Hands-On One-shot Learning with Python

Model architecture

The matching networks architecture is majorly inspired by the attention model and memory-based networks. In all these models, a neural attention mechanism is defined to access a memory matrix, which stores useful information to solve the task at hand. To begin with, first, we need to understand certain terminologies used in matching networks:

• Label set: This is the sample set of all possible categories. For example, if we are using the ImageNet dataset, it consists of thousands of categories (such as cats, dogs, and birds), but as part of the label set, we will only use five of those categories.
• Support set: This is the sampled input data points (for example, images) of our label set categories.
• Batch: Similar to the ...