dive-into-machine-learning / dive-into-machine-learning Goto Github PK

in this ticket, 'learning' refers to humans learning a topic, in general - it is not machine-learning hehe

I learned Python by hacking first, and getting serious later. [...] If this is your style, join me in getting a bit ahead of yourself

I've recently learned the term/concept of "impasse-driven learning" and it was a 💡 for me. Realized this is my preferred learning style. Some interesting papers exist (though I'm surprised how few): https://scholar.google.com/scholar?q=impasse-driven+learning

this guide is oriented towards that learning style. so I had the thought of adding a small note or link that points that out.

I agree that hacking is the way you get started. But after going through the same process I struggled for some foundation later.

The Data 8 course and the inferential thinking textbook cover the foundations of data science. The course was designed for undergrads because students were struggling with concepts once taking advance topics at the grad level.

Also, the idea behind the course came from Prof. Micheal Jordon who is also linked to AMP Lab (the birth place for Spark).

PS: This is the only textbook with atleast a definition of Data Science compared to the Venn diagram which now has potentially 7-8 versions.

Wanted to confirm if you want to include this information:

• Atleast a link
• Or link with some description highlighting the importance of this course
• Using the definition of data science as mentioned in the text book (separate PR)

I can raise a PR based on your response.

There is already a section about Big Data but it needs some revision.

I've accumulated a couple links I want to throw in, but I may need assistance from an expert to keep this section tiny.

Preview of the revised links (add onto the links to Apache Spark etc. -- those will remain there)

• 10 things statistics taught us about big data analysis (and some more food for thought: "What Statisticians think about Data Scientists")
• Talking Machines #12: Interviews Prof. Andrew Ng (from our main course, which has its own module on big data); the show covers some problems relevant to high-dimensional data
• Talking Machines #15: "Really Really Big Data and Machine Learning in Business":
• This preview edition of this book, Getting Data Right ... - http://jdp491bprdv1ar3uk2puw37i.wpengine.netdna-cdn.com/wp-content/uploads/2015/11/Getting_Data_Right_Preview_Edition_Nov2015.pdf
• the relevant module in Prof. Ng's class should be mentioned.
• anything else?

PULL REQUESTS WELCOME!

We already have this ...

Here's an IPython Notebook book about Probabilistic Programming and Bayesian Methods for Hackers: "An intro to Bayesian methods and probabilistic programming from a computation/understanding-first, mathematics-second point of view."

I've come across markdregan/Bayesian-Modelling-in-Python and it looks great. Should probably add a link to this, right around the quote above. But then it might be good to have a word or two link to some explanation of the exact relationship of Bayesian Modeling to ML ...

the inline link should link to 1+ of these:

• https://www.metacademy.org/roadmaps/rgrosse/bayesian_machine_learning
• http://mlg.eng.cam.ac.uk/zoubin/bayesian.html
• http://www.cs.utah.edu/~piyush/teaching/bayes-review.pdf

See Pull Request #4.

May want to add link to HackerRank: Statistics and Machine Learning, but want to vet it a bit. The site has many categories, so I would like to check if these are (indisputably) worthwhile exercises before adding link.

• https://github.com/rasbt/python-machine-learning-book-2nd-edition
• https://github.com/rasbt/deep-learning-book

https://developers.google.com/machine-learning/crash-course/

• add link in the Deep Learning section.
• add link in the Other Guides section.

• https://github.com/instacart/lore/
• https://tech.instacart.com/how-to-build-a-deep-learning-model-in-15-minutes-a3684c6f71e

http://www.thetalkingmachines.com/blog/2015/4/23/starting-simple-and-machine-learning-in-meds link doesn't work anymore.

This may be the correct one.

http://yerevann.com/a-guide-to-deep-learning/

UC Berkeley now offering their Data 8 course via edX(MOOC) platform. Update the Data 8 section accordingly.

Note: The previous issue #105 with the same name was not resolved, so opening again.

I should get wise to TensorFlow vs. Theano vs. Torch then update this snip of verbiage...

