Search

Multi-Party Computation in the Head (MPCitH) algorithms are appealing candidates in the additional US NIST standardization rounds for Post-Quantum Cryptography (PQC) with respect to key sizes and mathematical hardness assumptions. However, their complexity presents a significant challenge for platforms with limited computational capabilities. To address this issue, we present, to the best of our knowledge, the first design space exploration of MiRitH, a promising MPCitH algorithm, for embedded devices. We develop a library of mixed HW/SW blocks on the Xilinx ZYNQ 7000, and, based on this library, we explore optimal solutions under runtime or FPGA resource constraints for a given public key infrastructure. Our results show that MiRitH is a viable algorithm for embedded devices in terms of runtime and FPGA resource requirements.

Future cosmic microwave background (CMB) experiments are primarily targeting a detection of the primordial $B$-mode polarisation. The faintness of this signal requires exquisite control of systematic effects which may bias the measurements. In this work, we derive requirements on the relative calibration accuracy of the overall polarisation gain ($\Delta g_\nu$) for LiteBIRD experiment, through the application of the blind Needlet Internal Linear Combination (NILC) foreground-cleaning method. We find that minimum variance techniques, as NILC, are less affected by gain calibration uncertainties than a parametric approach, which requires a proper modelling of these instrumental effects. The tightest constraints are obtained for frequency channels where the CMB signal is relatively brighter (166 GHz channel, $\Delta {g}_\nu \approx 0.16 \%$), while, with a parametric approach, the strictest requirements were on foreground-dominated channels. We then propagate gain calibration uncertainties, corresponding to the derived requirements, into all frequency channels simultaneously. We find that the overall impact on the estimated $r$ is lower than the required budget for LiteBIRD by almost a factor $5$. The adopted procedure to derive requirements assumes a simple Galactic model. We therefore assess the robustness of obtained results against more realistic scenarios by injecting the gain calibration uncertainties, according to the requirements, into LiteBIRD simulated maps and assuming intermediate- and high-complexity sky models. In this case, we employ the so-called Multi-Clustering NILC (MC-NILC) foreground-cleaning pipeline and obtain that the impact of gain calibration uncertainties on $r$ is lower than the LiteBIRD gain systematics budget for the intermediate-complexity sky model. For the high-complexity case, instead, it would be necessary to tighten the requirements by a factor $1.8$., Comment: 29 pages, 11 figures

This paper introduces KwicKwocKwac 1.0 (KwicKK), a web application designed to enhance the annotation and enrichment of digital texts in the humanities. KwicKK provides a user-friendly interface that enables scholars and researchers to perform semi-automatic markup of textual documents, facilitating the identification of relevant entities such as people, organizations, and locations. Key functionalities include the visualization of annotated texts using KeyWord in Context (KWIC), KeyWord Out Of Context (KWOC), and KeyWord After Context (KWAC) methodologies, alongside automatic disambiguation of generic references and integration with Wikidata for Linked Open Data connections. The application supports metadata input and offers multiple download formats, promoting accessibility and ease of use. Developed primarily for the National Edition of Aldo Moro's works, KwicKK aims to lower the technical barriers for users while fostering deeper engagement with digital scholarly resources. The architecture leverages contemporary web technologies, ensuring scalability and reliability. Future developments will explore user experience enhancements, collaborative features, and integration of additional data sources., Comment: 10 pages, 5 figures

Although binary classification is a well-studied problem, training reliable classifiers under severe class imbalance remains a challenge. Recent techniques mitigate the ill effects of imbalance on training by modifying the loss functions or optimization methods. We observe that different hyperparameter values on these loss functions perform better at different recall values. We propose to exploit this fact by training one model over a distribution of hyperparameter values--instead of a single value--via Loss Conditional Training (LCT). Experiments show that training over a distribution of hyperparameters not only approximates the performance of several models but actually improves the overall performance of models on both CIFAR and real medical imaging applications, such as melanoma and diabetic retinopathy detection. Furthermore, training models with LCT is more efficient because some hyperparameter tuning can be conducted after training to meet individual needs without needing to retrain from scratch.

