May 2009 - May 2017 论坛八周年-你的足迹,我的骄傲


一亩三分地论坛

 找回密码
 获取更多干活,快来注册

一亩三分地官方iOS手机应用下载
查看: 5600|回复: 21
收起左侧

[DM|ML|计算] 不定时的看看深度学习deep learning

[复制链接] |试试Instant~ |关注本帖
小K 发表于 2016-9-1 13:10:31 | 显示全部楼层 |阅读模式

注册一亩三分地论坛,查看更多干货!

您需要 登录 才可以下载或查看,没有帐号?获取更多干活,快来注册

x
目的:不做研究,就是自己学起来,看看到底有什么用。不做严格计划了。

一楼贴总结和资料

==================================. 1point 3acres 璁哄潧
Hands on practice with kaggle https://www.kaggle.com/c/facial- ... p-learning-tutorial
Or just follow the udacity MOOC's assignment with tensorflow https://github.com/tensorflow/te ... ow/examples/udacity. visit 1point3acres.com for more.
Books. from: 1point3acres.com/bbs
先看这本书 http://www.deeplearningbook.org 鏉ユ簮涓浜.涓夊垎鍦拌鍧.
Microsoft book https://www.microsoft.com/en-us/ ... ng-Vol7-SIG-039.pdf
Tutorial
http://deeplearning.stanford.edu/tutorial/

Potential MOOC
Udacity: https://www.udacity.com/course/deep-learning--ud730
--- Finished the video part, it's a very very gentle intro, certainly not bad, but certainly not enough by itself.
to start in 2016/9: coursera: https://www.coursera.org/learn/neural-networks
Udacity还有一门reinforcement learning, 先放这里 https://www.udacity.com/course/reinforcement-learning--ud600-google 1point3acres
Udacity self driving car nano degree: to start

Stanford undergrad-google 1point3acres
visual recognition http://cs231n.stanford.edu/
NLP http://cs224d.stanford.edu/

======================================

Andrew Ng
The andrew Ng course on ML is obviously all good, but my fav is this section:
https://www.coursera.org/learn/machine-learning/lecture/x62iE/error-analysis
This really is a largely ignored part of ML system design, the problem is so commonly seen in practice


contrast this with a newer version of his talk with more emphasis on DL
Nuts and Bolts of Applying Deep Learning (Andrew Ng) - https://youtu.be/F1ka6a13S9I

fantastic stuff
=======================================
youtube上名人讲DL也是大堆大堆的,对于不喜欢看太多字,但是喜欢听人讲的来说也不错
Deep learning summer school 2016: http://videolectures.net/deeplearning2016_montreal
==================================
2016 deep learning school:. from: 1point3acres.com/bbs
From youtube:

Published on Sep 27, 2016
The talks at the Deep Learning School on September 24/25, 2016 were amazing. I clipped out individual talks from the full live streams and provided links to each below in case that's useful for people who want to watch specific talks several times (like I do). Please check out the official website (http://www.bayareadlschool.org) and full live streams below.

Having read, watched, and presented deep learning material over the past few years, I have to say that this is one of the best collection of introductory deep learning talks I've yet encountered. Here are links to the individual talks and the full live streams for the two days:

1. Foundations of Deep Learning (Hugo Larochelle, Twitter) - https://youtu.be/zij_FTbJHsk
2. Deep Learning for Computer Vision (Andrej Karpathy, OpenAI) - https://youtu.be/u6aEYuemt0M
3. Deep Learning for Natural Language Processing (Richard Socher, Salesforce) - https://youtu.be/oGk1v1jQITw
4. TensorFlow Tutorial (Sherry Moore, Google Brain) - https://youtu.be/Ejec3ID_h0w. visit 1point3acres.com for more.
5. Foundations of Unsupervised Deep Learning (Ruslan Salakhutdinov, CMU) - https://youtu.be/rK6bchqeaN8
6. Nuts and Bolts of Applying Deep Learning (Andrew Ng) - https://youtu.be/F1ka6a13S9I
7. Deep Reinforcement Learning (John Schulman, OpenAI) - https://youtu.be/PtAIh9KSnjo
8. Theano Tutorial (Pascal Lamblin, MILA) - https://youtu.be/OU8I1oJ9HhI
9. Deep Learning for Speech Recognition (Adam Coates, Baidu) - https://youtu.be/g-sndkf7mCs
10. Torch Tutorial (Alex Wiltschko, Twitter) - https://youtu.be/L1sHcj3qDNc.鏈枃鍘熷垱鑷1point3acres璁哄潧
11. Sequence to Sequence Deep Learning (Quoc Le, Google) - https://youtu.be/G5RY_SUJih4
12. Foundations and Challenges of Deep Learning (Yoshua Bengio) - https://youtu.be/11rsu_WwZTc

Full Day Live Streams: 鏉ユ簮涓浜.涓夊垎鍦拌鍧.
Day 1: https://youtu.be/eyovmAtoUx0
Day 2: https://youtu.be/9dXiAecyJrY

Go to http://www.bayareadlschool.org for more information on the event, speaker bios, slides, etc. Huge thanks to the organizers (Shubho Sengupta et al) for making this event happen.

