FastAI and TenserFlow
I have started learning Deep Learning from last month. I was familiar with some of the concepts and tools for it. I have started DL using FastAI library. This course is taught by Jeremy Howard at course.fast.ai. I really enjoyed this course and learned a lots of new things about deep learning. Initially I was familiar with TenserFlow library build my Google. After learning about FastAI, I was wondering that what are the basic difference between these two libraries since they are doing almost same task. I have got good feedback for fast ai forum. I have concluded this conversation here.
- TenserFlow is more lower level and different than FastAI. If we want to compare then we should compare TenserFlow and PyTorch(these two are also not quite same). FastAI library is build on top of PyTorch which provided lots of APIs that makes data science more fun.
- Whatever you can create using FastAI library, you can replicate most of them using TenserFlow but then you will need do more coding and take care of lots of attributes.
- TenserFlow is more popular than FastAI as of now. If you are new to this field than probably, at least, you might heard the name of TenserFlow but not FastAI. This is because FastAI is very young. Even though new to the field, FastAI is more powerful and flexible. You can customise all of it’s functions easily but if you need very standard things then, you need not to bother about a lots and a lots of attributes and math. In other words, I can say, TensorFlow is more popular but, FastAI works more effectively.
If you are intrested in history of FastAI then visit this blog: http://www.fast.ai/2016/10/07/fastai-launch/
Call To Action
Do you like such interested and personalised stories? Then subscribe to my mailing list. This tool will help you to know more interesting things.
About Author
Mayur is a Senior Software Engineer in his organisation. He has more than four years of experience in designing and building scale able applications using different technologies.