TensorFlow seems like a really big deal. [...]

maybe to something like

TensorFlow seems like a really big deal. It has to have its own bullet point. Now, it's still not magic. And it's not the only Deep Learning framework. But. You can bet people will do exciting things with TensorFlow, Theano, Torch and other machine learning frameworks that make complex algorithms more successful. Just remember: "More data beats a cleverer algorithm" (Domingos).

https://github.com/jakevdp/PythonDataScienceHandbook

https://fivethirtyeight.blogs.nytimes.com/fivethirtyeights-2012-forecast/ is invalid and was captured by CI in #90 pull request

Hello! I have a suggestion for the repo, and thought an issue might be the best place to post it.

I created a video series about scikit-learn, designed specifically for Python users with no background in machine learning or scikit-learn. I thought it might be useful for your readers. Here are the relevant links:

GitHub repo containing the notebooks - https://github.com/justmarkham/scikit-learn-videos
each notebook is related to a video - https://www.youtube.com/playlist?list=PL5-da3qGB5ICeMbQuqbbCOQWcS6OYBr5A
each notebook is also related to a blog post - http://blog.kaggle.com/author/kevin-markham/

If you'd like a quick overview of the series, it can be found here:
http://blog.kaggle.com/2015/04/08/new-video-series-introduction-to-machine-learning-with-scikit-learn/

As a side note, the first video in the series isn't really about scikit-learn, it's just an introduction to how machine learning "works". Thus, it could be worth a separate link in your repo:
https://www.youtube.com/watch?v=elojMnjn4kk&list=PL5-da3qGB5ICeMbQuqbbCOQWcS6OYBr5A&index=1

Anyway, I'm happy to answer any questions and appreciate your consideration!

Submitted by @vishwajeetv

This guide sets pragmatic philosophy towards becoming a machine learning developer.

But I need to add some framing or caveats. From #38 ... I said:

"Dive into Machine Learning" is very carefully curated so far ... While the Brownlee article has a similar hack-first focus, I have issues with Brownlee's tone. I'm trying to put myself in the shoes of someone who has more experience with ML, and did get a PhD in it.

Maybe this is my issue: this article provides advice that may help some developers get into machine learning ... yet it has a lot of attitude, and that attitude is discouraging towards inquiry and curiosity. If everyone had the attitude of this article (at least, at some of its moments) ... there would be no machine learning field for developers to get into; nobody would get into math or research let alone something as "un-pragmatic" as machine learning research!

I will mull this over a bit. Sorry to delay :) Appreciate the PR, for sure

Python Machine Learning

(and a blog post reflecting on the writing of this book)

The fsecurify link seems to be broken in section "A note about Machine Learning and Security (InfoSec, AppSec)"

Following on #36 , fix redirect links discovered by awesome_bot. https://travis-ci.org/hangtwenty/dive-into-machine-learning

Fix broken links as part of Travis CI build:

Issues :-(

Links

1. [L026] 522 https://www.continuum.io/downloads
2. [L121] http://nb.bianp.net/sort/views/ Connection timed out - connect(2) for "nb.bianp.net" port 80

Specifically in this section:
https://github.com/hangtwenty/dive-into-machine-learning#a-few-useful-things-to-know-about-machine-learning
The image at the end of this section (is from)/(hosted on) Photo Bucket and is not showing up. If possible we could use an alternative host for the image.

It's in a similar spirit, just want to add at the right spot.

• https://github.com/humphd/have-fun-with-machine-learning

Need to try out TPOT and probably add it to the guide.

👍

https://www.coursera.org/learn/learning-how-to-learn/

(braindump)

I've largely avoided having much deep learning links or neural nets links ... because it seems like it can be dangerous for beginners to jump ahead to those when they are not ready. I don't know enough about deep learning to really situate this beside just a couple links in a list ... let alone situate it in a sober, smart way :)

Recently, this MIT book about Deep Learning has been published, and judging by reactions it seems like a worthwhile resource to link to. I still don't want "everything and the kitchen sink" but this would be a good "essential resource" on the subject. Or maybe an expert knows a better choice.

So. Might want a small subsection to link to these two things, with the appropriate one or two sentence caveat that these are advanced topics, refer back to the pyramid showing that algorithm is least important (for many problems), etc etc. Maybe link to a Talking Machines episode too to give the appropriate context.

should go in with the other Deep Learning and Tensor Flow links

https://github.com/deepmind/sonnet

Machine Learning isn't just Kaggle competitions

UC Berkeley now offering their Data 8 course via edX(MOOC) platform. Update the Data 8 section accordingly.