==================================
Learning Goal
个人对application更感兴趣,尤其在除了NLP, img/audio之外,在传统统计学习,机器学习领域的应用
Disclaimer: I am a "traditional" data scientist/statistician, with hands-on experience in traditional ML, so I will tend to approach DL from this angle, and not from a traditional "CS" angle.

鏉ユ簮涓浜.涓夊垎鍦拌鍧.




. 鍥磋鎴戜滑@1point 3 acres
-google 1point3acres

评分

4

查看全部评分

 楼主| 小K 发表于 2016-9-1 16:57:27 | 显示全部楼层
关注一亩三分地公众号:
Warald_一亩三分地
The andrew Ng course on ML is obviously all good, but my fav is this section:
https://www.coursera.org/learn/machine-learning/lecture/x62iE/error-analysis
This really is a largely ignored part of ML system design, the problem is so commonly seen in practice


contrast this with a newer version of his talk with more emphasis on DL
Nuts and Bolts of Applying Deep Learning (Andrew Ng) - https://youtu.be/F1ka6a13S9I
. visit 1point3acres.com for more.
fantastic stuff
回复 支持 反对

使用道具 举报

18612824255 发表于 2016-9-1 20:47:02 | 显示全部楼层
关注一亩三分地微博:
Warald
非常感谢分享哇 现在在申请中 感觉对ps的帮助很大哈
回复 支持 反对

使用道具 举报

Rocketman456 发表于 2016-9-2 23:16:40 | 显示全部楼层
本帖最后由 Rocketman456 于 2016-9-2 23:18 编辑

感谢K姐分享!
回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-10-18 23:24:52 | 显示全部楼层
2016 deep learning school:
From youtube:

Published on Sep 27, 2016
The talks at the Deep Learning School on September 24/25, 2016 were amazing. I clipped out individual talks from the full live streams and provided links to each below in case that's useful for people who want to watch specific talks several times (like I do). Please check out the official website (http://www.bayareadlschool.org) and full live streams below.

Having read, watched, and presented deep learning material over the past few years, I have to say that this is one of the best collection of introductory deep learning talks I've yet encountered. Here are links to the individual talks and the full live streams for the two days:

1. Foundations of Deep Learning (Hugo Larochelle, Twitter) - https://youtu.be/zij_FTbJHsk
2. Deep Learning for Computer Vision (Andrej Karpathy, OpenAI) - https://youtu.be/u6aEYuemt0M
3. Deep Learning for Natural Language Processing (Richard Socher, Salesforce) - https://youtu.be/oGk1v1jQITw
4. TensorFlow Tutorial (Sherry Moore, Google Brain) - https://youtu.be/Ejec3ID_h0w
5. Foundations of Unsupervised Deep Learning (Ruslan Salakhutdinov, CMU) - https://youtu.be/rK6bchqeaN8
6. Nuts and Bolts of Applying Deep Learning (Andrew Ng) - https://youtu.be/F1ka6a13S9I
7. Deep Reinforcement Learning (John Schulman, OpenAI) - https://youtu.be/PtAIh9KSnjo
8. Theano Tutorial (Pascal Lamblin, MILA) - https://youtu.be/OU8I1oJ9HhI
9. Deep Learning for Speech Recognition (Adam Coates, Baidu) - https://youtu.be/g-sndkf7mCs
10. Torch Tutorial (Alex Wiltschko, Twitter) - https://youtu.be/L1sHcj3qDNc
11. Sequence to Sequence Deep Learning (Quoc Le, Google) - https://youtu.be/G5RY_SUJih4. from: 1point3acres.com/bbs
12. Foundations and Challenges of Deep Learning (Yoshua Bengio) - https://youtu.be/11rsu_WwZTc
.鏈枃鍘熷垱鑷1point3acres璁哄潧
Full Day Live Streams:
Day 1: https://youtu.be/eyovmAtoUx0. From 1point 3acres bbs
Day 2: https://youtu.be/9dXiAecyJrY

Go to http://www.bayareadlschool.org for more information on the event, speaker bios, slides, etc. Huge thanks to the organizers (Shubho Sengupta et al) for making this event happen.
回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-10-19 01:30:35 | 显示全部楼层
2016/10/18
- went through the udacity DL video part. I may not have time to go through all the tensorflow stuff until later.

will dab into traditional classification/regression tasks (for work), and then go through basic NLP/vision/sequence

https://www.coursera.org/learn/neural-networks/home/week/1
.1point3acres缃
to follow by reading this tutorial http://ufldl.stanford.edu/tutorial
回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-10-26 00:06:57 | 显示全部楼层
Udacity has an interesting course but it's too shallow to really "practice" DL. Good as a very first intro course, given you already know ML but knows little about DL.
The exercises were not helpful at all.

