In my (however limited) free time, I’ve started reading through the fast.ai book to learn how to create machine learning programs in Python for an upcoming project I’m working on. The course is available as a textbook published by O’reily and as a Jupyter notebook online.
For several years, I’ve had an interest in artificial intelligence, and machine learning specifically. AIs powered by machine learning have become incredibly powerful in recent years, with superhuman abilities for playing games like chess and specialized tasks like cancer detection, and surprisingly human-like abilities in natural language processing and object detection.
After reading through the first three chapters of the book, I was able to put together my own ML model for an image detection task. Using a total of around 500 images, it only took ~6 seconds to train the resnet18 model on an NVIDIA Quadro P5000 GPU (in the cloud). This in particular was astonishing to me because I’d always read about the hundreds of thousands of training examples and days of processing that were required to get a decent model; but with 500 images and 6 seconds of training, the model reached 96% accuracy.
I’m planning to continue reading through the book and build increasingly more complex and useful models, and would highly suggest this course for anyone looking to get into machine learning that knows a little Python. The fast.ai library is incredibly easy to use, and getting my simple model up and running was a breeze.
0 thoughts on “Getting Started with Machine Learning”