Neural networks are a type of machine learning algorithm based loosely on how our brains work. Over the past seven years or so, they have gone through a bit of a renaissance (commonly referred to as "deep learning") during which they have emerged from the periphery to become one of the most interesting and fruitful area in machine learning research today. Some of the advances have been so significant that it's renewed the possibility of AI (or software awfully close to it)--a goal that was largely abandoned after the 90s.
To demonstrate why neural networks are interesting, take a look at this demo (use Chrome or other WebGL browser): word vector demo
That demo is a visualization of word vectors trained with a neural network. Notice how similar words get grouped together? Well, the neural network figured all this out on its own. The only thing we provided it was a dataset of all the text from wikipedia. Simply by reading wikipedia, it was able to learn enough information to group all of these words together. That's incredible.
Results like these have companies like Google, Baidu, Microsoft, IBM, and more extremely excited, and many have already implemented deep neural network technology into their products. So where does that leave everyone else who maybe doesn't have the time or money to become a neural network expert or hire a bunch of them?
Enter Ersatz. Ersatz makes it super easy to build and deploy these neural networks and benefit from their functionality in your own software. Ersatz is a set of tools that make it possible for any company to derive powerful insights from their data.
Upload your data: First define your problem and collect the data you want your software to learn from. The data should be supplied as input/output pairs, as in standard classification problems. Once you have this, you upload the data to our secure servers.
Train a model: Now you're ready to train a model. We have a variety of models to choose from. Models (or layers, more specifically) can be combined to form more complex architectures. We currently support fully connected feed forward nets, randomly connected nets (dropout), convolutional nets w/ pooling, autoencoders, and recurrent neural networks. This lineup essentially represents the current state of the art and is a powerful set of tools from which you can solve machine learning problems from.
Use your models in your software: After training a model, it's ready to be used in your software. You interact with the model via our API. Anything you can do via our web interface, you can do with our API, so you can build a full training feedback loop into your software also.