So every layer has a purpose for processing data, but usually this is a "blackbox".
Recent scientific results reveal some inner workings of the layers (see link to PDF below)
The beauty of this layered approach is that by "re-training" ("fine tuning") of the model, one or a few top layers, already trained model can be reused for processing many related tasks, reducing training time significantly. This may enable distributed learning, since remote and mobile devices could have enough processing power to "learn" on top base models created on large cloud based neural networks and shared in compact form.
Metaphorically, like text-based books are methods of distributing human knowledge, there could be "AI Books" to distributed Neural Network models for AI tools and devices. A brave new world!
How convolutional neural networks see the world
"Keras is a Deep Learning library for Python, that is simple, modular, and extensible."
"VGG16 (also called OxfordNet) is a convolutional neural network architecture named after the Visual Geometry Group from Oxford, who developed it. It was used to win the ILSVR (ImageNet) competition in 2014."
Visualizing and Understanding Convolutional Networks : PDF
"Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark. However there is no clear understanding of why they perform so well, or how they might be improved. In this paper we address both issues."
ImageNet .org
"ImageNet is an image database organized according to the WordNet hierarchy (currently only the nouns), in which each node of the hierarchy is depicted by hundreds and thousands of images.
"Feature Learning Escapades
Stanford University CS231n: Convolutional Neural Networks for Visual Recognition
"Convolutional Neural Networks", mentioned even in Apple WWDC keynote: )
(as a feature of Mac OS)
Deep Visualization Toolbox - YouTube
Visualizing and Understanding Convolutional Networks : PDF
"Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark. However there is no clear understanding of why they perform so well, or how they might be improved. In this paper we address both issues."
ImageNet .org
"ImageNet is an image database organized according to the WordNet hierarchy (currently only the nouns), in which each node of the hierarchy is depicted by hundreds and thousands of images.
"Feature Learning Escapades
Stanford University CS231n: Convolutional Neural Networks for Visual Recognition
"Convolutional Neural Networks", mentioned even in Apple WWDC keynote: )
(as a feature of Mac OS)
Deep Visualization Toolbox - YouTube
.