Deep learning has emerged as a revolutionary field which has tremendous potential and applications. It is a subset of Machine Learning which is a very broad field.
Two of the major facilitators for the success are:
- Access to computational power: GPUs (graphics processing units) have made fast processing of huge matrices.
- Access to humongous volumes of data: We have more data than ever. In fact it may very well have become even more precious than oil.
Applications range from Healthcare to Autonomous Vehicles. In layman terms, deep learning models are black boxes that learn how to perform certain tasks by just looking at examples. The black box is not explicitly programmed to perform the task but rather “learns” how to do the task.
Traditionally we relied on hand crafted features developed by domain experts for computer vision tasks. However, neural networks are capable of learning complex representations that are excellent for these tasks through looking at lots of data.
Most of the advances that have been made are in supervised learning. This is where labels are available for data. For example you have a lot of images of apples and oranges and know which is which. However, there is a lot of scope in unsupervised learning techniques were there are no labels.
Different architectures exist today which are suitable for different tasks. For example, Convolutional Neural Networks for image classification and segmentation tasks, Recurrent Neural Networks for Natural Language processing etc.
For creating and training deep learning architecture there are several frameworks available. I’ve listed some of them below.
These are in no particular order and the list is not exhaustive. I am not going into the technical details in this post. Will start with future posts.
If you want to learn about deep learning quickly do consider doing the Deep Learning specialization at Coursera. It has helped me tremendously.
I’ll be sharing articles and projects on Deep Learning, Computer Vision and Machine Learning on this blog. Hopefully it will be useful for you. Thank you for reading the post. If you liked it do share and hit subscribe.