Style transfer using Generative adversarial networks (GANs) has been successful in recent publications. One field in style transfer is music style transfer, in which a piece of music is transformed in some way, be it through genre-, harmonic-, rhythmic transfer, etc. In this thesis, we have performed genre style transfer using a CycleGAN architecture and symbolic representation of data. Previous work using the same architecture and representation has focused solely on transferring the arrangement of the notes (composition). We have improved this work by including the transfer of multiple instruments (timbre) to create more convincing results. Additional discriminators were added to the CycleGAN architecture to achieve this, and they are individually tasked with enforcing the timbre and composition of a song. Previous works have also used variable autoencoders (VAEs) with sequential data representation for style transfer. The use of VAEs for genre style transfer using symbolic data representation instead of sequential was explored, and recommendations for future work include omitting faults found during exploration. Two different classifiers were created to evaluate the results of the CycleGAN model. One uses symbolic representation, in which all instruments are merged into one, thus evaluating the composition of the generated songs. The other classifier uses a spectrogram representation which evaluates the transfer as a whole, both timbre and composition. The evaluation of the improved CycleGAN model using the classifiers showed that it could perform genre style transfer successfully even when adding timbre to the style transfer.

One way of reducing the examination time in magnetic resonance imaging (MRI) is to reduce the amount of raw data acquired, by performing so-called undersampling. Conventionally, MRI data is acquired line-by-line on a Cartesian grid. In the field of Cardiovascular Magnetic Resonance (CMR), however, radial k-space sampling is seen as a promising emerging technique for rapid image acquisitions, mainly due to its robustness against motion disturbances occurring from the beating heart. Whereas Cartesian undersampling will result in image aliasing, radial undersampling will introduce streak artifacts. The objective of this work was to train the deep learning architecture, CycleGAN, to reduce streak artifacts in radially undersampled CMR images, and to evaluate the model performance. A benefit of using CycleGAN over other deep learning techniques for this application is that it can be trained on unpaired data. In this work, CycleGAN network was trained on 3060 radial and 2775 Cartesian unpaired CMR images acquired in human subjects to learn a mapping between the two image domains. The model was evaluated in comparison to images reconstructed using another emerging technique called GRASP. Whereas more investigation is warranted, the results are promising, suggesting that CycleGAN could be a viable method for effective streak-reduction in clinical applications.