In recent years, deep neural networks have attracted the attention of both the academic community and the general public. An effort to theoretically understand the intricacies of these systems is...Show moreIn recent years, deep neural networks have attracted the attention of both the academic community and the general public. An effort to theoretically understand the intricacies of these systems is ongoing and physics-inspired approaches may have a part to play. In this thesis, we will discuss recent results in the theoretical study of deep linear neural networks. This class of neural networks has very limited real-world applications, but it could provide a good training ground for developing theoretical techniques that could prove useful beyond the simple linear case. We will also argue that Fisher information, and in particular “sloppy model” logic, can be a useful tool for future research on deep neural networks, in particular for network architecture optimization.Show less
Parton distribution functions (PDFs) are vitally important for high energy physics calculations. Vast amounts of experimental evidence have shown that scattering processes involving nuclei cannot...Show moreParton distribution functions (PDFs) are vitally important for high energy physics calculations. Vast amounts of experimental evidence have shown that scattering processes involving nuclei cannot be solved using the free-nucleon formalism of perturbative QCD and therefore, a separate empirical determination of the nuclear modification of PDFs is necessary. Because the shape and size of nuclear modification are theoretically unmotivated, the NNPDF collaboration uses a neural network to achieve a model-independent parametrisation. In this thesis, we include new Z boson production data from pPb collisions into the NNPDF framework and examine its impact on the quality of the fit. We will also discuss the phenomenological implications of prompt photon production data in pPb collisions.Show less
