We are at crossroads in deep learning. Today, deep learning developers typically utilize one of the top two machine learning frameworks: Tensorflow, developed by Google/Deepmind, and PyTorch, developed by Facebook. In industry, Tensorflow is still more widely adopted. Still, PyTorch is rapidly up-and-coming in the research community, where 70%-80% of recently submitted conference research papers utilize PyTorch instead of Tensorflow. A recent 2020 Stack Overflow survey of the most popular frameworks and libraries reported that PyTorch was selected by an est 30% of respondents vs. 70% for Tensorflow, with PyTorch nearly doubling in popularity over the last two years. In the next couple of years, as these machine learning frameworks become equal in popularity, a book must well verse developers in both so they can choose the right methodology to help solve their deep learning problems.
The problem is that most deep learning books published today focus on just one of the machine learning frameworks. Python Deep Learning would identify both frameworks' pros and cons and then teach deep learning concepts utilizing practical examples from the framework best suited for particular problems. This book also features the APIs and libraries integrated with the respective framework, Keras for Tensorflow and fastai for PyTorch, that make application development and deployment even more straightforward.
What this Books Covers:
- Introduction and overview of deep learning concepts
- Description of the two machine learning frameworks: Tensorflow and PyTorch, as well as successful examples of their usage
- Detail the pros and cons of each machine learning framework
- Overview of the supportive libraries and APIs (including Keras and fastai) for each of the frameworks that make application development simpler
- Chapter-by-chapter review of the top neural network topologies (CNN, RNN, LSTM, MLP, and several newer variants)
- Interesting code examples of practical applications of the different neural networks, not the same tired MNIST and other examples often utilized today
- Final series of code examples (in Tensorflow or PyTorch) of real-world deep learning solutions that utilize more exotic neural network topologies
