=================================================================================

Few-shot learning (FSL), also referred to as low-shot learning (LSL), is a type of machine learning method where the training dataset contains limited information. Few shot learning aims to build accurate machine learning models with limited training data. As the dimension of input data is a factor that determines resource costs (e.g. time costs, computational costs, etc.), companies can reduce data analysis/machine learning (ML) costs by using few shot learning. Few-shot learning paradigm can be used to the cases that a few wafer images are available for each class, while the self-supervision information containing the intrinsic correlations of unlabeled wafer maps and their augmentations is expected to enhance the few-shot learner. [1]

Few-shot learning is a kind of meta learning. Metal learning is learn-to-learn. The goal of few-shot learning is to learn representations that generalize well to the minority defect pattern classes in which only a few wafer images are available. In the few-shot learning, a small amount of data from different classes are sampled evenly in one training batch, which is equivalent to the resampling tactics to combat imbalanced training data. By leveraging the self-supervised learning, the unlabeled wafer maps can analyzed easily.

Metal learning algorithms are:
    i) Gradient-based meta learning
       i.a) Model-agnostic meta learning (MAML)
       i.b) Meta-LSTM
       i.c) Reptile
     ii) Metric learning
        ii.a) Matching network
        ii.b) Prototypical network
         ii.c) Relation network

Table 4244. Application of few-shot learning (FSL).

============================================

[1] Hao Geng, Fan Yang, Xuan Zeng, Bei Yu, When Wafer Failure Pattern Classification Meets Few-shot Learning and Self-Supervised Learning, DOI: 10.1109/ICCAD51958.2021.9643518, 2021 IEEE/ACM International Conference On Computer Aided Design (ICCAD), 2021.