Week 5 — Topify
Last week we discussed on using one-shot learning for our project. But the problem with that was; one-shot learning takes one train data to test various other datas, hence the name. So we discussed that we should use few-shot learning.
Few-shot learning, or low-shot learning, is just a more flexible version of one-shot learning. With few-shot learning we can train our model with more than one train data and expect good results. Same with one-shot learning, few-shot learning is also mainly used for computer vision or natural language processing. There isn’t much examples we can look or articles we can read to use few-shot learning for our dataset which is not images or sentences. So we came up with an idea. We will draw the radar charts of our data. Then we will train our model with these images. For example
We tried few-shot learning with an example dataset called omniglot dataset, which is pretty famous for these kind of problems, and we figured it out how to implement it for our dataset.