Fast Flow Reconstruction via Robust Invertible n×n Convolution

Flow-based generative models have recently become one of the most efficient approaches to model data generation. Indeed, they are constructed with a sequence of invertible and tractable transformations. Glow first introduced a simple type of generative flow using an invertible 1×1 convolution. However, the 1×1 convolution suffers from limited flexibility compared to the standard convolutions. In this paper, we propose a novel invertible n×n convolution approach that overcomes the limitations of the invertible 1×1 convolution. In addition, our proposed network is not only tractable and invertible but also uses fewer parameters than standard convolutions. The experiments on CIFAR-10, ImageNet and Celeb-HQ datasets, have shown that our invertible n×n convolution helps to improve the performance of generative models significantly.

Four-dimensional tracer flow reconstruction in fractured rock through borehole ground-penetrating radar (GPR) monitoring

Two borehole ground-penetrating radar (GPR) surveys were conducted during saline tracer injection experiments in fully saturated crystalline rock at the Grimsel Test Site in Switzerland. The saline tracer is characterized by an increased electrical conductivity in comparison to formation water. It was injected under steady-state flow conditions into the rock mass that features sub-millimeter fracture apertures. The GPR surveys were designed as time-lapse reflection GPR from separate boreholes and a time-lapse transmission survey between the two boreholes. The local increase in conductivity, introduced by the injected tracer, was captured by GPR in terms of reflectivity increase for the reflection surveys, and attenuation increase for the transmission survey. Data processing and difference imaging was used to extract the tracer signal in the reflection surveys, despite the presence of multiple static reflectors that could shadow the tracer reflection. The transmission survey was analyzed by a difference attenuation inversion scheme, targeting conductivity changes in the tomography plane. By combining the time-lapse difference reflection images, it was possible to reconstruct and visualize the tracer propagation in 3D. This was achieved by calculating the potential radially symmetric tracer reflection locations in each survey and determining their intersections, to delineate the possible tracer locations. Localization ambiguity imposed by the lack of a third borehole for a full triangulation was reduced by including the attenuation tomography results in the analysis. The resulting tracer flow reconstruction was found to be in good agreement with data from conductivity sensors in multiple observation locations in the experiment volume and gave a realistic visualization of the hydrological processes during the tracer experiments. Our methodology was demonstrated to be applicable for monitoring tracer flow and transport and characterizing flow paths related to geothermal reservoirs in crystalline rocks, but it can be transferred in a straightforward manner to other applications, such as radioactive repository monitoring or civil engineering projects.

MLPF: efficient machine-learned particle-flow reconstruction using graph neural networks

In general-purpose particle detectors, the particle-flow algorithm may be used to reconstruct a comprehensive particle-level view of the event by combining information from the calorimeters and the trackers, significantly improving the detector resolution for jets and the missing transverse momentum. In view of the planned high-luminosity upgrade of the CERN Large Hadron Collider (LHC), it is necessary to revisit existing reconstruction algorithms and ensure that both the physics and computational performance are sufficient in an environment with many simultaneous proton–proton interactions (pileup). Machine learning may offer a prospect for computationally efficient event reconstruction that is well-suited to heterogeneous computing platforms, while significantly improving the reconstruction quality over rule-based algorithms for granular detectors. We introduce MLPF, a novel, end-to-end trainable, machine-learned particle-flow algorithm based on parallelizable, computationally efficient, and scalable graph neural network optimized using a multi-task objective on simulated events. We report the physics and computational performance of the MLPF algorithm on a Monte Carlo dataset of top quark–antiquark pairs produced in proton–proton collisions in conditions similar to those expected for the high-luminosity LHC. The MLPF algorithm improves the physics response with respect to a rule-based benchmark algorithm and demonstrates computationally scalable particle-flow reconstruction in a high-pileup environment.

4D Tracer Flow Reconstruction in Fractured Rock through Borehole GPR Monitoring

Two borehole ground penetrating radar (GPR) surveys were conducted during saline tracer injection experiments in fully-saturated crystalline rock at the Grimsel Test Site in Switzerland. The saline tracer is characterized by an increased electrical conductivity in comparison to formation water. It was injected under steady state flow conditions into the rock mass that features sub-mm fracture apertures. The GPR surveys were designed as time-lapse reflection GPR from separate boreholes and a time-lapse transmission survey between the two boreholes. The local increase in conductivity, introduced by the injected tracer, was captured by GPR in terms of reflectivity increase for the reflection surveys, and attenuation increase for the transmission survey. Data processing and difference imaging was used to extract the tracer signal in the reflection surveys, despite the presence of multiple static reflectors that could shadow the tracer reflection. The transmission survey was analyzed by a difference attenuation inversion scheme, targeting conductivity changes in the tomography plane. By combining the time-lapse difference reflection images, it was possible to reconstruct and visualize the tracer propagation in 3D. This was achieved by calculating the potential radially-symmetric tracer reflection locations in each survey and determining their intersections, to delineate the possible tracer locations. Localization ambiguity imposed by the lack of a third borehole for a full triangulation was reduced by including the attenuation tomography results into the analysis. The resulting tracer flow reconstruction was found to be in good agreement with data from conductivity sensors in multiple observation locations in the experiment volume and gave a realistic visualization of the hydrological processes during the tracer experiments. Our methodology proved to be successful for characterizing flow paths related with geothermal reservoirs in crystalline rocks, but it can be transferred in a straightforward manner to other applications, such as radioactive repository monitoring or civil engineering projects.