Line intensity maps (LIMs) are in principle sensitive to a large amount of information about faint, distant galaxies which are invisible to conventional surveys. However, actually extracting that information from a confused, foreground-contaminated map can be challenging. In this work we present the first application of convolutional neural network (CNN) to directly determine the underlying luminosity function of a LIM, including a treatment of extragalactic foregrounds and instrumental noise. We apply the CNN to simulations of mock Carbon Monoxide (CO) line intensity maps similar to those which will be produced by the currently-active COMAP experiment. We evaluate the trained CNN on a number of noise scenarios in order to determine how robust the network predictions are for application to realistic data. We find that, in the ideal case where the mock data capture all of the features of the real data, the CNN performs comparably to or better than conventional analyses. However, the network's accuracy degrades considerably when tested on signals and systematics outside of those it was trained on. For both intensity mapping and cosmology as a whole, this motivates a broad-based study of whether simulated data can ever be generated with sufficient detail to realize the enormous potential of machine learning methods.

We consider the possibility that tidal disruption events (TDEs) caused by supermassive black holes (SMBHs) in nearby galaxies can account for the ultra-high-energy cosmic-ray (UHECR) hotspot reported recently by the Telescope Array (TA) and the warm spot by Pierre Auger Observatory (PAO). We describe the expected cosmic-ray signal from a TDE and derive the constraints set by the timescale for dispersion due to intergalactic magnetic fields and the accretion time of the SMBH. We find that TDEs in M82 can explain the hotspot detected by the TA regardless of whether the UHECRs are composed of protons or heavier nuclei. We then check for consistency of the hot and warm spots from M82 and Cen A with the full-sky isotropic signal from all SMBHs within the GZK radius. This analysis applies to any scenario in which the hot/warm spots are real and due to M82 and Cen A, regardless of whether TDEs are the source of UHECRs. We find that the isotropic flux implied by the luminosity density inferred from M82 and Cen A is bigger than that observed by roughly an order of magnitude, but we provide several possible explanations, including the possibility of a local overdensity and the possibility of intermediate-mass nuclei in UHECRs, to resolve the tension.