Transfer channel pruning for compressing deep domain adaptation models

Deep unsupervised domain adaptation has recently received increasing attention from researchers. However, existing methods are computationally intensive due to the computational cost of convolutional neural networks (CNN) adopted by most work. There is no effective network compression method for such problem. In this paper, we propose a unified transfer channel pruning (TCP) method for accelerating deep unsupervised domain adaptation (UDA) models. TCP method is capable of compressing the deep UDA model by pruning less important channels while simultaneously learning transferable features by reducing the cross-domain distribution divergence. Therefore, it reduces the impact of negative transfer and maintains competitive performance on the target task. To the best of our knowledge, TCP method is the first approach that aims at accelerating deep unsupervised domain adaptation models. TCP method is validated on two main kinds of UDA methods: the discrepancy-based methods and the adversarial-based methods. In addition, it is validated on two benchmark datasets: Office-31 and ImageCLEF-DA with two common backbone networks - VGG16 and ResNet50. Experimental results demonstrate that our TCP method achieves comparable or better classification accuracy than other comparison methods while significantly reducing the computational cost. To be more specific, in VGG16, we get even higher accuracy after pruning 26% floating point operations (FLOPs); in ResNet50, we also get higher accuracy on half of the tasks after pruning 12% FLOPs for both discrepancy-based methods and adversarial-based methods.