King Abdulaziz University

Siya Yao

Yusuf Abdulaziz Al-Turki

Mengchu Zhou

Abdullah Abusorrah

Qi Kang

Kai Zhang

An AUC-maximized high-accuracy classification method and system for imbalanced datasets integrates an under-sampling-and-ensemble strategy, a true-outliers-removing strategy and a fake-outliers-concealing strategy, with the hope to effectively and robustly enhance both the AUC and the accuracy metrics in imbalanced classification. Applying under-sampling to construct multiple sub-datasets and assembling classification results of multiple classifiers greatly decline the risk of misclassification and lead to highly accurate and robust results in imbalanced classification task. Moreover, this invention pays attention to detect and identify extremely hidden outliers in a sub-dataset which includes a sub-majority dataset and the entire minority dataset. In this way, more hidden outliers can be located and thus exert less influence on the decision boundary, which contributes to both high AUC and accuracy. Furthermore, this invention proposes to conceal fake outliers when building decision boundary, which can achieve a higher classification accuracy of the majority class without changing that of the minority class.

AUC-maximized high-accuracy classifier for imbalanced datasets

In the field of computer vision, without sufficient labeled images, it is challenging to train an accurate model. But through visual adaptation from source to target domains, a relevant labeled dataset can help solve such problem. Many methods apply adversarial learning to diminish cross-domain distribution difference. They are able to greatly enhance the performance on target classification tasks. GAN (Generative Adversarial Networks) loss is widely used in adversarial adaptation learning methods to reduce a across-domain distribution difference. However, it becomes difficult to decline such distribution difference if generator or discriminator in GAN fails to work as expected and degrades its performance. To solve such cross-domain classification problems, an adaptation algorithm and system called as Generative Adversarial Distribution Matching (GADM) is implemented. In GADM, the objective function is improved by taking cross-domain discrepancy distance into consideration, and further minimize the difference through the competition between the generator and discriminator, thereby greatly decreasing the cross-domain distribution difference. Even when the performance of its generator or discriminator degrades, GADM is capable of decreasing the cross-domain distribution difference. The GADM algorithm and system employs a single GAN framework so as to achieve faster domain adaption with less computation resource. Specially, GADM transfers target data distribution to source one to keep accurate label dependence information, which ensures high accuracy and stability of source classifier and thus achieves better classification performance on target data.

System for fast and accurate visual domain adaptation

To ensure photovoltaic (PV) panels at a PV plant constantly operate in an ideal state, an unsupervised system and method of intelligent performance evaluation and data-driven fault detection is used. The system and method enables engineers to check PV panels in time and implement timely maintenance. Monitored data are classified into three subsets: ideal period A, transition period S, and downturn period B. Based on A and B data, two regression prediction models are built which are tree-based and fit the non-continuous PV data well. Real-time measured power is compared with upper and lower reference baselines derived from the two predictive models, respectively. Using threshold ranges, the system and method achieves the instantaneous performance monitoring of PV power generation, and provides failure identification and operation and maintenance (O&M) suggestions to engineers.

Growing community of inventors

Shanghai, China

Siya Yao