Spectral clustering of high-dimensional data
View Patent ↗A processor performing machine learning including spectral clustering can receive data from the sensor. Graph Laplacian of the data can be created and stored in a memory device. Spectral characteristic can be created by applying density of states and spectral gaps can be detected in an unsupervised manner in the spectral characteristic to determine r as number of clusters to cluster the data. A range space of a rational matrix of the graph Laplacian can be determined. K-means clustering can be performed on the range space of rational matrix of the graph Laplacian using r as the number of clusters, the K-means clustering returning r clusters of the received data.
1. A machine learning system comprising:
a processor;
a memory device coupled with the processor;
a sensor coupled with the processor;
the processor configured at least to:
receive data from the sensor;
create graph Laplacian of the data and store in the memory device;
compute spectral characteristic by applying density of states and detect spectral gaps in an unsupervised manner in the spectral characteristic to determine r number of clusters, r being a hyper-parameter for machine learning;
compute a range space of a rational matrix of the graph Laplacian, r being determined based on dynamically increasing projection subspace for capturing eigenvalues, in which the subspace is built by computing the range space, and without requiring estimation of the eigenvalues located inside a disk in a complex plane;
train an unsupervised machine learning model based on the hyper-parameter r to cluster the received data, wherein to train the unsupervised machine learning model, the processor is configured to perform K-means clustering on the range space of rational matrix of the graph Laplacian using r as the number of clusters, the K-means clustering trained to return r clusters of the received data;
the data including computer network traffic data, wherein the unsupervised machine learning model is trained to classify the computer network traffic data into r clusters of security levels; and
run the trained unsupervised machine learning model on incoming data traffic in real-time to detect a security level cluster to which the incoming data traffic belongs, and based on the security level cluster, filter the incoming data traffic to prevent unwanted data from entering a target computer system.
2. The system of claim 1 , wherein the range space is computed using a trapezoidal rule.
3. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a device to cause the device to:
receive data from the sensor;
create graph Laplacian of the data and store in the memory device;
compute spectral characteristic by applying density of states and detect spectral gaps in an unsupervised manner in the spectral characteristic to determine r number of clusters;
compute range space of a rational matrix of the graph Laplacian, r being determined based on dynamically increasing projection subspace for capturing eigenvalues, in which the subspace is built by computing the range space, and without requiring estimation of the eigenvalues located inside a disk in a complex plane;
train an unsupervised machine learning model based on the hyper-parameter r to cluster the received data, wherein to train the unsupervised machine learning model, the device is caused to perform K-means clustering on the range space of rational matrix of the graph Laplacian using r as the number of clusters, the K-means clustering trained to return r clusters of the received data;
the data including computer network traffic data, wherein the unsupervised machine learning model is trained to classify the computer network traffic data into r clusters of security levels; and
run the trained unsupervised machine learning model on incoming data traffic in real-time to detect a security level cluster to which the incoming data traffic belongs, and based on the security level cluster, filter the incoming data traffic to prevent unwanted data from entering a target computer system.
4. The computer program product of claim 3 , wherein the range space is computed using a trapezoidal rule.
5. A computer-implemented machine learning method comprising:
receiving data from the sensor;
creating graph Laplacian of the data and store in the memory device;
computing spectral characteristic by applying density of states and detect spectral gaps in an unsupervised manner in the spectral characteristic to determine r number of clusters, r being a hyper-parameter for machine learning;
computing a range space of a rational matrix of the graph Laplacian, r being determined based on dynamically increasing projection subspace for capturing eigenvalues, in which the subspace is built by computing the range space, and without requiring estimation of the eigenvalues located inside a disk in a complex plane;
training an unsupervised machine learning model based on the hyper-parameter r to cluster the received data, wherein to train the unsupervised machine learning model, the processor is configured to perform K-means clustering on the range space of rational matrix of the graph Laplacian using r as the number of clusters, the K-means clustering trained to return r clusters of the received data;
the data including computer network traffic data, wherein the unsupervised machine learning model is trained to classify the computer network traffic data into r clusters of security levels;
running the trained unsupervised machine learning model on incoming data traffic in real-time to detect a security level cluster to which the incoming data traffic belongs, and based on the security level cluster, filter the incoming data traffic to prevent unwanted data from entering a target computer system.
6. The method of claim 5 , wherein the range space is computed using a trapezoidal rule.