Nowadays, in the hot field of artificial intelligence, the most embarrassing problem is the contradiction between the increasingly large scale of the industry and the country’s annual demand for relevant talents of about 5 million. More and more young people choose to enter this industry because of its broad development prospect, huge talent gap and attractive salary. However, at present, there are not many colleges and universities offering artificial intelligence major in China, and the discipline construction is not perfect. Quite a number of developers are cross-border beginners, who need to learn a lot of knowledge by themselves and explore in practice. In the face of the fragmentation of network information and the situation of various and expensive training courses, how to efficiently learn has become the primary problem for the introduction of artificial intelligence.
Here are some of the most popular ai courses offered by Ng’s company, Coursera, Stanford and Berkeley, to help you learn.
Machine learning
Machine Learning
Speaker: Andrew Ng
Platform: Coursera
Course links:
https://www.coursera.org/learn/machine-learning#faqs
Course rating: On Coursera, nearly 50,000 people gave Ng an average of 4.9 out of 5 for the course, making it the most highly rated of any online Machine Learning course, according to Freecodecamp. This is an introductory theoretical course that lays the foundation for machine learning. It not only explains the basic concepts, but also attaches great importance to the connection between practice and experience summary: 1. In the course, Mr. Ng enumerated many practical examples of algorithm application. 2. He mentioned many of the problems they faced when they first started AI, and the experience of dealing with them. Considering its wide audience, this course does not involve too much math and is very friendly to students with weak basic knowledge of statistics and IT.
Learning from Data
Speaker: Yaser Abu-Mostafa
Release platform: edX, NetEase Open Class
Course links:
https://www.edx.org/course/caltechx/caltechx-cs1156x-learning-data-2516;
http://open.163.com/special/opencourse/learningfromdata.html
This is an introduction to machine learning by Professor Yaser Abu-Mostafa of Caltech, but the content is not easy. The course emphasizes data because machine learning is closely related to big data processing applications in various fields, such as finance and healthcare. This course covers fundamental theories, algorithms, and applications, balancing theory with practice, covering both mathematical statistics and heuristic conceptual understanding. Many have commented that the course is structured like storytelling and helps learners develop a deep, intuitive understanding of machine learning concepts and models. Learners agree that it is very informative, but the assignment module is highly controversial: some say it is difficult and lacks feedback, others say it is the best machine learning exercise available online.
Neural Networks for Machine Learning
Speaker: Geoffrey Hinton
Platform: Coursera
Course links:
https://www.coursera.org/learn/neural-networks#ratings
Course evaluation: This course by Geoffrey Hinton can be used as an advanced course of Ng’s machine learning, with relatively higher difficulty. It requires students to have basic knowledge of calculus and Python, and involves many proper nouns. It is difficult for beginners, so they need to find relevant materials by themselves. “Taking this course was a real eye-opener for me, and as far as I can tell, it’s pretty close to the cutting edge of deep learning. The problem sets were more detailed and challenging than Ng’s class, so I ended up learning more.”
Machine Learning
Speaker: Tom Mitchell
Release platform: CMU official website
Course links:
http://www.cs.cmu.edu/~tom/10701_sp11/
This course covers a wide range of topics, including algebra and probability theory, basic tools of machine learning, probabilistic graph models, AI, neural networks, active learning, and reinforcement learning, in order of importance. This course will help learners understand the evolution of machine learning. It is suitable for people planning systematic learning and investing a lot of time. For starters, it is advisable to take this course at least after attending Ng’s machine learning course.
Fundamentals of machine learning
Speaker: Lin Xuantian
Platform: Coursera
Course links:
On: https://www.coursera.org/learn/ntumlone-mathematicalfoundations
Under: https://www.coursera.org/learn/ntumlone-algorithmicfoundations
Course evaluation: This is an introductory course tailor-made for Chinese students, equivalent to the first half semester of machine learning course in Taiwan University, which teaches the most core knowledge of machine learning. Mr. Lin is one of the authors of the textbook “Learning From Data” and a promising young scholar in the field of machine Learning in China. The course was very thoughtful and slightly more substantial than Mr Ng’s introductory course. According to Lin, many students complain that English teaching is not easy to absorb as the top open machine learning courses are all taught in English. Therefore, by launching this course, we hope to help Chinese native language students to reduce the difficulty of entry.
