Last year I finished Machine Learning Coursera course by Stanford University and Andrews Ng
Is a free course about Machine Learning and a little of Deep Learning created by Andrew Ng and Stanford University. Although it’s free you can to purchase a certificate by $70. It is divided into 3 basis , videos, quizzies and programation excercises.
You see the videos, do a quizz and a practice exercices, designing a part of an algorithm and implementing and testing it with Matlab or Octave
What I think about
I like as an introduction, but I think is not fastest way to get in Machine Learning world. It is a very theoretical course. This is a good one to gain a general global vision
- How to select an algorithm.
- How to select and define its parameters.
- What problems you can get with algorithms and how to correct them.
Another bad point is to work with Matlab or Octave.
They argue that Matlab or the free Opensource alternative, Octave are a very fast tools for doing prototypes, and that is true, however, when you finish the course and try to dive into real life everyone is using python.
I worked with Octave and I spent a lot of time learning how to use it and its matrix tools.
Exercises are not trivial but there are guided, they guide you, giving and preparing the environment and datasheet. So that you only have to write a few lines of the core algorithm.
Algebra (Vector and Matrixes) and a little of Calculus is what you will find about Mathematics in the Andrew Ng Machine Learning Coursera course.
Here a slide with the logisti regression cost function. In exercisies you must to implement this formula with Matlab or Octave language and using Matrix notation
When I listened to talk about Machine Learning I imagined something similar to traditional programming, a lot of ifs and a lot of statistics. And this course has change my point of view giving me a better global vision of how this matter works.
Syllabus
If you want to know what course is in detail, here I let you de syllabus. We could divided it into two parts: Supervised and Unsupervised.
Supervised Learning
- Model and Cost function
- Gradient descent for linear regression
- Regularization
- Neural Networks
- Large Machine Classification y Kernels
- Principal Component Analysis (PCA)
- Machine Learning system design
- Support Vector Machines
Unsupervised Learning
- Dimensionality Reduction
- Anomaly Detection
- Recommender Systems
- Large Scale Machine Learning
If you want a more detailed summary by week
Week 1
Introduction
Linear Regression with One Variable
Linear Algebra Review
Week 2
Linear Regression with Multiple Variables
ctave/Matlab Tutorial
Week 3
Logistic Regression
Regularization
Week 4
Neural Networks: Representation
Week 5
Neural Networks: Learning
Week 6
Advice for Applying Machine Learning
Machine Learning System Design
Week 7
Support Vector Machines
Week 8
Unsupervised Learning
Dimensionality Reduction
Week 9
Anomaly Detection
Recommender Systems
Week 10
Large Scale Machine Learning
Week 11
Application Example: Photo OCR
Resources
- Use the coursera forums
- And although I don’t recommend it there are github with the solutions for exercises or quizzies. I know there are a lot of people who cheating with Github, but remember that you sign an ethic document. And YOU ARE TAKING THIS COURSE TO, NO TO FINISH IT.
What to do when you finish this course
I see it as an introduction to Machine Learning world to learn concepts, But you should study more when you finish the course, and do more focused courses. I recommend you someone in this post (Machine Learning and Deep Dearning courses)
Learn Python and frameworks like Tensor Flow, Keras, PyTorch or Numpy Python libraries.
Learn more deep learning and focus, focus, focus.
And work in a real project or take part in competitions. Kaggle is a good choice.
Do you have any experience? Please comment and tell it us.