Research On Semi-supervised Dimensionality Reduction Algorithms And Its Applications In Medical Expert System

With the development of information technology in recent years, theemerging of high-dimensional data in the scientific community as well asindustry is becoming more and more frequently, especially the high-techfield of computer vision, pattern recognition, aerospace, bio-informatics.The these data Witt often become obstacles when we deal with relatedalgorithms are facing high complexity cause the calculation results are notoptimal. Dimensionality reduction through the data with high dimensionalapproximately reduced to a low-dimensional to reveal the inherentlow-dimensional structure of the data itself. It as a way to overcome the"curse of dimensionality" plays an important role in many areas.In the past few decades, many dimensionality reduction algorithm isproposed by many researchers in-depth study, like traditionaldimensionality reduction algorithm PCA and LDA; manifold learningalgorithm LLE, LED, of ISOMAP, and LTSA. With the development ofinformation technology, the amount of data is growing, many times there isa sample of the label is often not easy to get, most of the unlabeled samples,which also makes the semi-supervised ideological dimensionalityreduction more widely.LLE algorithm is a classic in the field of manifold learning algorithm,but its dimension reduction for many types of samples mixed data set isnot very good. The author based on the more popular learning based onimproved LLE, and needs of the project incorporates the idea ofsemi-supervised learning, and its semi-supervised version.The tag propagation algorithm (LP) is a graph-based semi-supervisedlearning algorithm, its core idea is to maintain a certain structure between the raw data, some label sample label information to unlabeled samplespass through some of the methods, until a global stable state. Based labelpropagation algorithm based on a combination of thinking locally linearembedding local reconstruction algorithm, improved local alignment basedon discriminant information, raised its semi-supervised version of thepopular structure on the reconstruction of the right part of the labelinformation known sample label and dissemination, and use ofcommunication of all data the soft label information as categoryinformation. And experimental comparative analysis of the improvedalgorithm and the traditional semi-supervised dimensionality reductionalgorithm.In recent years, medical aided diagnosis system has been a very hot topic,its powerful ability to get the doctor and the patient certainly aideddiagnosis plays a very important role in promoting the development ofmedical diagnostic cases data particularity, and also to promote the work ofdata dimensionality reduction, feature selection, and a series of datamining. The most widely used in many medical aided diagnosis systemwas undoubtedly aided diagnosis system based on the image as well ascase-based input feature aided diagnosis system. Due to strong noisecharacteristics of the ultrasound image, in the image-based secondarydiagnosis is still exploring stage, but based on the input-aided diagnosiseffect than satisfactory. This article is for input case characteristics ofultrasound medical image diagnosis...