Machine Learning for Undergraduates
Speaker: Nando de Freitas
Platform: Youtube
Course links:
Nando de Freitas is a distinguished scholar in the field of machine learning. His course is also suitable for ng’s advanced machine learning course, supplementing some “machine learning” concepts with an emphasis on mathematics. Nando de Freitas gives a good explanation of basic mathematical principles such as probability theory and log Likelihood, and introduces more advanced mathematical and statistical concepts based on this.
Deep learning
Pratical Deep Learning for Coders, Part 1
Speaker: Jeremy Howard
Release platform: fast.ai
Course links:
http://course.fast.ai/
Course evaluation: it is an extremely practical course. Jeremy Howard, a Kaggle contest winner, teaches himself how to build the best deep neural networks in the industry. Jeremy Howard shares the methods that have actually been used and proven to work in engineering practice, not just theoretical definitions and formulas.
Deepinglearning.ai Specialization
Speaker: Andrew Ng
Platform: deeplearning.ai
Course links:
This is the first Deep Learning project (DEEplearning. ai) launched by Ng after he left Baidu. The course slogan is: Master Deep Learning,and Break into AI. This course is a bottom-up approach to the principles of neural networks. This course will help to strengthen the understanding of deep learning and, for those who already have some grounding in neural networks, the parameter search technique. It is considered by many to be the best introductory deep learning series available on the Internet to help scholars build a basic understanding of the field.
Deep learning at Oxford 2015
Speaker: Nando de Freitas
Release platform: Oxford official website
Course Home page:
http://www.cs.ox.ac.uk/people/nando.defreitas/
Assessment: Nando de Freitas moved to Oxford in 2013, and this is the full range of deep learning courses he took at Oxford for the 2014-2015 academic year. It introduces the basic background of neural networks, back propagation, Boltzmann machines, autoencoders, convolutional neural networks and recursive neural networks, illustrates how deep learning affects our understanding of intelligence and contributes to the practical design of intelligent machines.
CS231n: Convolutional Neural Networks for Visual Recognition
Speaker: Fei-fei Li
Platform: GitHub
Course Address:
http://cs231n.github.io/
Evaluation: This deep learning course for computer vision, directed by Professor Fei-Fei Li, is very authentic and highly rated for Stanford students. Although the name of the course is convolutional Neural network and Image recognition, a lot of basic knowledge about the building of Python development environment and the principle of neural network were introduced in the early stage, which is suitable for beginners to study seriously.
Deep Learning
Speaker: Yann Lecun
Release platform: Official website of College Francaise
Course Address:
https://www.college-de-france.fr/site/en-yann-lecun/course-2016-04-15-11h00.htm
Course Evaluation: Yann Lecun teaches eight courses on deep learning at the College de France in early 2016. Lessons were taught in French, with English subtitles added later. As a leader in artificial intelligence and head of Facebook’s AI Lab (FAIR), Yann Lecun is at the forefront of machine learning research in the industry. He has said publicly that some of the existing open machine learning courses are outdated. Through Yann Lecun’s course, we can learn the latest progress of deep learning research in recent years. This series serves as an advanced course in exploring deep learning.
Deep Learning for Natural Language Processing: 2016-2017
Speaker: Phil Blunsom
Platform: GitHub
Course links:
https://www.bilibili.com/video/av9817911/
This is Oxford University and DeepMind’s deep Learning application course for NLP. The course introduces mathematical definitions of relevant machine learning models and deduces relevant optimization algorithms. The course covers a range of applications of neural networks in NLP, including analyzing potential dimensions in text, transcribed speech to text, converting between languages, and answering questions. These topics are organized into three high-level topics, ranging from understanding sequential language modeling using neural networks, to understanding their use as conditional language models for transduction tasks, and ultimately ways to combine these techniques with other high-level applications. Practical implementations of these models on CPU and GPU hardware will also be discussed throughout the process.
