You know what I was hoping to have when I started learning Machine Learning. An all in one Machine Learning course. At the time, it was really tricky to find a good course with all the necessary concepts and algorithms. So we were forced to search all over the web, read research papers, and buy books.
Luckily that’s not the case any more. Now we are in the exact opposite situation. There are so many courses out there. How I am supposed to know which one is good, which includes all the things I need to learn. So here I compiled a list of the most popular and well- taught courses.
I have personal experience with most of them and I highly recommend all of them. Every Machine Learning Engineer or Data Scientist I know suggests one or many of them. So don’t look any further.
Ok, let’s get started.
This course by Stanford is considered by many the best Machine Learning course
around. It is taught by Andrew Ng himself ( for those of you who don’t know him,
he is a Stanford Professor, co-founder of Coursera, co-founder of Google Brain
and VP of Baidu) and it covers all the basics you need to know. Plus, it has a
rating of a whopping 4.9 out of 5.
The material is completely self-contained and is suitable for beginners as it
teaches you basic principles of linear algebra and calculus alongside with
supervised learning. The one drawback I can think of, is that it uses Octave (
an open-source version of Matlab) instead of Python and R because it really
wants you to focus on the algorithms and not on programming.
Cost: Free to audit, $79 if you want a Certificate
Time to complete: 76 hours
Syllabus: Linear Regression with One Variable
Linear Algebra Review
Linear Regression with Multiple Variables
Neural Networks: Representation
Neural Networks: Learning
Advice for Applying Machine Learning
Machine Learning System Design
Support Vector Machines
Large Scale Machine Learning
Application Example: Photo OCR
Again, a course taught by Andrew Ng and again it is considered on the best in
the field of Deep Learning. You see a pattern here? It actually consists of
5 different courses and it will give you a clear understanding of the most
important Neural Network Architectures. Seriously if you are interested in DL,
look no more.
It utilizes Python and the TensorFlow library ( some background is probably
necessary to follow along) and it gives you the opportunity to work in real-life
problems around natural language processing, computer vision, healthcare.
Cost: Free to audit, $49/month for a Certificate
Time to complete: 3 months (11 hours/week)
Neural Networks and Deep Learning
Improving Neural Networks: Hyperparameter Tuning, Regularization, and
Structuring Machine Learning Projects
Convolutional Neural Networks
The advanced Machine Learning specialization is offered by National Research
University Higher School of Economics and is structured and taught by Top Kaggle
machine learning practitioners and CERN scientists It includes 7 different
courses and covers more advanced topics such as Reinforcement Learning and
Natural Language Processing. You will probably need more math and a good
understanding of basic ML ideas, but the excellent instruction and the fun
environment will make up to you. It surely comes with my highest recommendation.
Cost: Free to audit, $49/month for a Certificate
Time to complete: 8-10 months (6-10 hours/week)
Introduction to Deep Learning
How to Win Data Science Competitions: Learn from Top Kagglers
Bayesian Methods for Machine Learning
Practical Reinforcement Learning
Deep Learning in Computer Vision
Natural Language Processing
Addressing the Large Hadron Collider Challenges by Machine Learning
If you need a holistic approach on the field and an interactive environment,
this is your course. I have to admit that I haven’t seen a more complete
curriculum than this. From supervised learning to unsupervised and
reinforcement, it has everything you can think of.
It won’t teach you Deep neural networks, but it will give you a clear
understanding of all the different ML algorithms, their strengths, their
weaknesses and how they can be used in real-world applications. Also, if you are
a fan of very short videos and interactive quizzes throughout the course, it’s a
perfect match for you.
Time to complete: 4 months
allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture"
This introductory class is designed and taught the co-founder of Udacity
Sebastian Thrun and the Director of Data Science Research and Development Katie
Malone. Its primary audience is beginners who are looking for a course to get
started with ML. Again if you like Udacity’s environment (which I personally do),
it is an amazing alternative to get your foot in the door.
Time to complete: 10 weeks
Welcome to Machine Learning
Support Vector Machines
Choose your own Algorithm
Datasets and Questions
The Deep Learning Nanodegree by Udacity will teach you all the cutting-edge DL
algorithms from convolutional networks to generative adversarial networks. It is
quite expensive but is the only course with 5 different hands-on projects. You
will build a dog breed classifier, a face generation system a sentiment analysis
model and you’ll also learn how to deploy them in production. And the best part
is that it is taught by real authorities such as Ian Goodfellow, Jun-Yan Zhuand,
Sebastian Thrun and Andrew Trask.
Cost: 1316 €
Time to complete: 4 months
Project 1: Predicting Bike-Sharing Patterns (Gradient Descent and Neural
Project 2: Dog Breed Classifier( CNN, AutoEncoders and PyTorch)
Project 3: Generate TV Scripts (RNN, LSTM and Embeddings)
Project 4: Generate Faces (GAN)
Project 5: Deploy a Sentiment Analysis Model
The next in our list is hosted in edX and is offered by the Columbia University.
