Hi! I am Rares Dolga, a PhD student in the Foundational AI program at UCL, supervised by Prof. David Barber. I also work as a part-time research intern at UiPath. Prior to starting my PhD, I worked as a quantitative developer and later as an AI researcher at J.P. Morgan. My work primarily revolves around probabilistic machine learning, with a focus on developing efficient methods for long-sequence generation in both text and multimodal contexts. Checkout my papers below and blog for a deeper dive into my research.
Journal Articles
Latte: Latent Attention for Linear Time Transformers
Published in NGSM ICML Workshop, 2024
This paper is about a linear latent variable re-parametrisation of attention and combining it with local standard attention
Recommended citation: Dolga, Rares, Marius Cobzarenco, and David Barber. "Latent Attention for Linear Time Transformers." arXiv preprint arXiv:2402.17512 (2024).
Download Paper
RotRNN: Modelling Long Sequences with Rotations
Published in NGSM ICML Workshop, 2024
This paper is about parameterising linear recursive neural networks with rotation matrices.
Recommended citation: Dolga, Rares, et al. "RotRNN: Modelling Long Sequences with Rotations." arXiv preprint arXiv:2407.07239 (2024).
Download Paper