Prodded by this comment on DataTau.

I replied ...

When I first started putting this together, it was still just -- ipython notebook, and other notebooks. The unifying effort of Jupyter hadn't been released yet. I need to catch up here... After I know the full picture about how Jupyter is different from ipython notebook, any compatibility issues, etc., I will make a full pass through the guide to update anything needing updates.

Please don't sell yourself as a Machine Learning expert while you're still in the Danger Zone. Don't build bad products or publish junk science. This guide can't tell you how you'll know you've "made it" into Machine Learning competence ... let alone expertise. It's hard to evaluate proficiency without schools or other institutions. This is a common problem for self-taught people. Your best bet may be: expert peers.

If you know a good way to evaluate Machine Learning proficiency, please submit a Pull Request to share it with us.

Need to tell people how they know they're out of the Danger Zone or how they know they are hire-able.

https://github.com/hangtwenty/dive-into-machine-learning#deep-learning

should have link to this in its list:

https://d2l.ai/

• Very start of the section should link to a page mentioning, "You Are Not the User" (a generally useful principal for developers to remember).
• And link/reference to "Rule #23: You are not a typical end user" from Martin Z's Rules of ML Engineering doc http://martin.zinkevich.org/rules_of_ml/rules_of_ml.pdf

The numpy array takes 13 milliseconds for its execution in an expression while a normal array takes about
several seconds for its operation

See Pull Request #4 .

https://distill.pub/about/

Include this great blog post from Berkley ML dept maybe.

I think it's better for README to have a table of contents, with clickables to jump directly to each of the high-level section headline.
I feel it's a good first task for me as a rookie open-source contributor. Kindly remind me of do's and dont's regarding making changes and pull requests.

Right at the beginning I like to Dive into Python ... and while I would still recommend that as a #1 choice, I should link to my document with more options.

• in notebooks section: http://clicksecurity.github.io/data_hacking/

• by security section: https://github.com/jivoi/awesome-ml-for-cybersecurity/blob/master/README.md

I want the same caution as before, that people should proceed carefully here. But still want some good resources :-)

or perhaps just a link to someone else's. or highlighting that one handy OSS timeseries-db-comparison, plus personal favorites like timescaledb

this is a link that would be added to that section

https://github.com/blue-yonder/tsfresh

This is a machine learning module I found here:

http://www.dcs.gla.ac.uk/~girolami/Machine_Learning_Module_2006/week_2/Lectures/wk_2_lect_2.pdf

None of this material is mine, it has all been created by Professor. M. A .Girolami.
This is hands down the best machine-learning tutorials I have found on the web, and I was afraid
the university link would be taken down, so now its on github.

i hope you enjoy this as much as i did.

https://github.com/josephmisiti/machine-learning-module

Thoughtful Machine Learning with Python: A Test-Driven Approach 1st Edition
by Matthew Kirk. Looks very interesting. I want to look into the book and/or early reviews of it, and most likely find a place to link it in the guide.

Hello, I wrote a tool that can validate README links (valid URLs, not duplicate). It can be run when someone submits a pull request.

It is currently being used by

• https://github.com/vsouza/awesome-ios
• https://github.com/matteocrippa/awesome-swift
• https://github.com/dkhamsing/open-source-ios-apps

Examples

• https://travis-ci.org/matteocrippa/awesome-swift/builds/96526196 ok ✅
• https://travis-ci.org/matteocrippa/awesome-swift/builds/96722421 link redirected / rename 🔴
• https://travis-ci.org/dkhamsing/open-source-ios-apps/builds/96763135 bad link / project deleted 🔴
• https://travis-ci.org/dkhamsing/open-source-ios-apps/builds/95754715 dupe 🔴

If you are interested, connect this repo to https://travis-ci.org/ and add a .travis.yml file to the project.

See https://github.com/dkhamsing/awesome_bot for options, more information
Feel free to leave a comment 😄

Elements of Statistical Learning and Understanding Machine Learning are both free books, and frequently recommended as reference textbooks. As you continue towards expertise you can also go deep into these books.

Issue: need to add link to UML in guide. Also need to link to context/comparison if possible.

I suggest to add Orange to the tools references.
Orange is an open source machine learning and data visualization for novice and expert.
Provides an interactive data analysis workflows with a large toolbox.