Apps written by other quantum programmers for your inspiration.
Simulating the folding of proteins could lead to a radical transformation of our understanding of complex biological systems and our ability to design powerful new drugs. This applications looks into how to use the quantum computer to explore the possible folding configurations of these interesting molecules.
Using unsupervised machine learning approaches, one can automate the discovery of a very sparse way to represent objects. This technique can be used for incredibly efficient compression. In this initial application, we used the algorithm to compress a simple movie. This movie is called 'Frey faces' and is a well-known test set for new compression algorithms. Our approach worked well, compressing the movie by a factor of approximately 50x.
Is there a car in this image? Quantum hardware, trained using a binary classification algorithm, is able to detect whether or not an image contains a car. Even though all the software receives is a collection of pixels, by training the system it can learn to spot cars, even if it has never seen the exact photograph before.
Searching a database of molecules
We developed an application for searching a database of three-dimensional molecules for a ranked list of molecules similar to a query molecule. This application was developed and run on the 28-qubit Leda processor, and demonstrated at Supercomputing 2007.
Labeling of news stories
In this application, we built software for automatically applying category labels to news stories. The corpus we used for training and testing performance was the REUTERS corpus, a well-known data set for testing multiple label assignment algorithms. We found that our approach worked extremely well on this problem, providing 15.4% better labeling accuracy than a state of the art conventional approach.
This application is related to the compressive sensing (video compression) application. The program first learns features from a sequence of video, and then uses the features to generate entirely new sequences of video. As the project progresses, the system will learn about more objects and scenes from the natural world, and the scope of its generated imagery will improve.
This application learns to assign sentiment labels to blog posts. In the initial application, there was a single binary label indicating whether the tone of the post was considered by a person to be positive or negative. This could readily be expanded to multiple labels with more types of sentiment assessed (such as anger, joy, etc.).