Papers and abstracts:
Researching scientific machine learning methods for predicting the sound field in dynamic environments; using domain decomposition with numerical methods for solving the wave equation.
Visiting Professor George Karniadakis working on applying neural operators to learn the wave equation for acoustic wave propagations in rooms.
Specialized in Mathematical Modelling and Computations.
Completed first year at the rhythmic section with the guitar as my main instrument.
This PhD thesis examines novel scientific machine learning methods to overcome some of the limitations of traditional numerical methods for learning the wave propagation in rooms for parameterized source positions. Also, a domain decomposition method was developed for coupling two numerical methods for accelerating the solutions to the wave equation.
The thesis investigated whether it was possible to implement finite-difference time-domain methods for solving the 3-D wave equation in real-time on the GPU using C++/CUDA with focus on physical correct simulations of the lower frequency sound field.
Bachelor thesis on the topic 'algorithmic composition' using different machine learning techniques. A software tool was developed for automatically generating music phrase variation and harmonisation from user input, each conforming to a predefined musical genre.
Notus is a domain-specific language for expressing musical structures in the high-level, declarative style of functional programming written in Swift.
A method for harmonising rhythmic music is presented. It uses a hidden Markov model to learn harmonisations of different artists in different genres and allows new chord sequences to be generated with respect to a given melody.
I have started composing and playing music again after years of hibernation. Some rough sketches for my project San Diego can be found on SoundCloud.
Gave a 45 min presentation as part of my PhD defence with the title "Accelerated methods for computing acoustic sound fields in dynamic virtual environments with moving sources".
Gave a 25 minutes talk about Livetake and how to engage concertgoers.
A method for harmonising rhythmic music, implemented in Haskell, will be presented. It uses a hidden Markov model to learn harmonisations of different artists in different genres and allows new chord sequences to be generated with respect to a given melody. The main focus of this talk is on how to perform the feature extraction for the chords and the melody lines necessary for the method to perform well.
Livetake participated in the TechCrunch Disrupt Event 2015 in San Francisco and demonstrated their latest technology. The prize for the technology that most enhance live concert experiences was given to Livetake.
Livetake won the prize for the most promising startup in Denmark within the audio industry. The award was given by the Danish Sound Cluster, an innovation network with the aim of connecting start-ups, established companies and knowledge institutions within the Danish Sound industry.