2016 Stanford Bay Area Deep Learning School Day 1
Course links:
https://www.youtube.com/watch?v=eyovmAtoUx0
Course evaluation: This video is a presentation of the first day of the Bay Area Deep Learning School in 2016. 1) Introduction on Feedforward Neural Network by Hugo Larochelle Andrej Karpathy: Deep Learning for Computer Vision; 3) Richard Socher teaches Deep Learning for NLP; Sherry Moore teaches the TensorFlow Tutorial. 5) Ruslan Salakhutdinov teaches Foundations of Deep Unsupervised Learning; 6) The Nuts and Bolts of Applying Deep Learning. These deep learning experts will explain the underlying concepts and principles of deep learning in an easy-to-understand way to give you a basic understanding of deep learning. They will also share examples of applications related to their respective topics.
2016 Bay Area Deep Learning School Day 2 at CEMEX Auditorium, Stanford
Course links:
https://www.youtube.com/watch?v=9dXiAecyJrY
Course evaluation: This is the second day of the Deep Learning school in the Bay Area. The video covers the following contents: 1) John Schulman teaches Foundation of Deep Reinforcement Learning; 2) Pascal Lamblin teaches Theano: An extremely fast Python library for model building and training. A Fast Python Library for Modelling & Training); 3) Speech Recognition and Deep Learning by Adam Coates and Vinay Rao Machine Learning with Torch & Autograd by Alex Wiltschko 5) Quoc Le teaches Seq2Seq (Sequence to Sequence by Deep Learning); Yoshua Bengio teaches the Foundation and Challenges of Deep Learning. These deep learning users are experts in frequently searched deep learning applications, and they also work for large companies such as Google Brain and Twitter.
Natural Language Processing with Deep Learning
Course links:
https://www.bilibili.com/video/av9285496/
This course is led by NLP leaders Chris Manning and Richard Socher. It is a classic course in deep learning natural language processing. This course provides a comprehensive introduction to cutting-edge research in deep learning as it applies to NLP. In terms of models, we will discuss word vector representation, window-based neural networks, recursive neural networks, long term short-term memory models, recursive neural networks, convolutional neural networks, and some recent models involving memory components.
Deep Learning (Chinese/English) by Google
Speaker: Vincent Vanhoucke, Arpan Chakraborty
Release platform: Youda School City
Course links:
https://cn.udacity.com/course/deep-learning–ud730
Course evaluation: Developed in collaboration with Vincent Vanhoucke, Google’s chief scientist and technology manager for the Google Think Tank, this course offers free lessons on training and optimizing basic neural networks, convolutional neural networks, and short and long term memory networks. Learners are exposed to TensorFlow, a complete machine learning system, through projects and tasks.
2016 Monterier Deep Learning Summer School
Course links:
Evaluation: The Montreal Deep Learning Summer Program has many experts and practitioners from different ages. The purpose of this tutorial is to give people a basic understanding of deep learning and neural networks. Yoshua Bengio: Recurrent Neural networks, Surya Ganguli: Theoretical Neuroscience and Deep Learning, Sumit Chopra: Reasoning Summit, Attention, Jeff Dean on TensorFlow large-scale machine learning, Ruslan Salakhutdinov on deep generative models of learning, Ryan Olson on GPU programming for deep learning, and many more.
Machine Learning and having it Deep and Structured
Speaker: Li Hongyi
Course links:
https://www.bilibili.com/video/av9770302/
Course evaluation: This is a rare free Chinese course, some netizens comment that this course speaks GAN too well, and others think that Li Hongyi is good at giving students an intuitive image and understanding of the algorithm. This course is not very beginner friendly and is suitable for learning after his Machine Learning course.
