be a good choice. This library lets you build your own deep nets with a sophisticated set

of layer configuration options. You can even access premade nets that were uploaded to

a community website. Let’s take a look.

The Caffe Deep Learning Library was created by Google’s Yangqing Jia, who won an ImageNet

Challenge in 2014.

Caffe was originally designed for machine vision tasks, so it’s well-suited for convolutional

nets. However, recent versions of the library provide support for speech and text, reinforcement

learning, and recurrent nets for sequence processing. Since the library is written in

C++ with CUDA, applications can easily switch between a CPU and a GPU as needed. Matlab

and Python interfaces are also available for Caffe.

With Caffe, you can build a deep net by configuring its hyper-parameters. In fact, the layer configuration

options are very sophisticated. You can create a net with many different types of layers,

such as a vision layer, a loss layer, an activation layer, and a few others. So each layer can

perform a different function or take on a different role. This flexibility allows you

to develop extremely complex deep nets for your application. Caffe is supported by a

large community where users can contribute their own deep net to a repository known as

the “Model Zoo”. AlexNet and GoogleNet are two popular user-made nets available to

the community. There are also a few educational resources like demos and slides, so if you’re

going to use Caffe, it’s a great place to start.

Caffe vectorizes input data through a special data representation called a “blob”. A

“blob” is a type of array that speeds up data analysis and provides synchronization

capabilities between a CPU and a GPU.

Have you ever used the Caffe library in one of your own Deep Net projects? Please comment

and share your experiences.