scholarly journals Automated Spectral Kernel Learning

2020 ◽  
Vol 34 (04) ◽  
pp. 4618-4625
Author(s):  
Jian Li ◽  
Yong Liu ◽  
Weiping Wang
Keyword(s):  
Error Bounds ◽  
Kernel Methods ◽  
Experimental Results ◽  
Kernel Learning ◽  
Generalization Error ◽  
Learning Framework ◽  
Model Training ◽  
Efficient Learning ◽  
Spectral Representations ◽  
Spectral Kernel

The generalization performance of kernel methods is largely determined by the kernel, but spectral representations of stationary kernels are both input-independent and output-independent, which limits their applications on complicated tasks. In this paper, we propose an efficient learning framework that incorporates the process of finding suitable kernels and model training. Using non-stationary spectral kernels and backpropagation w.r.t. the objective, we obtain favorable spectral representations that depends on both inputs and outputs. Further, based on Rademacher complexity, we derive data-dependent generalization error bounds, where we investigate the effect of those factors and introduce regularization terms to improve the performance. Extensive experimental results validate the effectiveness of the proposed algorithm and coincide with our theoretical findings.

Using API Call Sequences for IoT Malware Classification Based on Convolutional Neural Networks

2021 ◽  
Vol 31 (04) ◽  
pp. 587-612
Author(s):  
Qianguang Lin ◽  
Ni Li ◽  
Qi Qi ◽  
Jiabin Hu
Keyword(s):  
Time Series ◽  
Time Series Data ◽  
Comprehensive Evaluation ◽  
Optimal Solution ◽  
Classification Problem ◽  
Experimental Results ◽  
Series Data ◽  
Chi Square ◽  
Malware Classification ◽  
Model Training

Internet of Things (IoT) devices built on different processor architectures have increasingly become targets of adversarial attacks. In this paper, we propose an algorithm for the malware classification problem of the IoT domain to deal with the increasingly severe IoT security threats. Application executions are represented by sequences of consecutive API calls. The time series of data is analyzed and filtered based on the improved information gains. It performs more effectively than chi-square statistics, in reducing the sequence lengths of input data meanwhile keeping the important information, according to the experimental results. We use a multi-layer convolutional neural network to classify various types of malwares, which is suitable for processing time series data. When the convolution window slides down the time sequence, it can obtain higher-level positions by collecting different sequence features, thereby understanding the characteristics of the corresponding sequence position. By comparing the iterative efficiency of different optimization algorithms in the model, we select an algorithm that can approximate the optimal solution to a small number of iterations to speed up the convergence of the model training. The experimental results from real world IoT malware sample show that the classification accuracy of this approach can reach more than 98%. Overall, our method has demonstrated practical suitability for IoT malware classification with high accuracies and low computational overheads by undergoing a comprehensive evaluation.

FKPIndexNet: An efficient learning framework for finger-knuckle-print database indexing to boost identification

2021 ◽  
pp. 108028
Author(s):  
Geetika Arora ◽  
Avantika Singh ◽  
Aditya Nigam ◽  
Hari Mohan Pandey ◽  
Kamlesh Tiwari
Keyword(s):  
Learning Framework ◽  
Database Indexing ◽  
Efficient Learning

scholarly journals Multisignal VGG19 Network with Transposed Convolution for Rotating Machinery Fault Diagnosis Based on Deep Transfer Learning

2020 ◽  
Vol 2020 ◽  
pp. 1-12
Author(s):  
Jianye Zhou ◽  
Xinyu Yang ◽  
Lin Zhang ◽  
Siyu Shao ◽  
Gangying Bian
Keyword(s):  
Fault Diagnosis ◽  
Transfer Learning ◽  
High Efficiency ◽  
Image Resolution ◽  
Fine Tuning ◽  
Test Accuracy ◽  
Multiple Sensors ◽  
Learning Framework ◽  
Training Samples ◽  
Model Training

To realize high-precision and high-efficiency machine fault diagnosis, a novel deep learning framework that combines transfer learning and transposed convolution is proposed. Compared with existing methods, this method has faster training speed, fewer training samples per time, and higher accuracy. First, the raw data collected by multiple sensors are combined into a graph and normalized to facilitate model training. Next, the transposed convolution is utilized to expand the image resolution, and then the images are treated as the input of the transfer learning model for training and fine-tuning. The proposed method adopts 512 time series to conduct experiments on two main mechanical datasets of bearings and gears in the variable-speed gearbox, which verifies the effectiveness and versatility of the method. We have obtained advanced results on both datasets of the gearbox dataset. The dataset shows that the test accuracy is 99.99%, achieving a significant improvement from 98.07% to 99.99%.

Ensemble Pruning for Data Dependant Learners

2011 ◽  
Vol 135-136 ◽  
pp. 522-527 ◽  
Author(s):  
Gang Zhang ◽  
Shan Hong Zhan ◽  
Chun Ru Wang ◽  
Liang Lun Cheng
Keyword(s):  
Unlabeled Data ◽  
Experimental Results ◽  
Clear Relationship ◽  
Data Sets ◽  
Ensemble Pruning ◽  
Generalization Ability ◽  
Learning Framework ◽  
Single Instance ◽  
Better Than ◽  
Individual Kernel

Ensemble pruning searches for a selective subset of members that performs as well as, or better than ensemble of all members. However, in the accuracy / diversity pruning framework, generalization ability of target ensemble is not considered, and moreover, there is not clear relationship between them. In this paper, we proof that ensemble formed by members of better generalization ability is also of better generalization ability. We adopt learning with both labeled and unlabeled data to improve generalization ability of member learners. A data dependant kernel determined by a set of unlabeled points is plugged in individual kernel learners to improve generalization ability, and ensemble pruning is launched as much previous work. The proposed method is suitable for both single-instance and multi-instance learning framework. Experimental results on 10 UCI data sets for single-instance learning and 4 data sets for multi-instance learning show that subensemble formed by the proposed method is effective.

The Machine of Teaching Quality Valuation Studies an Analytical Method

2013 ◽  
Vol 303-306 ◽  
pp. 1510-1513
Author(s):  
Yun Xia Wang
Keyword(s):  
Machine Learning ◽  
Research Work ◽  
Teaching Quality ◽  
Experimental Results ◽  
Quality Analysis ◽  
Positive Contribution ◽  
Learning Framework ◽  
Machine Learning Methods ◽  
Network Method ◽  
Artificial Neural Network Method

This text describes the research work in machine learning framework for the assessment of teaching quality , mainly focused on the analysis of data on information technology in the teaching process , and the use of artificial neural network method, the experiment , the experimental results reflect the level of teaching quality analysis . Experimental results show that the use of machine learning methods can indeed make a positive contribution to the teaching quality assessment .

Sign in / Sign up

Export Citation Format

Share Document