Biomedical subjects

Keshu Zhang

Publications and source records attributed to Keshu Zhang.

4 recordsLinked to original sources

PubMed2006

Efficient and robust feature extraction by maximum margin criterion.

In pattern recognition, feature extraction techniques are widely employed to reduce the dimensionality of data and to enhance the discriminatory information. Principal component analysis (PCA) and linear discriminant analysis (LDA) are the two most popular linear dimensionality reduction methods. However, PCA is not very effective for the extraction of the most discriminant features, and LDA is not stable due to the small sample size problem. In this paper, we propose some new (linear and nonlinear) feature extractors based on maximum margin criterion (MMC). Geometrically, feature extractors based on MMC maximize the (average) margin between classes after dimensionality reduction. It is shown that MMC can represent class separability better than PCA. As a connection to LDA, we may also derive LDA from MMC by incorporating some constraints. By using some other constraints, we establish a new linear feature extractor that does not suffer from the small sample size problem, which is known to cause serious stability problems for LDA. The kernelized (nonlinear) counterpart of this linear feature extractor is also established in the paper. Our extensive experiments demonstrate that the new feature extractors are effective, stable, and efficient.

Algorithms↗

PubMed2006

A general framework for biclustering gene expression data.

A large number of biclustering methods have been proposed to detect patterns in gene expression data. All these methods try to find some type of biclusters but no one can discover all the types of patterns in the data. Furthermore, researchers have to design new algorithms in order to find new types of biclusters/patterns that interest biologists. In this paper, we propose a novel approach for biclustering that, in general, can be used to discover all computable patterns in gene expression data. The method is based on the theory of Kolmogorov complexity. More precisely, we use Kolmogorov complexity to measure the randomness of submatrices as the merit of biclusters because randomness naturally consists in a lack of regularity, which is a common property of all types of patterns. On the basis of algorithmic probability measure, we develop a Markov Chain Monte Carlo algorithm to search for biclusters. Our method can also be easily extended to solve the problems of conventional clustering and checkerboard type biclustering. The preliminary experiments on simulated as well as real data show that our approach is very versatile and promising.

Algorithms↗

PubMed2005

Robust and accurate cancer classification with gene expression profiling.

Robust and accurate cancer classification is critical in cancer treatment. Gene expression profiling is expected to enable us to diagnose tumors precisely and systematically. However, the classification task in this context is very challenging because of the curse of dimensionality and the small sample size problem. In this paper, we propose a novel method to solve these two problems. Our method is able to map gene expression data into a very low dimensional space and thus meets the recommended samples to features per class ratio. As a result, it can be used to classify new samples robustly with low and trustable (estimated) error rates. The method is based on linear discriminant analysis (LDA). However, the conventional LDA requires that the within-class scatter matrix S(w) be nonsingular. Unfortunately, Sw is always singular in the case of cancer classification due to the small sample size problem. To overcome this problem, we develop a generalized linear discriminant analysis (GLDA) that is a general, direct, and complete solution to optimize Fisher's criterion. GLDA is mathematically well-founded and coincides with the conventional LDA when S(w) is nonsingular. Different from the conventional LDA, GLDA does not assume the nonsingularity of S(w), and thus naturally solves the small sample size problem. To accommodate the high dimensionality of scatter matrices, a fast algorithm of GLDA is also developed. Our extensive experiments on seven public cancer datasets show that the method performs well. Especially on some difficult instances that have very small samples to genes per class ratios, our method achieves much higher accuracies than widely used classification methods such as support vector machines, random forests, etc.

Algorithms↗

PubMed2004

Minimum entropy clustering and applications to gene expression analysis.

Clustering is a common methodology for analyzing the gene expression data. In this paper, we present a new clustering algorithm from an information-theoretic point of view. First, we propose the minimum entropy (measured on a posteriori probabilities) criterion, which is the conditional entropy of clusters given the observations. Fano's inequality indicates that it could be a good criterion for clustering. We generalize the criterion by replacing Shannon's entropy with Havrda-Charvat's structural alpha-entropy. Interestingly, the minimum entropy criterion based on structural alpha-entropy is equal to the probability error of the nearest neighbor method when alpha = 2. This is another evidence that the proposed criterion is good for clustering. With a non-parametric approach for estimating a posteriori probabilities, an efficient iterative algorithm is then established to minimize the entropy. The experimental results show that the clustering algorithm performs significantly better than k-means/medians, hierarchical clustering, SOM, and EM in terms of adjusted Rand index. Particularly, our algorithm performs very well even when the correct number of clusters is unknown. In addition, most clustering algorithms produce poor partitions in presence of outliers while our method can correctly reveal the structure of data and effectively identify outliers simultaneously.

Algorithms↗