- G is a generation module
- D is a classification module
- z is a random variable input as a latent variable
- y is a normal vector
- G(x) means a generated vector that is a generated multidimensional vector. Therefore, according to Equation 1, the loss function of the generation module and the classification module is D when the generation module is not sufficiently trained and the classification module perfectly classifies y and G(z) through the random noise z, which is a latent variable.
- the region loss function for distinguishing the normal or abnormality of the production vector may be further included through comparison of the production vectors generated by.
- the zone loss function according to an embodiment of the present invention may be configured as follows.
- L BP (G, D) is the loss function for each section, the section loss function, i is the i-th layer of the classification module, T is the entire layer of the classification module, and N i is the area in the i-th layer. It can mean the number of features. Accordingly, if the classification module accurately distinguishes normal or abnormal areas of a specific region from a specific layer of the generation vector,
- 1, and the normal vector If a specific region in a specific layer of y and the generation vector G(z) is not separated by the classification module,
- 0. Therefore, the loss function of the abnormal data classification module 14 to which the above-described zone loss function is applied may be configured as follows.
- ⁇ may be a weight constant
- L BP may mean a zone loss function of Equation 2
- the generation module is learned in a direction to minimize the zone loss function. As a result, the generation module is learned so that a generation vector closer to the normal vector is output more precisely.
- a loss function may be configured so that the generation module may consider the frame order.
- generation module generates vector at a certain time the latent variables in t z t and the previous time, the t-1 generation vector of G time t to a (z t-1) to the input data in the It can be configured to output G (z t ).
- the sequence loss function of the abnormal data classification module 14 for this may be configured as shown in the following equation.
- FIG. 11 is a schematic diagram showing an abnormal data classification process of the abnormal data classification module 14 according to an embodiment of the present invention.
- the generation module receives a random noise z to generate a multidimensional vector (generated vector) close to a normal vector, and of input user data (division target data that is an object for classifying whether it is abnormal data).
- the multidimensional vector (division target vector) and the generation vector, whether the user data (division target data) input is normal or abnormal is distinguished.
- the generation module may be configured such that the parameters are fixed after learning, and the potential variable through Back Propagation is reduced so that the division loss function (L), which is the difference between G(z) and y, is reduced. It can be configured to adjust the random noise z. Equation 5 below relates to the division loss function (L) for the difference between G(z) and y, and Equation 6 relates to the control of random noise as a latent variable.
- L is the difference between the generated vector generated close to the normal vector and the multidimensional vector of the user data (division target data) (division target vector)
- G(z) is the generated vector
- z is A potential variable random noise
- y is a multidimensional vector (a classification object vector) of user data (a classification object data)
- ⁇ is a learning rate.
- the loss value of L is specified even if z is adjusted to reduce the classification loss function L with the parameters of the generation module G fixed. It will not fall below the value. That is, Loss L when y is abnormal data has a relatively higher value than L when y is normal data. Therefore, the effect of being able to perform classification of abnormal data, classification of abnormal data, and detection of abnormal data is generated by using L as an abnormal data score.
- an effect of being able to classify and collect unexpected data when data is collected is generated. Since abnormal data is often generated in unexpected situations, it is very difficult to build a deep learning classification model through labeling using the existing general statistical approach or Manual Feature Engineering. In addition, unsupervised learning of classification, classification, and detection of abnormal data is possible without additional labeling or tagging of abnormal data. In addition, when attempting to classify, classify, or detect abnormal data based on other deep learning models and clustering algorithms, the distance between vectors embedded into the abstract space must be used as an anomaly score.