Slate recommendation is a technique commonly used on streaming platforms and e-commerce sites to present multiple items together. A significant challenge with slate recommendation is managing the complex combinatorial choice space. Traditional methods often simplify this problem by assuming users engage with only one item at a time. However, this simplification does not reflect the reality, as users often interact with multiple items simultaneously. In this paper, we address the general slate recommendation problem, which accounts for simultaneous engagement with multiple items. We propose a generative approach using Diffusion Models, leveraging their ability to learn structures in high-dimensional data. Our model generates high-quality slates that maximize user satisfaction by overcoming the challenges of the combinatorial choice space. Furthermore, our approach enhances the diversity of recommendations. Extensive offline evaluations on applications such as music playlist generation and e-commerce bundle recommendations show that our model outperforms state-of-the-art baselines in both relevance and diversity., Comment: 9 pages, 5 figures, 3 tables

Large angular scale surveys in the absence of atmosphere are essential for measuring the primordial $B$-mode power spectrum of the Cosmic Microwave Background (CMB). Since this proposed measurement is about three to four orders of magnitude fainter than the temperature anisotropies of the CMB, in-flight calibration of the instruments and active suppression of systematic effects are crucial. We investigate the effect of changing the parameters of the scanning strategy on the in-flight calibration effectiveness, the suppression of the systematic effects themselves, and the ability to distinguish systematic effects by null-tests. Next-generation missions such as LiteBIRD, modulated by a Half-Wave Plate (HWP), will be able to observe polarisation using a single detector, eliminating the need to combine several detectors to measure polarisation, as done in many previous experiments and hence avoiding the consequent systematic effects. While the HWP is expected to suppress many systematic effects, some of them will remain. We use an analytical approach to comprehensively address the mitigation of these systematic effects and identify the characteristics of scanning strategies that are the most effective for implementing a variety of calibration strategies in the multi-dimensional space of common spacecraft scan parameters. We also present Falcons, a fast spacecraft scanning simulator that we developed to investigate this scanning parameter space.

The chaotic dynamics of quantum many-body systems are expected to quickly randomize any structured initial state, delocalizing it in Fock space. In this work, we study the spreading of an initial product state in Hilbert space under dual-unitary dynamics, captured by the inverse participation ratios and the distribution of overlaps (bit-string probabilities). We consider the self-dual kicked Ising model, a minimal model of many-body quantum chaos which can be seen as either a periodically driven Floquet model or a dual-unitary quantum circuit. Both analytically and numerically, we show that the inverse participation ratios rapidly approach their ergodic values, corresponding to those of Haar random states, and establish the emergence of the Porter-Thomas distribution for the overlap distribution. Importantly, this convergence happens exponentially fast in time, with a time scale that is independent of system size. We inspect the effect of local perturbations that break dual-unitarity and show a slowdown of the spreading in Fock space, indicating that dual-unitary circuits are maximally efficient at preparing random states., Comment: 6+2 pages

We assess the capabilities of the LiteBIRD mission to map the hot gas distribution in the Universe through the thermal Sunyaev-Zeldovich (SZ) effect. Our analysis relies on comprehensive simulations incorporating various sources of Galactic and extragalactic foreground emission, while accounting for specific instrumental characteristics of LiteBIRD, such as detector sensitivities, frequency-dependent beam convolution, inhomogeneous sky scanning, and $1/f$ noise. We implement a tailored component-separation pipeline to map the thermal SZ Compton $y$-parameter over 98% of the sky. Despite lower angular resolution for galaxy cluster science, LiteBIRD provides full-sky coverage and, compared to the Planck satellite, enhanced sensitivity, as well as more frequency bands to enable the construction of an all-sky $y$-map, with reduced foreground contamination at large and intermediate angular scales. By combining LiteBIRD and Planck channels in the component-separation pipeline, we obtain an optimal $y$-map that leverages the advantages of both experiments, with the higher angular resolution of the Planck channels enabling the recovery of compact clusters beyond the LiteBIRD beam limitations, and the numerous sensitive LiteBIRD channels further mitigating foregrounds. The added value of LiteBIRD is highlighted through the examination of maps, power spectra, and one-point statistics of the various sky components. After component separation, the $1/f$ noise from LiteBIRD is effectively mitigated below the thermal SZ signal at all multipoles. Cosmological constraints on $S_8=\sigma_8\left(\Omega_{\rm m}/0.3\right)^{0.5}$ obtained from the LiteBIRD-Planck combined $y$-map power spectrum exhibits a 15% reduction in uncertainty compared to constraints from Planck alone. This improvement can be attributed to the increased portion of uncontaminated sky available in the LiteBIRD-Planck combined $y$-map., Comment: 38 pages, 13 figures, abstract shortened. Updated to match version accepted by JCAP