Artificial intelligence
Intro to Artificial Intelligence
Speakers: Peter Norvig, Sebastian Thrun
Platform: Udacity
Course Address:
https://cn.udacity.com/course/intro-to-artificial-intelligence–cs271
Course evaluation: This course has a long reputation and is recognized as one of the best open courses for AI introduction. It introduces several major areas of AI: probabilistic reasoning, information retrieval, robotics, natural language processing, etc. It tends to introduce practical applications of AI, and the course exercises are well received. The course’s two speakers, Peter Norvig and Sebastian Thrun, a research director at Google and a renowned machine learning professor at Stanford, are both top AI experts in the same class as Ng and Yann Lecun.
Knowledge-based artificial Intelligence: Cognitive Systems
Speaker: Ashok Goel David Joyner
Release platform: Youda School City
Course links:
https://cn.udacity.com/course/knowledge-based-ai-cognitive-systems–ud409
Course Evaluation: This course is the core of artificial intelligence and is highly challenging, involving significant independent work, reading, tasks and projects. It covers structured knowledge presentation and knowledge-based problem solving, planning, decision making and learning methods. “The most fascinating part is the course project: building an ARTIFICIAL intelligence agent that solves Raven’s evolution matrix, which is basically a visual IQ test that is very interesting and challenging. The ‘easy’ problems are easy to solve, but the hard ones are unbelievably hard to solve.”
Computer science
Principles of Scala functional programming
Speaker: Martin Odersky
Platform: Coursera
Course links:
https://www.coursera.org/learn/progfun1
Evaluation: This course is very hands-on. Most of the units in the course use short programs to illustrate basic principles and concepts. Listeners can try to run these programs and try to rewrite them. Netizens generally agree that the course is a little difficult, while others feel that the course structure is not reasonable and requires students to consult a lot of other materials. Proponents say, “Having the creator of a language teach the language himself gave me insights I wouldn’t have gotten otherwise.” “This course is amazing and highly recommended, and it shows how much effort and skill went into Scala’s design.
The database
Speaker: Jennifer Widom
Website: Stanford University
Course links:
https://lagunita.stanford.edu/courses/DB/2014/SelfPaced/about
Evaluation: This course is one of three large open online courses offered by Stanford in the fall of 2011; It was offered again in moOCs in 2013 and 2014. Netizens to the course evaluation of the lesson is “organization” the best online courses, “all of the content in the lectures are related, all content will be applied in the problem sets and test, exercise a lot, the weekly homework assignments from simple unfolding to moderate difficulty, well-designed Web environment, provides an excellent feedback and can guide you to answer a question correctly.”
Probabilistic Graphical Models Course
Speaker: Daphne Koller
Release platform: Couresa
Course links:
https://www.coursera.org/specializations/probabilistic-graphical-models
Course evaluation: One user said this course was the most interesting online course he had ever taken, but it required a lot of effort to apply the lecture content to the homework. Another commented, “The first course of this particular course had a very good and very engaging start, but after that the gap between lectures and problem sets quickly opened up (perhaps this is why the author brags that this is a challenging course and not for everyone). PGM is a powerful tool for solving many machine learning problems, but it can be difficult. According to a netizen, “At Stanford, students are thrilled to pass the PGM exam.”
5. Platform recommendation
Abroad: Coursera, edX, Udacity, Udemy, etc
Domestic: NetEase open class, July online, bilibili, etc
If your English is good, an overseas Q&A forum like Hacker News will be helpful. If English isn’t enough, most of Coursera’s and Yoda’s machine learning resources add Chinese subtitles; There are also many subtitle translations for NetEase open courses; If you need to supplement the basic knowledge of mathematics and statistics, Khan Academy is highly recommended.
The original article was published on April 16, 2018
Author: Data
This article is from “Data THU”, a partner of the cloud community. For more information, follow “Data THU”.