search

cbfinn / maml

Code for "Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks"
MIT License
2.55k stars 604 forks source link

First-order approximation #77

Open AceChuse opened 4 years ago

AceChuse commented 4 years ago

Thanks for your works! It is really great idea. There are one thing I want to ask. In paper, you mentioned the first-order approximation of MAML. I have seen details about it or the name of algorithm. Can you talk more details about it?

AceChuse commented 4 years ago

In addtion, did you try to use a large scale neural network training by a first-order approximation. Can it get a better result on dataset like miniImageNet?

Runist commented 4 years ago

In addtion, did you try to use a large scale neural network training by a first-order approximation. Can it get a better result on dataset like miniImageNet?

I think you can read "How to train your MAML".This paper asks five questions about MAML.

AceChuse commented 4 years ago

In addtion, did you try to use a large scale neural network training by a first-order approximation. Can it get a better result on dataset like miniImageNet?

I think you can read "How to train your MAML".This paper asks five questions about MAML.

Thank you for your response! I have read "How to train your MAML". However, what I want to know is the math process of the first-order approximation. Would you tell more detail about it?

Runist commented 4 years ago

In addtion, did you try to use a large scale neural network training by a first-order approximation. Can it get a better result on dataset like miniImageNet?

I think you can read "How to train your MAML".This paper asks five questions about MAML.

Thank you for your response! I have read "How to train your MAML". However, what I want to know is the math process of the first-order approximation. Would you tell more detail about it?

Sorry about that.I just write a code and training it.I'm not good at math, so I can't tell you.

AceChuse commented 4 years ago

In addtion, did you try to use a large scale neural network training by a first-order approximation. Can it get a better result on dataset like miniImageNet?

I think you can read "How to train your MAML".This paper asks five questions about MAML.

Thank you for your response! I have read "How to train your MAML". However, what I want to know is the math process of the first-order approximation. Would you tell more detail about it?

Sorry about that.I just write a code and training it.I'm not good at math, so I can't tell you.

Can I see the code of this process and how to use? I ask for the math process because I cannot find the code. So if I know how you write the code, I can write the math process.

Runist commented 4 years ago

In addtion, did you try to use a large scale neural network training by a first-order approximation. Can it get a better result on dataset like miniImageNet?

I think you can read "How to train your MAML".This paper asks five questions about MAML.

Thank you for your response! I have read "How to train your MAML". However, what I want to know is the math process of the first-order approximation. Would you tell more detail about it?

Sorry about that.I just write a code and training it.I'm not good at math, so I can't tell you.

Can I see the code of this process and how to use? I ask for the math process because I cannot find the code. So if I know how you write the code, I can write the math process.

可以，老哥还是说汉语吧。顶不住了，但其实我写的代码没有用到二阶导，其实我也对这一段挺好奇的。code

AceChuse commented 4 years ago

好吧，所以我才希望看到她本人来回答一下。

Runist commented 4 years ago

好吧，所以我才希望看到她本人来回答一下。

根据李宏毅教授的讲解，他是把二阶导近似等效为0或1。但是这样对结果的精度不好。

AceChuse commented 4 years ago

好吧，所以我才希望看到她本人来回答一下。

根据李宏毅教授的讲解，他是把二阶导近似等效为0或1。但是这样对结果的精度不好。

这个有讲义资料吗？求地址啊

Runist commented 4 years ago

好吧，所以我才希望看到她本人来回答一下。

根据李宏毅教授的讲解，他是把二阶导近似等效为0或1。但是这样对结果的精度不好。

这个有讲义资料吗？求地址啊

B站上搜“李宏毅”第一个2020的你拉到下面有个Meta-Learning章节的，就是了

AceChuse commented 4 years ago

我看到了，十分感谢！