We study the effect of random perturbations in the Human and Nature Dynamics (HANDY) model. HANDY models the interactions between human population, depletion, and consumption of natural resources. HANDY explains how endogenous human--nature interactions could lead to sustainability or collapse in past societies. We introduce a Gaussian random noise perturbation on the population change to represent generic external perturbations. The robustness of the results is investigated with statistical analysis based on probability distributions of specific events. Our study shows that the results of the unperturbed HANDY model are robust under small perturbations of $\lesssim$ 10\% of the Human population. Our results confirm that endogenous dynamics drive the societal cycles. However, exogenous perturbations, such as floods, droughts, earthquakes, volcanic eruptions, infectious disease, epidemics, and wars, can accelerate or delay a collapse cycle., Comment: 26 pages 12 figures

Given the wide use of forgery throughout history, scholars have and are continuously engaged in assessing the authenticity of historical documents. However, online catalogues merely offer descriptive metadata for these documents, relegating discussions about their authenticity to free-text formats, making it difficult to study these assessments at scale. This study explores the generation of structured data about documents' authenticity assessment from natural language texts. Our pipeline exploits Large Language Models (LLMs) to select, extract and classify relevant claims about the topic without the need for training, and Semantic Web technologies to structure and type-validate the LLM's results. The final output is a catalogue of documents whose authenticity has been debated, along with scholars' opinions on their authenticity. This process can serve as a valuable resource for integration into catalogues, allowing room for more intricate queries and analyses on the evolution of these debates over centuries.

This work introduces GPTCast, a generative deep-learning method for ensemble nowcast of radar-based precipitation, inspired by advancements in large language models (LLMs). We employ a GPT model as a forecaster to learn spatiotemporal precipitation dynamics using tokenized radar images. The tokenizer is based on a Quantized Variational Autoencoder featuring a novel reconstruction loss tailored for the skewed distribution of precipitation that promotes faithful reconstruction of high rainfall rates. The approach produces realistic ensemble forecasts and provides probabilistic outputs with accurate uncertainty estimation. The model is trained without resorting to randomness, all variability is learned solely from the data and exposed by model at inference for ensemble generation. We train and test GPTCast using a 6-year radar dataset over the Emilia-Romagna region in Northern Italy, showing superior results compared to state-of-the-art ensemble extrapolation methods., Comment: 20 pages, 10 figures

Downscaling techniques are one of the most prominent applications of Deep Learning (DL) in Earth System Modeling. A robust DL downscaling model can generate high-resolution fields from coarse-scale numerical model simulations, saving the timely and resourceful applications of regional/local models. Additionally, generative DL models have the potential to provide uncertainty information, by generating ensemble-like scenario pools, a task that is computationally prohibitive for traditional numerical simulations. In this study, we apply a Latent Diffusion Model (LDM) to downscale ERA5 data over Italy up to a resolution of 2 km. The high-resolution target data consists of 2-m temperature and 10-m horizontal wind components from a dynamical downscaling performed with COSMO_CLM. Our goal is to demonstrate that recent advancements in generative modeling enable DL to deliver results comparable to those of numerical dynamical models, given the same input data, preserving the realism of fine-scale features and flow characteristics. A selection of predictors from ERA5 is used as input to the LDM, and a residual approach against a reference UNET is leveraged in applying the LDM. The performance of the generative LDM is compared with reference baselines of increasing complexity: quadratic interpolation of ERA5, a UNET, and a Generative Adversarial Network (GAN) built on the same reference UNET. Results highlight the improvements introduced by the LDM architecture and the residual approach over these baselines. The models are evaluated on a yearly test dataset, assessing the models' performance through deterministic metrics, spatial distribution of errors, and reconstruction of frequency and power spectra distributions., Comment: 24 pages, 14 figures

