机器学习（八）
meta learning
Meta Learning and Its Applications to Human Language Processing (ntu.edu.tw)
和机器学习的3个step类似
Using the optimization approach you know
If you know how to compute 𝜕𝐿(𝜙)/𝜕𝜙
Gradient descent is your friend
What if 𝐿(𝜙) is not differentiable?
Reinforcement Learning / Evolutionary Algorithm
few shot learning和meta learning的goal 不一样 (用meta去完成few shot. few shot 是目标，meta是手段）
meta：testing task is what we really care about.
Machine Learning ≈ find a function f
within task testing
Meta Learning ≈ find a function F that finds a function f
Training Tasks产生一个learned “learning algorithm”
within task training + within task testing
loss
What you know about ML can usually apply to meta learning
• Overfitting on training tasks • Get more training tasks to improve performance • Task augmentation • There are also hyperparameters when learning a learning algorithm …… • Development task
What is learnable in a learning algorithm?
Model-Agnostic Meta-Learning (MAML)
NAS
data processing
Data Augmentation
Sample Reweighting
Beyond Gradient Descent
application
- Few-shot Image Classification
课程总结
李宏毅《机器学习》（2021） (xiaoe-tech.com)
