Name: lectures
Owner: Biomedical & Health Informatics
Description: null
Forked from: MachinesWhoLearn/workshops
Created: 2016-09-26 00:27:30.0
Updated: 2016-09-26 00:27:32.0
Pushed: 2016-05-05 00:50:00.0
Homepage: null
Size: 562
Language: Jupyter Notebook
GitHub Committers
| User | Most Recent Commit | # Commits |
|---|
Other Committers
| User | Most Recent Commit | # Commits |
|---|
The ISL and ESL textbooks are both FREE to download (legally!). See links below.
An Introduction to Statistical Learning with Applications in R An excellent textbook intended for students in the non-mathematical sciences. Also useful for the R examples. Coauthored by UW's very own Daniela Witten.
Elements of Statistical Learning My personal favorite reference. Accessible, yet advanced enough to give you the computational and mathematical reasoning behind many models.
Machine Learning (A Probabilistic Perspective) Probably the most well known textbook on machine learning. Recommended for mathematically mature readers.
A list of books from related fields that Marina Meila (teaches STAT 391, STAT 535, and others) recommends for those who are “serious about Statistical Learning”.
Marina also has a list of books on probability, for STAT 391 students.
UW faculty member Jake VanderPlas is working on an O'Reilly book “Python Data Science Handbook”. Some previews are posted.
Either Python or R are both excellent languages to know for those getting started in machine learning. Lectures will initially be taught in R.
We recommend you install the free Anaconda Python distribution, which will automatically install Python along with all of the machine learning and scientific computing libraries you need.
The following are indispensable tools for any type of data analysis in the Python environment:
scikit-learn ? Python's primary machine learning library
NumPy ? an optimized Python numerical library
matplotlib ? graphing and plots
SciPy ? domain specific scientific functions
Pandas ? data analytics
Git ? version control for software
Jupyter Notebook ? interactive computing
This collection of online notebooks created by UW faculty member Jake VanderPlas is an excellent resource for becoming familiar with the above development stack. 00_intro.ipynb and 01_basic_training.ipynb introduce the Python programming language. Subsequent notebooks cover the aforementioned list of tools.
Reinforcement learning is a fascinating subfield of machine learning with connections to artificial intelligence, control theory and behavioral psychology.
For an introduction to the field, read the first chapter of Sutton and Barto's classic book, “Reinforcement Learning: An Introduction”. It's really an excellent book, fun to read, and available online in PDF or HTML.
Deep Reinforcement Learning online course notes from Berkeley
Reinforcement Learning course videos and slides from David Silver of Google DeepMind/University College London
The Berkeley AI course offers great introductory online lectures on reinforcement learning and a fun PacMan reinforcement learning project, which will be familiar to those who took CSE 473 (Intro to Artificial Intelligence) at UW.
The term “deep learning” most commonly refers to “neural nets with multiple hidden layers”, and is currently an extremely active area of research.
Deep Learning: an excellent overview article published in Nature in 2015, written by three leaders in the field.
Neural Networks and Deep Learning, free e-book.
Deep Learning, a comprehensive, in-progress textbook freely available online.
Convolutional Neural Networks for Visual Recognition, Berkeley online course notes.
Juergen Schmidhuber's website has many interesting resources.
Deep Learning course videos and slides by Nando de Freitas of Google DeepMind/Oxford.
Keras is a powerful and easy to use deep learning library for Python that supports the popular TensorFlow and Theano backends.
Torch is another commonly used deep learning framework.
A Few Useful Things to Know about Machine Learning is a short essay describing important things to know, from UW CSE professor Pedro Domingos.
Lecture notes for STAT 535 “Foundations of Machine Learning”.
Andrew Ng's (Stanford) machine learning course on Coursera.
UW professor Pedro Domingos online Machine Learning course.
The FAQ in r/MachineLearning.
STAT 391 (Spring, non-majors okay) - Probability and Statistics for Computer Scientists. MATH/STAT 394 or 390 comes in incredibly handy for this class. MATH/STAT 395 doesn't hurt either. Prerequisite: minimum grade of 2.5 in MATH 126; 2.5 in MATH 308; either CSE 326, CSE 332, CSE 373, CSE 417, or CSE 421.
CSE 446 (Winter/Spring, majors only) - Machine Learning. Prerequisite: CSE 332; either STAT 390, STAT 391, or CSE 312.
STAT 535 (Autumn, non-majors okay) - Foundations of Machine Learning. Prerequisite: “experience with programming in a high level language”, but in reality you need to have taken at least STAT 391 or CSE 446. Intended for grads and newly anointed Phd candidates, but some undergrads take it as well.
Related courses that cover stuff you will need to know eventually anyways:
Helpful but not necessary:
Google DeepMind's Deep Q-learning playing Atari Breakout [1:42] You can read more about it in Wired, read the excellent academic paper and even download an open source implementation.
MarI/O - Machine Learning for Video Games [5:58] Neural network learns how to play Super Mario Bros. from simplified screen input.
Computer program that learns to play classic NES games [16:17] Explanation ends and gameplay begins at [6:13]. A more general program, which reads in previously recorded bytes of memory from a human “teacher” playing a video game, then attempts to find a sequence of those bytes which make its level in the game go up. (So all it is trying to do is make (specific) bytes in memory increase, it knows nothing about the game). Notice how often the player gets stuck in local optima. This is a problem in optimization problems in general, but particularly in neural networks.
This material would more accurately fall under artificial intelligence, a superfield of machine learning, but it's beautiful nonetheless.
Emily Howell, a computer program that composes music by listening to recordings of performances by humans.
EMI “Experiments in Musical Intelligence” is another program by David Cope. Here is an eerily well composed symphony in the style of Mozart.
Artificial General Intelligence, which is defined as “the capacity of an engineered system to display the same rough sort of general intelligence as humans; or, display intelligence that is not tied to a highly specific set of tasks”, depends heavily on the field on machine learning.
Scholarpedia overview of the field of Artificial General Intelligence
The Hidden Pattern by Dr. Ben Goertzel, Chairman of the Artificial General Intelligence Society and OpenCog
A Formal Theory of Universal Intelligence, an introductory video by Dr. Marcus Hutter
Machine Superintelligence, the PhD thesis of Shane Legg, cofounder of Google DeepMind
How Khan Academy is using Machine Learning to Assess Student Mastery
How Companies Know Your Secrets The classic article on how Target found out a girl was pregnant before her father did.
Other links will be posted as they pertain to lectures!