LiteBIRD, the next-generation cosmic microwave background (CMB) experiment, aims for a launch in Japan's fiscal year 2032, marking a major advancement in the exploration of primordial cosmology and fundamental physics. Orbiting the Sun-Earth Lagrangian point L2, this JAXA-led strategic L-class mission will conduct a comprehensive mapping of the CMB polarization across the entire sky. During its 3-year mission, LiteBIRD will employ three telescopes within 15 unique frequency bands (ranging from 34 through 448 GHz), targeting a sensitivity of 2.2\,$\mu$K-arcmin and a resolution of 0.5$^\circ$ at 100\,GHz. Its primary goal is to measure the tensor-to-scalar ratio $r$ with an uncertainty $\delta r = 0.001$, including systematic errors and margin. If $r \geq 0.01$, LiteBIRD expects to achieve a $>5\sigma$ detection in the $\ell=$2-10 and $\ell=$11-200 ranges separately, providing crucial insight into the early Universe. We describe LiteBIRD's scientific objectives, the application of systems engineering to mission requirements, the anticipated scientific impact, and the operations and scanning strategies vital to minimizing systematic effects. We will also highlight LiteBIRD's synergies with concurrent CMB projects., Comment: 23 pages, 9 figures, 1 table, SPIE Astronomical Telescopes + Instrumentation 2024

The formal representation of cultural metadata has always been a challenge, considering both the heterogeneity of cultural objects and the need to document the interpretive act exercised by experts. This article provides an overview of the revalorization of the digital collection Mythologiae in Linked Open Data format. The research aims to explore the data of a collection of artworks (Mythologiae) by promoting the potential of the Semantic Web, focusing particularly on the formal representation of the association of cultural objects with literary sources, as realized by experts, also documenting their interpretations. The workflow consisted of defining the data model, cleaning and disambiguating the data, converting it (from tabular structure to graph), and conducting testing activities (particularly expert domain review of the dataset through competency questions and data visualizations). The result is the mythLOD platform, which presents the dataset and detailed research documentation. Additionally, the platform hosts two data visualization spaces (the online catalogue and a data storytelling experiment on the case study of the Aeneid) that enrich the project documentation as user-friendly test units for the dataset and constitute an additional project documentation tool and exploration of the collection., Comment: in Italian language

We discuss the entanglement entropy for a massive Klein-Gordon field in two Schwarzschild-like quantum black hole spacetimes, also including a nonminimal coupling term with the background scalar curvature. To compute the entanglement entropy, we start from the standard spherical shell discretisation procedure, tracing over the degrees of freedom residing inside an imaginary surface. We estimate the free parameters for such quantum metrics through a simple physical argument based on Heisenberg uncertainty principle, along with alternative proposals as asymptotic safety, trace anomaly, and graviton corpuscular scaling. Our findings reveal a significant decrease in entropy compared to the area law near the origin for the quantum metrics. In both scenarios, the entanglement entropy converges to the expected area law sufficiently far from the origin. We then compare these results to the entropy scaling in regular Hayward and corrected-Hayward spacetimes to highlight the main differences with such regular approaches., Comment: 15 pages, 7 figures

In-context learning is a promising approach for online policy learning of offline reinforcement learning (RL) methods, which can be achieved at inference time without gradient optimization. However, this method is hindered by significant computational costs resulting from the gathering of large training trajectory sets and the need to train large Transformer models. We address this challenge by introducing an In-context Exploration-Exploitation (ICEE) algorithm, designed to optimize the efficiency of in-context policy learning. Unlike existing models, ICEE performs an exploration-exploitation trade-off at inference time within a Transformer model, without the need for explicit Bayesian inference. Consequently, ICEE can solve Bayesian optimization problems as efficiently as Gaussian process biased methods do, but in significantly less time. Through experiments in grid world environments, we demonstrate that ICEE can learn to solve new RL tasks using only tens of episodes, marking a substantial improvement over the hundreds of episodes needed by the previous in-context learning method., Comment: Published at ICLR 2024

We present an analysis of the recent near threshold BESIII data for the nucleon time-like effective form factors. The damped oscillation emerging from the subtraction of the dipole formula is treated in non-perturbative-QCD, making use of the light cone distribution amplitudes expansion. Non-perturbative effects are accounted for by considering Q2-dependent coefficients in such expansions, whose free parameters are determined by fitting to the proton and neutron data. Possible implications and future analysis have been discussed., Comment: 8 pages, 4 figures

Although binary classification is a well-studied problem in computer vision, training reliable classifiers under severe class imbalance remains a challenging problem. Recent work has proposed techniques that mitigate the effects of training under imbalance by modifying the loss functions or optimization methods. While this work has led to significant improvements in the overall accuracy in the multi-class case, we observe that slight changes in hyperparameter values of these methods can result in highly variable performance in terms of Receiver Operating Characteristic (ROC) curves on binary problems with severe imbalance. To reduce the sensitivity to hyperparameter choices and train more general models, we propose training over a family of loss functions, instead of a single loss function. We develop a method for applying Loss Conditional Training (LCT) to an imbalanced classification problem. Extensive experiment results, on both CIFAR and Kaggle competition datasets, show that our method improves model performance and is more robust to hyperparameter choices. Code is available at https://github.com/klieberman/roc_lct.

