arXiv subjects
Publications and source records attributed to Andrey Grabovsky.
Motivated by the need to increase the precision of theoretical predictions to test saturation physics at both the LHC and the EIC, we compute the cross-sections for the diffractive single and di-hadron production at the NLO in the shockwave formalism.
We calculate the cross-sections of diffractive single hadron photo- or electroproduction with large $p_T$, on a nucleon or a nucleus in the shockwave formalism. We use the hybrid formalism mixing collinear factorization with high energy small-$x$ factorization with the impact factors computed at next-to-leading order accuracy. We prove the cancellation of divergence and we determine the finite parts of the differential cross-sections. We work in general kinematics such that both photoproduction and leptoproduction are considered. The results can be used to detect saturation effects, at both the future EIC or already at LHC, using Ultra-Peripheral Collisions.
We analyze algorithmic and computational aspects of biological phenomena, such as replication and programmed death, in the context of machine learning. We use two different measures of neuron efficiency to develop machine learning algorithms for adding neurons to the system (i.e. replication algorithm) and removing neurons from the system (i.e. programmed death algorithm). We argue that the programmed death algorithm can be used for compression of neural networks and the replication algorithm can be used for improving performance of the already trained neural networks. We also show that a combined algorithm of programmed death and replication can improve the learning efficiency of arbitrary machine learning systems. The computational advantages of the bio-inspired algorithms are demonstrated by training feedforward neural networks on the MNIST dataset of handwritten images.