What is different between Machine learning,Deep learning,transfer learning,meta learning,few shot learning, zero shot learning and one shot learning?

Question

我确实搜索了这个概念，但我无法清楚而简要地理解它们的区别，有人可以用简单的语言来描述它们的区别吗？

Answer 1

I think there are plenty of good explanations online (way better than mine), but I'll try to explain in on sentence each:

• ML: Train some kind of algorithm using labeled data eg. for classification
• DL: subclass of ML using big NN architectures to perform a task eg. classification
• TL: Use a pretrained architecture and fine-tune it for your use-case using your custom data
• Few/one/zero shot learning: Use only a few/one/zero new training samples to train your ML algorithm for a new task (used in the area of generalization)
• Meta learning: Not using eg. labeled data from supervisor but use output of exisiting ML algorithms to train your algorithm - can combine knowledge of different algortithms