Neighborhood Preserving and Weighted Subspace Learning Method for Drift Compensation in Gas Sensor
ABSTARCT :
This article presents a novel discriminative subspace-learning-based unsupervised domain adaptation (DA) method for the gas sensor drift problem. Many existing subspace learning approaches assume that the gas sensor data follow a certain distribution such as Gaussian, which often does not exist in real-world applications. In this article, we address this issue by proposing a novel discriminative subspace learning method for DA with neighborhood preserving (DANP). We introduce two novel terms, including the intraclass graph term and the interclass graph term, to embed the graphs into DA. Besides, most existing methods ignore the influence of the subspace learning on the classifier design. To tackle this issue, we present a novel classifier design method (DANP+) that incorporates the DA ability of the subspace into the learning of the classifier. The weighting function is introduced to assign different weights to different dimensions of the subspace. We have verified the effectiveness of the proposed methods by conducting experiments on two public gas sensor datasets in comparison with the state-of-the-art DA methods.
EXISTING SYSTEM :
? This can effectively avoid the so-called dimensional disaster, in which the expected sum converges to a constant due to the existence of a large number of independent components in the data.
? However, this method does not exploit the label information in the source data and ignores the case that multimodal data distribution might exist.
? If we use particles instead of birds to describe this process, each particle in the model is an individual in the N-dimensional space to search in the space.
? The current position of each particle in the particle swarm can be regarded as a candidate solution of the current optimization problem.
DISADVANTAGE :
? Machine learning tools, particularly, domain adaptation (DA) methods, are utilized to deal with the gas sensor drift problems.
? However, the gas sensor drift problem is quite common due to the environment change, such as temperature and humidity variations.
? The focus of this article is on the subspace-learning-based unsupervised DA methods for the gas sensor drift problem, where we have no access to the label information of gas sensor data from the target domain.
? To be specific, we proposed a novel discriminative subspace-learning-based DA method to conquer the gas sensor drift problem.
? The novel SVM classifier can be efficiently solved by reformulating the problem as a smooth support vector machine (SSVM).
PROPOSED SYSTEM :
• In this article, a local discriminant subspace projection (LDSP) method is proposed to tackle the gas sensor drift problem.
• The proposed approach is a significant extension of a recently proposed subspace projection approach, i.e., domain regularization component analysis (DRCA) in.
• The proposed approach exploits the class label information of the source data in order to reduce the possibility of the case that data with different labels in the latent subspace stay close to each other.
• We have proposed to introduce the locality-preserving projection for the subspace learning to deal with multimodal data.
• We have shown that the proposed approach performs better than the state-of-the-art methods on two public gas sensor drift datasets.
ADVANTAGE :
? The proposed methods inherited the merits of extreme learning machines, i.e., high efficiency of computation.
? We will take a more complicated technique of feature augmentation into consideration in the future work for further performance improvement.
? However, the deep learning-based methods cannot be better than the proposed methods on the average recognition performance.
? LDSP has a better performance than DANP, which means that DANP might not be good enough to deal with the multimodal data.
? This technique is also used in other DA methods such as the transfer component analysis (TCA).
|