Search

Black holes are a recently observed theoretical prediction of General Relativity, characterized by event horizons, from which information cannot escape. Examined through the lenses of quantum mechanics, they can radiate at a definite temperature inverse to their mass and horizon radius. Hawking radiation, whose spectrum was calculated considering particles scattering off black holes, is connected to the paradox of the loss of information falling into them. Information can become non-fungible, due to scrambling. We demonstrate this feature not to be restricted to curved space-times: soft radiation scattering in a flat space-time does scramble information as well. To this end, we compute the scrambling of information through the tripartite mutual information in a scattering process off a black hole and compare it with the flat space-time analog. We show that the scrambling power of the gravitational field of a black hole is negligible with respect to the scrambling power of flat space-time., Comment: 90 pages

Observable scattering processes entail emission-absorption of soft photons. As these degrees of freedom go undetected, some information is lost. Whether some of this information can be recovered in the observation of the hard photons, depends of the actual pattern of the scrambling of information. We compute the information scrambling of photon scattering by the tripartite mutual information in terms of the 2-Renyi entropy, and find a finite amount of scrambling is present. The developed procedure thus sheds novel light on the black hole information loss paradox, showing that scrambling is a byproduct of decoherence achieved by the scattering system in its interaction with the environment, due to the emission-absorption of soft photons in fully unitary processes., Comment: 23 pages, 4 figures

Common image-to-image translation methods rely on joint training over data from both source and target domains. The training process requires concurrent access to both datasets, which hinders data separation and privacy protection; and existing models cannot be easily adapted for translation of new domain pairs. We present Dual Diffusion Implicit Bridges (DDIBs), an image translation method based on diffusion models, that circumvents training on domain pairs. Image translation with DDIBs relies on two diffusion models trained independently on each domain, and is a two-step process: DDIBs first obtain latent encodings for source images with the source diffusion model, and then decode such encodings using the target model to construct target images. Both steps are defined via ordinary differential equations (ODEs), thus the process is cycle consistent only up to discretization errors of the ODE solvers. Theoretically, we interpret DDIBs as concatenation of source to latent, and latent to target Schrodinger Bridges, a form of entropy-regularized optimal transport, to explain the efficacy of the method. Experimentally, we apply DDIBs on synthetic and high-resolution image datasets, to demonstrate their utility in a wide variety of translation tasks and their inherent optimal transport properties., Comment: 18 pages, 12 figures, in the Eleventh International Conference on Learning Representations (ICLR 2023)

We propose a new sampling-based approach for approximate inference in filtering problems. Instead of approximating conditional distributions with a finite set of states, as done in particle filters, our approach approximates the distribution with a weighted sum of functions from a set of continuous functions. Central to the approach is the use of sampling to approximate multiplications in the Bayes filter. We provide theoretical analysis, giving conditions for sampling to give good approximation. We next specialize to the case of weighted sums of Gaussians, and show how properties of Gaussians enable closed-form transition and efficient multiplication. Lastly, we conduct preliminary experiments on a robot localization problem and compare performance with the particle filter, to demonstrate the potential of the proposed method.

Citation function and provenance are two cornerstone tasks in citation analysis. Given a citation, the former task determines its rhetorical role, while the latter locates the text in the cited paper that contains the relevant cited information. We hypothesize that these two tasks are synergistically related, and build a model that validates this claim. For both tasks, we show that a single-layer convolutional neural network (CNN) outperforms existing state-of-the-art baselines. More importantly, we show that the two tasks are indeed synergistic: by jointly training both of the tasks in a multi-task learning setup, we demonstrate additional performance gains. Altogether, our models improve the current state-of-the-arts up to 2\%, with statistical significance for both citation function and provenance prediction tasks.