arXiv subjects

Jonathan Cohen

Publications and source records attributed to Jonathan Cohen.

39 records · Page 3Linked to original sources

arXiv2016

Contact Interactions in Higgs-Vector Boson Associated Production at the ILC

We explore new physics (NP) effects in Higgs-Vector boson associated production at a future International Linear Collider (ILC) via e+e- -> Zh,Zhh, using effective field theory (EFT) techniques. In particular, we focus on a certain class of dimension 6 operators, which are generated by tree-level exchanges of a new heavy vector field in the underlying theory. These operators induce new contact terms of the form ψψϕDϕ, involving the Standard Model (SM) fermions (ψ), gauge-bosons (D is the covariant derivative) and the SM Higgs field (ϕ). We investigate the high-energy behaviour of these new effective interactions in e+e- -> Zh,Zhh, imposing bounds from electroweak precision measurements, and show that the ILC is an excellent testing ground for probing this type of NP via e+e- -> Zh,Zhh. We also address the validity of the EFT expansion and we study the correlation between the hZ and hhZ signals, which can be utilized in future searches for NP in these channels.

hep-ph↗

arXiv2014

cuDNN: Efficient Primitives for Deep Learning

We present a library of efficient implementations of deep learning primitives. Deep learning workloads are computationally intensive, and optimizing their kernels is difficult and time-consuming. As parallel architectures evolve, kernels must be reoptimized, which makes maintaining codebases difficult over time. Similar issues have long been addressed in the HPC community by libraries such as the Basic Linear Algebra Subroutines (BLAS). However, there is no analogous library for deep learning. Without such a library, researchers implementing deep learning workloads on parallel processors must create and optimize their own implementations of the main computational kernels, and this work must be repeated as new parallel processors emerge. To address this problem, we have created a library similar in intent to BLAS, with optimized routines for deep learning workloads. Our implementation contains routines for GPUs, although similarly to the BLAS library, these routines could be implemented for other platforms. The library is easy to integrate into existing frameworks, and provides optimized performance and memory usage. For example, integrating cuDNN into Caffe, a popular framework for convolutional networks, improves performance by 36% on a standard model while also reducing memory consumption.

cs.NE↗

arXiv2012

A Parallel Auxiliary Grid AMG Method for GPU

In this paper, we develop a new parallel auxiliary grid algebraic multigrid (AMG) method to leverage the power of graphic processing units (GPUs). In the construction of the hierarchical coarse grid, we use a simple and fixed coarsening procedure based on a region quadtree generated from an auxiliary grid. This allows us to explicitly control the sparsity patterns and operator complexities of the AMG solver. This feature provides (nearly) optimal load balancing and predictable communication patterns, which makes our new algorithm suitable for parallel computing, especially on GPU. We also design a parallel smoother based on the special coloring of the quadtree to accelerate the convergence rate and improve the parallel performance of this solver. Based on the CUDA toolkit [40], we implemented our new parallel auxiliary grid AMG method on GPU and the numerical results of this implementation demonstrate the efficiency of our new method. The results achieve an average speedup of over 4 on quasi-uniform grids and 2 on shape regular grids when compared to the AMG implementation in CUSP.

math.NA↗