Black Hole Weather Forecasting with Deep Learning: A Pilot Study

This repository contains the data files containing the convolutional neural network weights resulting from our multi GPU training using Tensorflow v1.8.0. Both files are in the HDF5 binary format. The results from this architecture was published in <em>Black Hole Weather Forecasting with Deep Learning: A Pilot Study</em> (Duarte, Nemmen &amp; Navarro, MNRAS, in press). • dl_fluids.h5: one-sim model• dl_fluids-1.h5: multi-sim modelPlease refer to the source file<code>inference.py</code> available in the github repository (https://github.com/black-hole-group/DL_BH_fluids), which demonstrates how to create predictions using these data files. In particular, you can load the weights using <code>model.load("/path/to/.h5/dl_fluids.h5").</code><br>