Deep Learning with Modern Java Code

An airhacks.fm conversation with Dr. Zoran Sevarac (@zsevarac) about:

DeepNetts is targeting Java developers, nice Java code with DeepNetts, DeepNeetts with two dependencies only, Image Recognition with Duke, the data augmentation for variation generation, DeepNetts supports all formats from java image IO, Convolutional Layer, max Pooling Layer, Fully Connected Layer, max pool layer reduces the dimension of a problem, convolutional layer is about pattern recognition, convolutional layer slides a square shape over an image to recognise a pattern, max pool layer is about downsizing, Fully Connected Layer are classifying the images, the output layers is uses a mathematical soft max function, output layer provides the prediction, the VisRec JSR-381 library, DeepNetts does not rely on the existence of GPU,

Dr. Zoran Sevarac on twitter: @zsevarac and Zoran's deepnetts.com