In this paper we describe the global on-sky calibration strategy of the LSPE-Strip instrument. Strip is a microwave telescope operating in the Q- and W-bands (central frequencies of 43 and 95 GHz respectively) from the Observatorio del Teide in Tenerife, with the goal to observe and characterise the polarised Galactic foreground emission, and complement the observations of the polarisation of the cosmic microwave background to be performed by the LSPE-SWIPE instrument and other similar experiments operating at higher frequencies to target the detection of the B-mode signal from the inflationary epoch of the Universe. Starting from basic assumptions on some of the instrument parameters (NET, 1/f noise knee frequency, beam properties, observing efficiency) we perform realistic simulations to study the level of accuracy that can be achieved through observations of bright celestial calibrators in the Strip footprint (sky fraction of 30 %) on the determination and characterisation of the main instrument parameters: global and relative gain factors (in intensity and in polarisation), polarisation direction, polarisation efficiency, leakage from intensity to polarisation, beams, window functions and pointing model., Comment: 46 pages, 20 figures. This paper is part of of the paper series "The LSPE/Strip instrument description and testing", to appear in Jinst. Comments are welcome

In this work, we investigate the fate of the non-Hermitian skin effect in one-dimensional systems that conserve the dipole moment and higher moments of an associated global $\text{U}(1)$ charge. Motivated by field theoretical arguments and lattice model calculations, we demonstrate that the key feature of the non-Hermitian skin effect for $m$-pole conserving systems is the generation of an $(m+1)$th multipole moment. For example, in contrast to the conventional skin effect where charges are anomalously localized at one boundary, the dipole-conserving skin effect results in charges localized at both boundaries, in a configuration that generates an extremal quadrupole moment. In addition, we explore the dynamical consequences of the $m$-pole skin effect, focusing on charge and entanglement propagation. Both numerically and analytically, we provide evidence that long-time steady-states have Fock-space localization and an area-law scaling of entanglement entropy, which serve as quantum indicators of the skin effect., Comment: 5+2 pages, 3+4 figures

We analyze time-dependent dark energy equations of state through linear and nonlinear structure formation and their quintessence potentials, characterized by fast, recent transitions, inspired by parameter space studies of selected classes of the more general Horndeski models. The influence of dark energy on structures comes from modifications to the background expansion rate and from perturbations as well. In order to compute the structures growth, we employ a generalization of the \emph{spherical collapse} formalism that includes perturbations of fluids with pressure. We numerically solve the equations of motion for the perturbations and the field. Our analysis suggests that a true Heaviside step transition is a good approximation for most of the considered models, since most of the quantities weakly depend on the transition speed. We find that transitions occurring at redshifts $z_{\rm t}\gtrsim 2$ cannot be distinguished from the $\Lambda$CDM model if dark energy is freezing, i.e, the corresponding equation of state tends to $-1$. For fast, recent transitions, the redshift at which the properties of dark energy have the most significant effect is $z=0.6\pm 0.2$. We also find that in the freezing regime, the $\sigma_8$ values can be lowered by about $8\%$, suggesting that those models could relieve the $\sigma_8$-tension. Additionally, freezing models generally predict faster late-time merging rates but a lower number of massive galaxies at $z=0$. Finally, the nonlinear matter power spectrum for smooth dark energy shows a valley centered in $k\approx1\,h\,{\rm Mpc}^{-1}$ which in the clustering case is replaced by a sharp increase for $k\gtrsim 0.2\,h\,{\rm Mpc}^{-1}$ and a peak at $k\approx 2\,h\,{\rm Mpc}^{-1}$., Comment: 26 pages, 32 figures. Accepted for publication in PRD

