Bettering Cross-domain Few-shot Classification with Multilayer Perceptron
Authors: Shuanghao Bai, Wanqi Zhou, Zhirong Luan, Donglin Wang, Badong Chen
Abstract: Cross-domain few-shot classification (CDFSC) is a tough and tough course of because of vital distribution discrepancies all through completely totally different domains. To deal with this downside, many approaches intention to review transferable representations. Multilayer perceptron (MLP) has confirmed its performance to review transferable representations in assorted downstream duties, akin to unsupervised image classification and supervised concept generalization. Nonetheless, its potential inside the few-shot settings has however to be comprehensively explored. On this analysis, we study the potential of MLP to assist in addressing the challenges of CDFSC. Significantly, we introduce three distinct frameworks incorporating MLP in accordance with three sorts of few-shot classification methods to substantiate the effectiveness of MLP. We reveal that MLP can significantly enhance discriminative capabilities and alleviate distribution shifts, which might be supported by our expensive experiments involving 10 baseline fashions and 12 benchmark datasets. Furthermore, our method even compares favorably in opposition to totally different state-of-the-art CDFSC algorithms.
