Practical Deep Learning
Crashcourse in Machine Learning
Ever trouble figuring out how to model your data with AI? This week I have been diving into an awesome free course on Deep Learning to start building my own:
This course teaches the foundations for machine learning, building models from the ground up, and how to deploy models to production. A few things that I already love about this course:
- The teacher believes in a context first teaching style, showing in depth examples first then explaining how they work
- The whole course is on Jupyter Notebook, which makes following along with the code stupid easy
- It’s practical! It’s in the name, the course provides you with industry standard tools and how to apply it to your personal project … all while explaining how things work in a digestible language
If you’re further interested in Jeremy Howard, the founder of FastAI and teacher of this course, this is a great listen for understanding the ideology that helped build this course:
But Why?
When it comes to things like image recognition or language inference, using a pre-built model makes sense. Images and language have already been thoroughly trained on well-known constraints; so no matter the picture or sentence you pass through, the model will likely recognize the data and be able to deduce from there. But when you introduce a custom dataset that with undefined context, you will have a hard time finding an existing model that can accurately digest your data… Thus, sometimes you have to make your own model!
Although this course is mainly introductory, at the end of it you should have the tools to produce a working production model for your dataset. I’ve only just started, so I will check back in with my results after I have completed the course
Please share any other great resources on deep learning that you’ve come across