Deep Learning for Numerical Applications with SAS

Chapter 9: Conclusions

Data-Driven Programming

The Quest for Speed

From Tasks to GPUs

Training and Inference

FPGA

Hybrid Architectures

 

In this concluding book, we look back at our accomplishments and give some pointers for the evolution of deep learning for numerical applications (DL4NA).

Data-Driven Programming

In 2007, Jim Gray of Microsoft Research gave a speech in which he argued that we were entering a fourth paradigm in science (Hey et al. 2009).

His argument was that science and research have gone through several phases:

1.   Experimental Science

This phase started thousands of years ago. It was mostly characterized by the observation of natural phenomena.

2.   Theoretical Science

This phase started a few hundred years ago. It began ...