- the abnormal data classification device effectively uses the feature extraction performance of deep learning and includes a generation module G that creates normal data at the same time, the meaning of the abnormal degree of user data (classified data) is anomaly score The effect expressed in is generated.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
Claims (2)
- 시계열 분포 데이터를 입력받아 비정상 데이터인지 여부를 구분하는 비정상 데이터 스코어를 출력하는 비정상 데이터 구분 모듈의 프로그램 코드 및 상기 시계열 분포 데이터 또는 구분 대상 데이터의 미세 변화를 표현하는 미세 변화 데이터를 출력하는 미세 변화 모듈의 프로그램 코드를 저장하는 메모리 모듈; 및상기 비정상 데이터 구분 모듈의 상기 프로그램 코드 및 상기 미세 변화 모듈의 상기 프로그램 코드를 처리하는 처리 모듈;을 포함하고,상기 비정상 데이터 구분 모듈의 상기 프로그램 코드는,잠재 변수를 기초로 정상인 상기 시계열 분포 데이터의 다차원 벡터인 정상 벡터를 생성하도록 학습된 상기 비정상 데이터 구분 모듈의 일구성인 생성 모듈이 생성 벡터를 생성하는 생성 벡터 단계;비정상 데이터 구분의 대상이 되는 구분 대상 데이터의 다차원 벡터인 구분 대상 벡터를 수신하는 구분 대상 벡터 단계;상기 생성 벡터와 상기 구분 대상 벡터의 차이를 기초로 손실값인 비정상 데이터 스코어를 출력하는 비정상 데이터 스코어 출력 단계; 및상기 비정상 데이터 스코어가 낮아지는 방향으로 상기 잠재 변수를 조절하는 잠재 변수 조절 단계;를 포함하여 컴퓨터 상에서 수행되도록 구성되고,상기 미세 변화 모듈의 상기 프로그램 코드는,상기 시계열 분포 데이터 또는 상기 구분 대상 데이터를 수신하는 수신 단계;수신된 상기 시계열 분포 데이터 또는 상기 구분 대상 데이터의 분포의 변화량에 대한 데이터인 변화량 데이터를 생성하는 변화량 데이터 생성 단계;상기 변화량 데이터를 매트릭스 형태인 변화량 분포 데이터로 변환하는 변화량 분포 데이터 생성 단계; 및시간에 따른 상기 변화량 분포 데이터를 기초로 미세 변화 데이터를 생성하는 미세 변화 데이터 생성 단계;를 포함하여 컴퓨터 상에서 수행되도록 구성되고,상기 정상 벡터는 상기 시계열 분포 데이터의 상기 미세 변화 데이터를 포함하도록 구성되고, 상기 구분 대상 벡터는 상기 구분 대상 데이터의 상기 미세 변화 데이터를 포함하도록 구성되며,상기 비정상 데이터 구분 모듈은, 상기 잠재 변수 조절 단계에서 상기 잠재 변수의 조절에 의해 낮아진 상기 비정상 데이터 스코어가 특정 값 이상인 경우 상기 구분 대상 데이터가 상기 비정상 데이터인 것으로 구분하는 것을 특징으로 하는,미세 분포 변화를 이용한 비정상 데이터 구분 장치.
- 미세 변화 모듈이, 시계열 분포 데이터 또는 비정상 데이터 구분의 대상이 되는 구분 대상 데이터를 수신하는 수신 단계;상기 미세 변화 모듈이, 수신된 상기 시계열 분포 데이터 또는 상기 구분 대상 데이터의 분포의 변화량에 대한 데이터인 변화량 데이터를 생성하는 변화량 데이터 생성 단계;상기 미세 변화 모듈이, 상기 변화량 데이터를 매트릭스 형태인 변화량 분포 데이터로 변환하는 변화량 분포 데이터 생성 단계;상기 미세 변화 모듈이, 시간에 따른 상기 변화량 분포 데이터를 기초로 미세 변화 데이터를 생성하는 미세 변화 데이터 생성 단계;생성 벡터 모듈이, 상기 시계열 분포 데이터를 입력받아 비정상 데이터인지 여부를 구분하는 비정상 데이터 스코어를 출력하는 비정상 데이터 구분 모듈의 일구성인 생성 모듈이 생성 벡터를 생성하는 생성 벡터 단계;구분 대상 벡터 모듈이, 상기 구분 대상 데이터의 상기 미세 변화 데이터가 포함된 상기 구분 대상 데이터의 다차원 벡터인 구분 대상 벡터를 수신하는 구분 대상 벡터 단계;비정상 데이터 스코어 출력 모듈이, 상기 생성 벡터와 상기 구분 대상 벡터의 차이를 기초로 손실값인 비정상 데이터 스코어를 출력하는 비정상 데이터 스코어 출력 단계; 및잠재 변수 조절 모듈이, 상기 비정상 데이터 스코어가 낮아지는 방향으로 상기 잠재 변수를 조절하는 잠재 변수 조절 단계;를 포함하여 컴퓨터 상에서 수행되도록 구성되고,상기 생성 모듈은, 잠재 변수를 기초로 정상인 상기 시계열 분포 데이터의 다차원 벡터인 정상 벡터를 생성하도록 학습되며, 상기 정상 벡터는 상기 시계열 분포 데이터의 상기 미세 변화 데이터가 포함되고,상기 비정상 데이터 구분 모듈은, 상기 잠재 변수 조절 단계에서 상기 잠재 변수의 조절에 의해 낮아진 상기 비정상 데이터 스코어가 특정 값 이상인 경우 상기 구분 대상 데이터가 상기 비정상 데이터인 것으로 구분하는 것을 특징으로 하는,미세 분포 변화를 이용한 비정상 데이터 구분 방법.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/KR2018/015874 WO2020122287A1 (ko) | 2018-12-13 | 2018-12-13 | 미세 분포 변화를 이용한 비정상 데이터 구분 장치 및 방법 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/KR2018/015874 WO2020122287A1 (ko) | 2018-12-13 | 2018-12-13 | 미세 분포 변화를 이용한 비정상 데이터 구분 장치 및 방법 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020122287A1 true WO2020122287A1 (ko) | 2020-06-18 |
Family
ID=71076510
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2018/015874 WO2020122287A1 (ko) | 2018-12-13 | 2018-12-13 | 미세 분포 변화를 이용한 비정상 데이터 구분 장치 및 방법 |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2020122287A1 (ko) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110119374A1 (en) * | 2009-10-20 | 2011-05-19 | Jan Matthias Ruhl | Method and System for Detecting Anomalies in Time Series Data |
