Chongyangzi Teng

Pengwei Yang, Chongyangzi Teng, Jack George Mangos

Label noise is a significant obstacle in deep learning model training. It can have a considerable impact on the performance of image classification models, particularly deep neural networks, which are especially susceptible because they have a strong propensity to memorise noisy labels. In this paper, we have examined the fundamental concept underlying related label noise approaches. A transition matrix estimator has been created, and its effectiveness against the actual transition matrix has been demonstrated. In addition, we examined the label noise robustness of two convolutional neural network classifiers with LeNet and AlexNet designs. The two FashionMINIST datasets have revealed the robustness of both models. We are not efficiently able to demonstrate the influence of the transition matrix noise correction on robustness enhancements due to our inability to correctly tune the complex convolutional neural network model due to time and computing resource constraints. There is a need for additional effort to fine-tune the neural network model and explore the precision of the estimated transition model in future research.

Pengwei Yang, Chongyangzi Teng, Jack George Mangos

Non-negative matrix factorisation (NMF) has been extensively applied to the problem of corrupted image data. Standard NMF approach minimises Euclidean distance between data matrix and factorised approximation. The traditional NMF technique is sensitive to outliers since it utilises the squared error of each data point, despite the fact that this method has proven effective. In this study, we theoretically examine the robustness of the traditional NMF, HCNMF, and L2,1-NMF algorithms and execute sets of experiments to demonstrate the robustness on ORL and Extended YaleB datasets. Our research indicates that each algorithm requires a different number of iterations to converge. Due to the computational cost of these approaches, our final models, such as the HCNMF and L2,1-NMF model, fail to converge within the iteration parameters of this work. Nonetheless, the experimental results illustrate, to some extent, the robustness of the aforementioned techniques.