The coursera NN for ML course is dry as ****. Maybe good content, but I can't go through with it. Jeff Hinton is the giant, but that course is supremely dry.....maybe will revisit after going through the shallower tutorials.

Stanford 231n is a fantastic course, and like many other stanford courses (in all kinds of other areas), this one is engaging, and establishes the intuition superbly ---- this is a huge advantage that I see in learning. I am so green with envy, re Stanford students!!!!!

Also, instead of Andrew Ng's coursera one, I will review CS229 (the full course) instead.
https://see.stanford.edu/Course/CS229
Apparantly, there is a lot more content....
回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-10-26 01:38:51 | 显示全部楼层
Also the optimization courses here would be a nice refresher
https://see.stanford.edu/Course

回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-10-27 02:26:08 | 显示全部楼层
following the good tradition on this board, here's the learning plan for ML (not necessarily DL)
.1point3acres缃
Stanford 231n
Stanford 229 (Andrew Ng's ML) review: Notes 1-12, Lecture 1-20, I don't plan to do much assignment
==========================
10/26 cs231n:
linear classification; back prop

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-10-28 12:17:40 | 显示全部楼层
10/27 cs231n:
neural net, train nn

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-11-4 08:47:01 | 显示全部楼层
11/2 stanford DL tutorial section 1

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-11-4 10:44:05 | 显示全部楼层
11/3 stanford DL tutorial, multi-layer NN, CNN
unsupervised learning: autoencoders, PCA/whitening
============
TODO: remainder of unsupervised learning and self-taught learning
exercise on CNN
then on with 231n.

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-11-8 07:44:42 | 显示全部楼层
11/7
tSNE
online learning https://www.coursera.org/learn/ml-classification/lecture/vEF6f/timeline-of-scalable-machine-learning-stochastic-gradient
231n - visualize CNN/RNN intro 鏉ユ簮涓浜.涓夊垎鍦拌鍧.
============
TODO:
stanford tutorial: remainder of unsupervised learning and self-taught learning
exercise on CNN

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-11-14 09:05:56 | 显示全部楼层
came across an interesting playlist on youtube:
2 min papers:
https://www.youtube.com/user/keeroyz/playlists
quite fun
kinda like "talking machines" podcast (sadly, it seems to be no longer actively updating)
also found some other podcasts on software engineering or ML, but they weren't as good as talking machines

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

RobinEJ 发表于 2016-11-17 11:18:05 | 显示全部楼层
谢谢K姐分享!
回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-12-3 22:30:21 | 显示全部楼层
finished video part of cs231n. now on to toy examples and finding some real use cases from work, maybe
recently started a refresher using the udacity legacy course intro to AI. since my experience in ML/DL etc is completely self taught, this course filled some gaps in the knowledge landscape.

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

DL 发表于 2016-12-4 13:20:45 | 显示全部楼层
谢谢K姐分享这么多好东西
可惜我基础太差,根本没能力看这些。先存下来当胡萝卜挂着吧。
回复 支持 反对

使用道具 举报

 楼主| 小K 发表于 2016-12-14 14:14:33 | 显示全部楼层
Docker image
最好的宝贝,没有之一 https://github.com/saiprashanths/dl-docker 好用的想哭。不然就会抓狂到想哭。
caffe model zoo:. Waral 鍗氬鏈夋洿澶氭枃绔,
https://github.com/BVLC/caffe/wiki/Model-Zoo
有人说可以用这个https://github.com/ethereon/caffe-tensorflow 把其变成TF zoo Deep Viz toolbox
https://github.com/yosinski/deep-visualization-toolbox

评分

1

查看全部评分

回复 支持 反对

使用道具 举报

time2036 发表于 2017-1-16 13:48:23 | 显示全部楼层
谢谢分享!好多东东!
回复 支持 反对

使用道具 举报

time2036 发表于 2017-1-30 10:40:57 | 显示全部楼层
谢谢分享!收藏学习!
回复 支持 反对

使用道具 举报

本版积分规则

关闭

一亩三分地推荐上一条 /5 下一条

手机版|小黑屋|一亩三分地论坛声明 ( 沪ICP备11015994号 )

custom counter

GMT+8, 2017-5-30 07:25

Powered by Discuz! X3

© 2001-2013 Comsenz Inc. Design By HUXTeam

快速回复 返回顶部 返回列表