US20120166142A1 (en) * | 2009-09-07 | 2012-06-28 | Hitachi, Ltd. | Anomaly Detection and Diagnosis/Prognosis Method, Anomaly Detection and Diagnosis/Prognosis System, and Anomaly Detection and Diagnosis/Prognosis Program |
US20150269050A1 (en) * | 2014-03-18 | 2015-09-24 | Microsoft Corporation | Unsupervised anomaly detection for arbitrary time series |
-
2018
- 2018-12-13 WO PCT/KR2018/015874 patent/WO2020122287A1/ko active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120166142A1 (en) * | 2009-09-07 | 2012-06-28 | Hitachi, Ltd. | Anomaly Detection and Diagnosis/Prognosis Method, Anomaly Detection and Diagnosis/Prognosis System, and Anomaly Detection and Diagnosis/Prognosis Program |
US20110119374A1 (en) * | 2009-10-20 | 2011-05-19 | Jan Matthias Ruhl | Method and System for Detecting Anomalies in Time Series Data |
US20150269050A1 (en) * | 2014-03-18 | 2015-09-24 | Microsoft Corporation | Unsupervised anomaly detection for arbitrary time series |
Non-Patent Citations (2)
Title |
---|
DAN LI ET AL.: "Anomaly Detection with Generative Adversarial Networks for Multivariate Time Series", ARXIV, 13 September 2018 (2018-09-13), pages 1 - 10, XP055716405, Retrieved from the Internet <URL:https://arxiv.org/pdf/1809.04758vl.pdf> [retrieved on 20190628] * |
PANKAJ MALHOTRA ET AL.: "Long Short Term Memory Networks for Anomaly Detection in Time Series", ELEN.UCL.AC.BE, 22 April 2015 (2015-04-22), XP055346996, Retrieved from the Internet <URL:https://www.elen.ucl.ac.be/Proceedings/esann/esannpdf/es2015-56.pdf> [retrieved on 20190628] * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102570278B1 (ko) | 교사 모델로부터 학생 모델을 트레이닝하는데 사용되는 학습 데이터를 생성하는 장치 및 방법 | |
KR102036963B1 (ko) | Cnn 기반의 와일드 환경에 강인한 얼굴 검출 방법 및 시스템 | |
Mohandes et al. | Arabic sign language recognition using the leap motion controller | |
KR101925907B1 (ko) | 신경망 생성 모델을 이용한 객체 움직임 패턴 학습장치 및 그 방법 | |
KR102418596B1 (ko) | 소규모 학습 데이터 셋 기반 딥러닝 시스템과 이를 이용한 딥러닝 학습방법 | |
CN107301376B (zh) | 一种基于深度学习多层刺激的行人检测方法 | |
CN113378676A (zh) | 基于多特征融合的图像中人物交互检测方法 | |
CN109063625A (zh) | 一种基于级联深度网络的人脸关键点检测方法 | |
CN111612100B (zh) | 对象再识别方法、装置、存储介质及计算机设备 | |
CN111862065B (zh) | 基于多任务深度卷积神经网络的输电线路诊断方法和系统 | |
CN117116498A (zh) | 移动查房数据处理系统及其方法 | |
CN110688980A (zh) | 基于计算机视觉的人体姿态分类方法 | |
CN112149962A (zh) | 一种施工事故致因行为的风险定量评估方法及系统 | |
KR102274581B1 (ko) | 개인화된 hrtf 생성 방법 | |
CN113127857A (zh) | 针对对抗性攻击的深度学习模型防御方法及深度学习模型 | |
CN112087443A (zh) | 一种大规模工业传感网络信息物理攻击下传感数据异常智能化检测方法 | |
CN117723158A (zh) | 非接触式体温监测系统及其方法 | |
CN115797701A (zh) | 目标分类方法、装置、电子设备及存储介质 | |
KR102198224B1 (ko) | 공간 데이터를 기초로 한 인공신경망을 이용한 비정상 데이터 구분 장치 | |
WO2021137320A1 (ko) | 인공신경망을 이용한 비정상 데이터 구분 장치 | |
Abdul-Ameer et al. | Development smart eyeglasses for visually impaired people based on you only look once | |
KR20080100097A (ko) | 2차원 정준상관 분석법을 이용한 얼굴 인식 방법 | |
WO2020122287A1 (ko) | 미세 분포 변화를 이용한 비정상 데이터 구분 장치 및 방법 | |
WO2018092957A1 (ko) | 신경망 모델에서 입력값에 대한 재학습 여부 결정 방법, 장치 및 프로그램 | |
KR102119891B1 (ko) | 미세 변화 데이터 및 공간 데이터를 기초로 한 인공신경망을 이용한 비정상 데이터 구분 장치 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 18943180 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18943180 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 26/04/2022) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18943180 Country of ref document: EP Kind code of ref document: A1 |