Our goal in this chapter is to expose you to some of the powerful functionality that is available with Kelp.Net.
In this chapter, you will learn:
- How to use Kelp.Net to perform your own testing
- How to write tests
- How to do benchmarks of functions
- How to extend Kelp.Net
Kelp.Net4 is a deep learning library written in C# and .NET. With the ability to chain functions into a function stack, it provides an incredible amount of power in a very flexible and intuitive platform. It also takes heavy advantage of the OpenCL language platform to enable seamless operation on both CPU-and GPU-enabled devices. Deep learning is an incredibly powerful tool, and native support for Caffe and Chainer model loading makes this platform even more powerful. As you will see, you can create a 1 million hidden layer deep learning network in just a few lines of code.
Kelp.Net also makes it very easy to save and load models to and from disk storage. This is a very powerful feature, allowing you to perform your training, save the model, and then load and test as required. It also makes it much easier to productionize code and truly separate the training and the test phases.
Among other things, Kelp.Net is an incredibly powerful tool for you to be able to learn and understand better various types of functions, their interactions, and performance. For instance, you can run tests against the same network with different optimizers and see the results by changing a single line of code. Also, you can design your tests easily to see the difference in using various batch sizes, number of hidden layers, epochs, and more. There really is no deep learning workbench for .NET that offers the power and flexibility found in Kelp.Net.
Let's begin by talking a little bit about deep learning.