Radar is a central system in the field of electronic warfare used to estimate anobject’s location, speed, and direction. A pulse radar emits pulses at predeter-mined time intervals. The interval between two pulses is often referred to asthe pulse repetition interval (PRI) and is an important parameter used for theradar identification problem in a radar warning receiver (RWR) system. Beingable to properly classify PRI modulation is crucial to avoiding casualties in thebattlefield and being able to map the opponents’ assets.One of the challenges in the defense industry is the lack of data sharing be-tween different parties. It can, therefore, be hard and expensive for companiesto collect real authentic data. A framework which could increase the size ofa data set from an existing pool of data without duplicating it, is therefore at-tractive in the military industry. Due to the essential role the PRI parameterplays, a framework of this kind could assist in improving an RWR system withmachine learning by expanding the amount of PRI data available for training.This thesis investigates two different kinds of generative deep learning modelsto generate PRI sequences of six different classes. The models experimentedwith are a traditional long short-term memory (LSTM) network and a gener-ative adversarial network (GAN). The models are evaluated on how similarthey can generate data in comparison to the data trained on. The results showthat the GAN model performs better than the LSTM model, but neither of themodels achieves results that can be used to improve an RWR system as of to-day. Future work involving experimentation with network configurations, dataset sizes and other settings, is therefore suggested to improve the results.