It requires substantial knowledge in mathematics (linear algebra and calculus)
and Programming( Python or Octave) so if I were a beginner I wouldn’t start
here. Nevertheless, it can be ideal for more advanced students if they want to
develop a mathematical understanding of the algorithms.
One thing that makes this course unique is the fact that it focuses on the
probabilistic area of Machine Learning covering topics such as Bayesian linear
regression and Hidden Markov Models.
Cost: Free to audit, $227 for Certificate
Time to complete: 12 weeks
Week 1: maximum likelihood estimation, linear regression, least squares
Week 2: ridge regression, bias-variance, Bayes rule, maximum a posteriori
Week 3: Bayesian linear regression, sparsity, subset selection for linear
Week 4: nearest neighbor classification, Bayes classifiers, linear
Week 5: logistic regression, Laplace approximation, kernel methods, Gaussian
Week 6: maximum margin, support vector machines, trees, random forests,
Week 7: clustering, k-means, EM algorithm, missing data
Week 8: mixtures of Gaussians, matrix factorization
Week 9: non-negative matrix factorization, latent factor models, PCA and
Week 10: Markov models, hidden Markov models
Week 11: continuous state-space models, association analysis
Week 12: model selection, next steps
Practical Deep Learning for Coders is an amazing free resource for people with
some coding background (but not too much) and includes a variety of notes,
assignments and videos. It is built around the idea to give students practical
experience in the field so expect to code your way through. You can even learn
how to use a GPU server on the cloud to train your models. Pretty cool.
Time to complete: 12 weeks (8 hours/week)
Introduction to Random Forests
Random Forest Deep Dive
Performance, Validation, and Model Interpretation
Feature Importance. Tree Interpreter
Extrapolation and RF from Scratch
Data Products and Live Coding
RF From Scratch and Gradient Descent
Gradient Descent and Logistic Regression
Regularization, Learning Rates, and NLP
More NLP and Columnar Data
Complete Rossmann. Ethical Issues
Definitely, the most popular AI course on Udemy with half a million students
enrolled. It is created by Kirill Eremenko, Data Scientist & Forex Systems
Expert and Hadelin de Ponteves, Data Scientist. Here you can expect an analysis
of the most important ML algorithms with code templates in Python and R. With 41
hours of learning + 31 articles, it is certainly worth a second look.
Cost: 199 € (but with discounts. At the time of writing the cost was 13.99€)
Time to complete: 41 hours
Part 1 - Data Preprocessing
Part 2 - Regression: Simple Linear Regression, Multiple Linear
Regression, Polynomial Regression, SVR, Decision Tree Regression, Random
Part 3 - Classification: Logistic Regression, K-NN, SVM, Kernel SVM, Naive
Bayes, Decision Tree Classification, Random Forest Classification
Part 4 - Clustering: K-Means, Hierarchical Clustering
Part 5 - Association Rule Learning: Apriori, Eclat
Part 6 - Reinforcement Learning: Upper Confidence Bound, Thompson Sampling
Part 7 - Natural Language Processing: Bag-of-words model and algorithms for
Part 8 - Deep Learning: Artificial Neural Networks, Convolutional Neural
Part 9 - Dimensionality Reduction: PCA, LDA, Kernel PCA
Part 10 - Model Selection & Boosting: k-fold Cross Validation, Parameter
Tuning, Grid Search, XGBoost
The most difficult course on the list for sure because arguably Reinforcement
Learning is much more difficult. But if you want to dive into it, there is no
better way to do it. It is in fact actual recorded lectures from Stanford
University. So be prepared to become a Stanford student yourself. The professor
Emma Brunskill makes it very easy to understand all these complex topics and
gives you amazing introduction to the RL systems and algorithms. Of course, you
will find many mathematical equations and proofs, but there is no way around it
when it comes to Reinforcement Learning.
Time to complete: 19 hours
Given a model of the world
Model-Free Policy Evaluation
Value Function Approximation
CNNs and Deep Q Learning
Policy Gradient I
Policy Gradient II
Policy Gradient III and Review
Fast Reinforcement Learning
Fast Reinforcement Learning II
Fast Reinforcement Learning III
Batch Reinforcement Learning
Monte Carlo Tree Search
Here you have it. The ultimate list of Machine and Deep Learning Courses. Some
of them may be too advanced, some may contain too much math, some may be too
expensive but each one of them is guaranteed to teach all you need to succeed in
the AI field.
And to be honest, it doesn’t really matter which one you’ll choose. All of them
are top-notch. The important thing is to pick one and just start learning.
Originally published in AI Summer