Machine learning is a method of data analysis that automates analytical model building. It is a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention. The function and application of Neural Network are at the heart of Machine Learning.
Advances in Machine Learning have developed our current day Consumer Industry and changed the way we shop and interact in this industry. Social corners from Social Media (think Facebook and Twitter) to Finance (think AmEx monitoring your charge card for fraud activity) have developed Machine Learning to make choices and provide services for the consumer.
And now Google, as they often do, has taken Machine Learning one step further. They let the network dream and develop it’s own thoughts and take it’s own direction.
Like other Neural Network developers, Google trained a network by showing it many examples (pictures) of what they wanted it to learn in the hopes the significant features that define an object will be retained by the network. For example a tire needs to be round, can be any color and size, and have multiple different textures.
Google trained one particular network to recognize animals, among many other images. They then created a feedback loop, telling the machine “Whatever you see there, I want more of it!”. By introducing higher-level layers the network can identify complex features. When showing the network an image of a big sky speckled with cloud formations, the network “saw” images of animals in the clouds, like a fish for example, much like a person would. Using the feedback loop, the network would then make the cloud look more like the fish it “saw” on each iteration until a highly detailed image of a fish appeared.
The network can dream – and the results are fascinating. Even a simple network can be trained to gaze at the clouds like a person and “See” images that we recognize.