Reconstructing the local density field with combined convolutional and point cloud architecture
Baptiste Barthe--Gold · Nhat-Minh Nguyen · Leander Thiele
Abstract
We construct a neural network to perform regression on the local density field given line-of-sight peculiar velocities of tracers. Our architecture combines a convolutional U-Net with a point-cloud DeepSets. This combination enables efficient use of small-scale information and improves the model's reconstruction quality relative to a U-Net-only approach.
Chat is not available.
Successful Page Load