A novel logistic regression model combining semi-supervised learning and active learning for disease classification

Abstract

Traditional supervised learning classifier needs a lot of labeled samples to achieve good performance, however in many biological datasets there is only a small size of labeled samples and the remaining samples are unlabeled. Labeling these unlabeled samples manually is difficult or expensive. Technologies such as active learning and semi-supervised learning have been proposed to utilize the unlabeled samples for improving the model performance. However in active learning the model suffers from being short-sighted or biased and some manual workload is still needed. The semi-supervised learning methods are easy to be affected by the noisy samples. In this paper we propose a novel logistic regression model based on complementarity of active learning and semi-supervised learning, for utilizing the unlabeled samples with least cost to improve the disease classification accuracy. In addition to that, an update pseudo-labeled samples mechanism is designed to reduce the false pseudo-labeled samples. The experiment results show that this new model can achieve better performances compared the widely used semi-supervised learning and active learning methods in disease classification and gene selection.

Image Processing and Machine Learning for Automated Identification of Chemo-/Biomarkers in Chromatography–Mass Spectrometry

“…Consequently, the number of training data increased by five times the primary ones. The discrimination of the original aroma odor/breath samples and the molecule additive-containing samples and the calculation of the feature score for each data set were performed by the logistic regression model (Figure e) . Machine learning was performed to optimize the following equation: log p /(1 – p ) = β 0 + x 1 β 1 + x 2 β 2 + x 3 β 3 + ... + x n β n , where p is the probability of which the data sets can be classified, x n is the intensity of each segment in the 2D MS map, β n ( n ≥ 1) is the model’s learned weight (i.e., feature score), and β 0 is the bias.…”

Concepts of Artificial Intelligence for Computer-Assisted Drug Discovery

“…While conventional semisupervised approaches attempt to exploit the latent structure of unlabeled data with the specific goal of improving label predictions, the goal of active learning is to reduce the number of labeled examples needed for learning at the same time. 78 As an active research area, there are many different active learning methods, such as pool-based active learning, 79 where the learner chooses which sample to label next in a pool of unlabeled data; selective sampling, 80,81 where unlabeled data come as a stream and the learner decides to query or discard each arriving point; and batch-mode active learning, 82 where the learner queries multiple instances at once. Interested readers may refer to a report by Settles 83 to learn more about the details of the query strategy frameworks.…”

Judging a Book by Its Cover: The Effect of Facial Perception on Centrality in Social Networks

“…The idea of this method is very naive, namely that if samples are closer in the feature space, they are more likely to be in the same class. Then, we observed two linear classification modelsâĂŹ performance, linear support vector classifier (linear SVC) [3,17] and logistic regression (LR) [7]. Both identify hyperplanes in the feature space to split samples into different classes.…”