Deep Learning
My biggest passion the last few years have been deep learning and machine learning. As a theorist I have focused on reading papers and taking online courses,
but I also have a few pet projects with somewhat decent documentation. You can find some of these projects below, and some of extra tidbits can also be found on my
github.
Deep reinforcement learning for quantum optimization
This is a project that resulted in a publication (see the publications section), where I use Deep Q-learning to optimize a quantum protocol. The quantum system I optimized is rather complicated to explain, but if you're interested, check out the publication. The code is explained in the documentation below, where the code itself can also be found.
Recurrent neural network for rap generation
In this project I create a tool for scraping lyrics from a lyric database, and use a recurrent neural network to generate rap lyrics. I chose the artist MF Doom, because he is an artist that is known to rhyme within every line, and not just at the end of them. This one is well documented, so go ahead and check out the link below. By the way, the lyrics generated is not safe for children, but I doubt you are a child.
Autoencoders
In this small project I demonstrate some of the applications of autoencoders. Mainly, using them for dimensional reduction in a similar vein as Principal Component Analysis and then performing classification on the reduced data. Another use is denoising of input data, as well as for pretraining when the set of labled datapoints is a subset of the full unlabled data. I will also soon implement variational autoencoders for data generation.
Quantum Neural Networks
One of the hottest topics in physics these days is quantum computing. The hottest topic in computer science is neural networks. Why not combine them? In this presentation that I've held a few times, I first introduce neural networks and quantum computing, and then I go through a few examples of proposed models that combine the two.