Machine Learning Night: Fastai 2019 6.2 - Convolutional NN In Depth

This week we'll finish fastai Lesson 6 with an in-depth
look at convolutional neural networks.

Slack channel for the meetup (thanks, Wenchang!):
https://fatcatml.slack.com
Invite link to join:
https://join.slack.com/t/fatcatml/shared_invite/enQtNzE3ODA0MDgxNDU3LWE2Yjc4M2Q2Y2FlYmE4Njk0ZTBiMGRjM2Y4N2I3MDVmYTE2NzFiMzhjNWQzM2E3MTQ4ZWViZGYyOTdlNTg2MmI

Note: this isn't sponsored by fastai - we'll just be going
through their 2019 course week by week and learning it together.

Fastai is one of the more exciting ML projects out there.
Over the next weeks I want to try going through the new (2019)
fastai deep learning course, formally called:
Practical Deep Learning for Coders, v3
This course is now the first link on the fastai page: https://www.fast.ai/
or it can be accessed directly at : https://course.fast.ai/

We'll go through the notebooks for the course lessons together,
run them, and discuss. If you want to, you can follow along on
your laptop, and do your own experiments with the notebook code.

There are 7 lessons in the course, and I plan to do each lesson over two weeks.
This will allow you to get all the lessons even if you have to miss a week.
Also, if anyone does some experimenting on their own after the first week
and wants to talk about what they did they can do so in the second week.

To follow along, you should be familiar with Python and with using
Jupyter notebooks. You should also have an environment set up for
executing the course notebooks - there are many options for this
ranging from easy prepackaged environments like Kaggle or Colab,
to setting up a remote server on AWS, to setting up your own system.
The easy and free options include Kaggle and Colab.
Kaggle is a very easy way to get started - you just need to sign
up for an account (one-click), fork a kernel for one of the
course lessons, and start editing. For more details, see:
https://course.fast.ai/start_kaggle.html
Note: if using Kaggle, be sure to turn Internet on in the settings
so you can download external datasets from your notebook.

For info on other options, go to https://course.fast.ai
and click on "Server setup" on the left side.

If you have your own ML setup with Anaconda and NVidia GPU drivers
already installed, you can probably set up the required fastai
libraries as follows:
conda install -c pytorch -c fastai fastai
Note - this installs fastai in your current environment (must be Python >= 3.6).
You might instead want to set up a separate environment for fastai.
Then download the course notebooks:
git clone https://github.com/fastai/course-v3
You can then start the jupyter notebook server:
cd course-v3/nbs/dl1
jupyter notebook
and access it through localhost:8888 on your browser.
You should duplicate the course notebooks and do your experiments on
the duplicates to avoid problems updating the notebooks from github
in the future.

Machine Learning Night happens every Monday at Fat Cat Fab Lab. Our group will discuss Machine Learning, Information Retrieval, Natural Language Processing, Knowledge Representation, and Artificial Intelligence.

We are a small group of students, computer scientists, engineers, academics, and entrepreneurs with a deep interest in these fields & related technologies.

Meeting topics are varied and can focus on current trends and technologies, best practices for development in big-data environments, and introductory discussions about machine learning and how to incorporate it into your life and business

• To Attend, you must join our Meetup group and RSVP. Please note: All meetups held at Fat Cat Fab Lab are private events. They are for members of the Fat Cat Fab Lab Meetup group only. All events are for persons 18 years and older.