ABSTRACT
Learning invariant representations via contrastive learning has seen state-of-the-art performance in domain generalization (DG). Despite such success, in this paper, we find that its core learning strategy - feature alignment - could heavily hinder model generalization. Drawing insights in neuron interpretability, we characterize this problem from a neuron activation view. Specifically, by treating feature elements as neuron activation states, we show that conventional alignment methods tend to deteriorate the diversity of learned invariant features, as they indiscriminately minimize all neuron activation differences. This instead ignores rich relations among neurons - many of them often identify the same visual concepts despite differing activation patterns. With this finding, we present a simple yet effective approach, Concept Contrast (CoCo), which relaxes element-wise feature alignments by contrasting high-level concepts encoded in neurons. Our CoCo performs in a plug-and-play fashion, thus it can be integrated into any contrastive method in DG. We evaluate CoCo over four canonical contrastive methods, showing that CoCo promotes the diversity of feature representations and consistently improves model generalization capability. By decoupling this success through neuron coverage analysis, we further find that CoCo potentially invokes more meaningful neurons during training, thereby improving model learning.
ABSTRACT
This paper focuses on the domain generalization task where domain knowledge is unavailable, and even worse, only samples from a single domain can be utilized during training. Our motivation originates from the recent progresses in deep neural network (DNN) testing, which has shown that maximizing neuron coverage of DNN can help to explore possible defects of DNN (i.e., misclassification). More specifically, by treating the DNN as a program and each neuron as a functional point of the code, during the network training we aim to improve the generalization capability by maximizing the neuron coverage of DNN with the gradient similarity regularization between the original and augmented samples. As such, the decision behavior of the DNN is optimized, avoiding the arbitrary neurons that are deleterious for the unseen samples, and leading to the trained DNN that can be better generalized to out-of-distribution samples. Extensive studies on various domain generalization tasks based on both single and multiple domain(s) setting demonstrate the effectiveness of our proposed approach compared with state-of-the-art baseline methods. We also analyze our method by conducting visualization based on network dissection. The results further provide useful evidence on the rationality and effectiveness of our approach.
ABSTRACT
In this paper, we have systematically studied the holographic fabrication of three-dimensional (3D) structures using a single 3D printed reflective optical element (ROE), taking advantage of the ease of design and 3D printing of the ROE. The reflective surface was setup at non-Brewster angles to reflect both s- and p-polarized beams for the interference. The wide selection of reflective surface materials and interference angles allow control of the ratio of s- and p-polarizations, and intensity ratio of side-beam to central beam for interference lithography. Photonic bandgap simulations have also indicated that both s and p-polarized waves are sometimes needed in the reflected side beams for maximum photonic bandgap size and certain filling fractions of dielectric inside the photonic crystals. The flexibility of single ROE and single exposure based holographic fabrication of 3D structures was demonstrated with reflective surfaces of ROEs at non-Brewster angles, highlighting the capability of the ROE technique of producing umbrella configurations of side beams with arbitrary angles and polarizations and paving the way for the rapid throughput of various photonic crystal templates.
