Go Deep In Deep Learning:
My parallel master degree in deep learning

Realized in 2018 Tags: machine learning, coursera, supervised learning, unsupervised learning, neural network

andrew_hinton_khan

Hi, my name is Luca Sorrentino and I live in Turin, Italy.
I graduated in 2017 in computer science and then I started a Master Degree at Università di Torino with specialization in artificial intelligence.
At university, we study a lot of interesting things, but not enough on neural networks, beacuse dedicating the right space to them would take too much time for the other courses, so we only had an introduction and some practical examples. This isn't enough for me, I want to go deep in deep learning.
So, in 02/18/2018 I started my parallel master degree.


Index


Why a parallel master degree?

While I was searching for more detail about the Boltzmann Machine, I discovered the Hinton's course on Coursera. It was unbelievable because Hinton is the inventor of them and his course is completely free. The whole concept of Coursera is fantastic and I was very excited, so I followed the videos (two weeks of the course) in order to prepare my university's exam.

During my studies, I discovered Daniel Bourke and he was a game changer.

I realize that I could start from scratch and go beyond the exam, becoming really confident in this field. So I prepared a plan for 2 years, taking the best online courses that I found from Coursera, Khan Academy, Google, fast.ai, Mit... following them during my study at university. In the follows paragraphs I make available all my personal notes, all my solved questionnaires with relative scores (all with 100% but I covered my solutions to respect the honor code on Coursera) and a screenshot from each programming assignments.

NB: My personal notes are not explanations of the topics covered in the courses, they are just notes. So this means that they can have some errors and moreover that they are written in Italian (my first language).


Khan Academy

Do you want to play with magic? First of all, you will need to know the basics, so mathematics and statistics.

A good resource to study is The Deep Learning Textbook but if you are on entry-level I really suggest you Khan Academy.
Khan Academy is a non-profit educational organization created in 2006 by educator Salman Khan with a goal of creating a set of online tools that help educate students.[7] The organization produces short lectures in the form of YouTube videos.[8] Its website also includes supplementary practice exercises and materials for educators. All resources are available to users of the website.

Khan Academy Introduction to Matrices
Khan Academy Linear Algebra
Khan Academy Multivariable Calculus
Khan Academy Statistics and Probability


Machine Learning on Coursera

This course provides a broad introduction to machine learning, datamining, and statistical pattern recognition. Topics include:

  1. (I) Supervised learning (linear regression, logistic regression, neural networks, support vector machines).
  2. (II) Unsupervised learning (clustering, dimensionality reduction, recommender systems, PCA, Anomaly detection).
  3. (III) Best practices in machine learning (bias/variance theory,Regoularization,evaluation of learning algorithms, error analysis, ciling analysis, large scle machine learning, debuggin ML algorithms).
This course is taught by Andrew Ng: Co-founder, Coursera; Adjunct Professor, Stanford University; formerly head of Baidu AI Group/Google Brain.

  1. Week:
  2. Week:
  3. Week:
  4. Week:
  5. Week:
  6. Week:
  7. Week:
  8. Week:
  9. Week:
  10. Week:
  11. Week:

Neural Networks for Machine Learning on Coursera

Through this course, you can learn about artificial neural networks and how they're being used for machine learning, as applied to speech and object recognition, image segmentation, modeling language and human motion, etc. We'll emphasize both the basic algorithms and the practical tricks needed to get them to work well.
This course is taught by Geoffrey Hinton : one of the researchers who introduced the back-propagation algorithm that has been widely used for practical applications. His other contributions to neural network research include Boltzmann machines, distributed representations, time-delay neural nets, mixtures of experts, variational learning, products of experts and deep belief nets. You can find all the slides at this Link.

  1. Week:
  2. Week:
  3. Week:
  4. Week:
  5. Week:
  6. Week:

    What's happened in week 6?
    I realized that I was losing my time without increase my skills. Of course, I learned a lot of new things but i'm not able to replicate them in code, I think that 4 programming assignments in 16 weeks are too few and that the arguments must be explained in more details to have a full understanding of them. This course is not for beginners or intermediate students of machine learning. Most appropriate as an overview of the field by a master that experts and practitioners may find interesting. BUT...

  7. Last lessons before quitting the course:

    I could not miss the opportunity to learn RBM from one of the men who invented them. So, I skipped a week and I followed another 3 weeks before quitting. I left here my last notes of this course.


What's Next?

After 125 hours of video lessons, quizzes, programming assignments, writing notes and building this blog page, I have a solid foundation on ML. Now I am ready to move on to phase 2 of my parallel degree:"Deep Learning Specialization" by Andrew Ng on Coursera. He's a great teacher, I'm really happy to learn from him.

You can find the second part here.