The art of neural networks: Artificial intelligence and art (Ted talk inspired)
Recently, with parallel processing (huge computation power at very small cost) machine learning has come up with interesteing techniques such as neural networks.
Neural networks is a form of computing when people started thinking of brain as computer and tried to make a computing architecture with very simple units units like neurons which can do simple calculations (such as a function of addition) and have connections with other neurons and the connections are tunable. Each neurons can receive information from many other neurons and sends it downstream to other neurons. These are all connected in a sort of network with layers as shown below. Note that each layer only connects to the next layer. The information such as a pixel comes in to the input layer in this case. Connections have weights (random at first). At the far side of the network there are fewer and fewer neurons.
We need to train this network with thousands of images and we tell the network the answer over and over iteratively so as to strengthen (incrense the weight) the right connections. The computer can finally identify even from pictures it has never seen before. It thus generalizes what makes a cat and dog. The initial layers are reactive to features such as edges, in the inner layers the layers can extract higher order features such as like an eye and final layers can understand it is a cat. Our brain is seen to work similarly.
The reverse of this should also work, it is called a generative network/ recurrent network. Someone came up with an algorithm where you show a network that is trained on a say a picture, run the network forward, whatever the network saw we will adjust the pixels in the picture towards this interpretation.
Early layers care about simple features, but as we move through the network we have higher order features.
You may thing these are turtle but not really: it is the brain which interprets it as turtle:
This machine would organize itself to perform an action similar to how the brain figures it out. Same thing can be done with text, music etc. (Extract patterns from large amounts of data)
How can these networks be used for arts or to inspire arts? Its an exciting time.
Deep neural networks is an extension of neural networks where the computer goes one step further and each neuron figures out its own function as well. In one of the recent papers treating art with deep networks (linked below): we take a photograph, choose a painting and the network applies the artistic sytle of the painting to the image and transforms it as shown below: