Making sure that AI is accessible around the world is one important challenge. There is already a lot of resources around the web to learn machine learning for free. However, understanding machine learning concepts car be hard for someone that wants to get started. This is even harder for a nonnative English speaker that have to deal with the language barriers and the theory behind machine learning. This is the reason why I decided to create a full tutorial series of more than 20 vidéos for French students that want to get involved in this domain. I used the alpha version fo Tensorflow 2.0 to teach the course and to make it so that all people that follow my course can be ready to use the library once the official 2.0 version came out.
What it does
The course is designed for anyone that wants to learn machine learning and practice with projects. The courses are designed for beginners with a particular focus on practics. All the code is available on my Github and linked into the description of each video.
The series is split into 5 weeks:
- How does a perceptron work?
- Gradient descent
- Neural network
- How to code a simple neural network
- Data normalization
- The error function
- Train, Test, Validation set
- Which activation function should I use?
- How to use and save a model?
- Eager mode vs Graph mode
- How to train a model with tf.Gradient
- Subclassing API
- Customize Layers
Week 4 (Project 1):
- Recognize drawings
- Handle data with tf.data
- How to create a convolutional neural network
Week 5 (Project 2):
- How to generate Victor Hugo's poetries
- Sequential batches
- One hot encoding
- How to code a recurrent neural network
- How to generate random poetries
How I built it
Just after the first release of Tf2 (tf-nightly-2.0-preview) I used Jupyter notebook to first prepare all the projects and to experiment with the main features of Tensorflow 2.0. Once I got a sense of what's new in Tensorflow I started to prepare the slides of the classes. Once everything has been set up I have recorded the videos. The videos are either theoretical or practical.
Challenges I ran into
During the creation process of the course, I tried to make sure that everything I did can be followed by most of the student with basic knowledge in programming and python as a requirement. What helped me to bring a full course is the other videos that I have on my Youtube channel where I talk about AI. Thus, I used as complementary resources my other videos in the projects I made in the course.
Accomplishments that I'm proud of
Until know the course have been viewed more than 10k times in less than 2 months, which is a lot for a French course in Machine learning. I also have a lot of good feedback and I have the feeling that this course has an impact to help a lot of French students to get started in AI and with Tensorflow 2.0 for free.
What's next for Tensorflow 2.0 course (French)
The course is now finished, but my Youtube channel is not. I created this youtube channel more than 1 year ago and I am determined to continue creating French videos to educate and make AI more accessible to viewers. As of now, all courses have accumulated more than 350,000 views.