P-KDGAN: Progressive Knowledge Distillation with GANs for One-class Novelty Detection
This work addresses the challenge of efficient novelty detection for deployment on constrained devices, presenting an incremental improvement over existing GAN-based methods.
The paper tackles the problem of deploying over-parameterized GANs for one-class novelty detection on resource-limited devices by proposing P-KDGAN, a progressive knowledge distillation method that transfers knowledge from a teacher to a student GAN, resulting in performance improvements of 2.44%, 1.77%, and 1.73% on CIFAR-10, MNIST, and FMNIST datasets while achieving compression ratios of 24.45:1, 311.11:1, and 700:1, respectively.
One-class novelty detection is to identify anomalous instances that do not conform to the expected normal instances. In this paper, the Generative Adversarial Networks (GANs) based on encoder-decoder-encoder pipeline are used for detection and achieve state-of-the-art performance. However, deep neural networks are too over-parameterized to deploy on resource-limited devices. Therefore, Progressive Knowledge Distillation with GANs (PKDGAN) is proposed to learn compact and fast novelty detection networks. The P-KDGAN is a novel attempt to connect two standard GANs by the designed distillation loss for transferring knowledge from the teacher to the student. The progressive learning of knowledge distillation is a two-step approach that continuously improves the performance of the student GAN and achieves better performance than single step methods. In the first step, the student GAN learns the basic knowledge totally from the teacher via guiding of the pretrained teacher GAN with fixed weights. In the second step, joint fine-training is adopted for the knowledgeable teacher and student GANs to further improve the performance and stability. The experimental results on CIFAR-10, MNIST, and FMNIST show that our method improves the performance of the student GAN by 2.44%, 1.77%, and 1.73% when compressing the computation at ratios of 24.45:1, 311.11:1, and 700:1, respectively.