The nonresonant mechanism in the reactions $e^+\,e^-\to \pi^-\,p\,\bar{n}$ and $e^+\,e^-\to \pi^+\,n\,\bar{p}$ is investigated in frame of the one-photon exchange approximation. The description of the hadronic phase space and the invariant amplitude formalism, earlier developed for the neutral pion channel, are applied here. Some modifications are required to fulfill the invariance and to account for an additional contribution due to diagram with the pion pole. Two different variants which lead to dipole and monopole asymptotic behaviour of the pion electromagnetic form factor, are considered. The double and single distributions over the invariant variables as well the total cross section, are obtained. The results are plotted for $s=$5,\,6,\,10,\,16\, GeV$^2$. The calculations are performed for the $\pi^-\,p\,\bar{n}$ channel and the rules to proceed to the $\pi^+\,n\,\bar{p}$ channel are formulated for every distribution., Comment: 30 pages, 9 figures

The Q&U Bolometric Interferometer for Cosmology (QUBIC) is the first bolometric interferometer designed to measure the primordial B-mode polarization of the Cosmic Microwave Background (CMB). Bolometric interferometry is a novel technique that combines the sensitivity of bolometric detectors with the control of systematic effects that is typical of interferometry, both key features in the quest for the faint signal of the primordial B-modes. A unique feature is the so-called "spectral imaging", i.e., the ability to recover the sky signal in several sub-bands within the physical band during data analysis. This feature provides an in-band spectral resolution of \Delta{\nu}/{\nu} \sim 0.04 that is unattainable by a traditional imager. This is a key tool for controlling the Galactic foregrounds contamination. In this paper, we describe the principles of bolometric interferometry, the current status of the QUBIC experiment and future prospects., Comment: To appear in Proc. of the mm Universe 2023 conference, Grenoble (France), June 2023, published by F. Mayet et al. (Eds), EPJ Web of conferences, EDP Sciences

The differential cross section and polarization observables for the elastic reaction induced by deuteron scattering off electrons at rest are calculated in the one-photon-exchange (Born) approximation. Specific attention is given to the kinematical conditions, that is, to the specific range of incident energy and transferred momentum. The specific interest of this reaction is to access very small transferred momenta. Numerical estimates are given for polarization observables that describe the of single- and double-spin effects, provided that the polarization components (both, vector and tensor) of each particle in the reaction are determined in the rest frame of the electron target., Comment: 39 pages 11 figures, AddedvAcknowledgment section

In this work, we investigate the localization properties of a paradigmatic model, coupled to a monitoring environment and possessing a many-body localized (MBL) phase. We focus on the post-selected no-click limit with quench random rates, i.e., random gains and losses. In this limit, the system is modeled by adding an imaginary random potential, rendering non-Hermiticity in the system. Numerically, we provide an evidence that the system is localized for any finite amount of disorder. To analytically understand our results, we extend the quantum random energy model (QREM) to the non-Hermitian scenario. The Hermitian QREM has been used previously as a benchmark model for MBL. The QREM exhibits a size-dependent MBL transition, where the critical value scales as $W_c\sim \sqrt{L} \ln{L}$ with system size and presenting many-body mobility edges. We reveal that the non-Hermitian QREM with random gain-loss offers a significantly stronger form of localization, evident in the nature of the many-body mobility edges and the value for the transition, which scales as $W_c\sim \ln^{1/2}{L}$ with the system size., Comment: 11 pages, 4 figures, 94 references

Personalization of playlists is a common feature in music streaming services, but conventional techniques, such as collaborative filtering, rely on explicit assumptions regarding content quality to learn how to make recommendations. Such assumptions often result in misalignment between offline model objectives and online user satisfaction metrics. In this paper, we present a reinforcement learning framework that solves for such limitations by directly optimizing for user satisfaction metrics via the use of a simulated playlist-generation environment. Using this simulator we develop and train a modified Deep Q-Network, the action head DQN (AH-DQN), in a manner that addresses the challenges imposed by the large state and action space of our RL formulation. The resulting policy is capable of making recommendations from large and dynamic sets of candidate items with the expectation of maximizing consumption metrics. We analyze and evaluate agents offline via simulations that use environment models trained on both public and proprietary streaming datasets. We show how these agents lead to better user-satisfaction metrics compared to baseline methods during online A/B tests. Finally, we demonstrate that performance assessments produced from our simulator are strongly correlated with observed online metric results., Comment: 10 pages. KDD 23

Both optical flow and stereo disparities are image matches and can therefore benefit from joint training. Depth and 3D motion provide geometric rather than photometric information and can further improve optical flow. Accordingly, we design a first network that estimates flow and disparity jointly and is trained without supervision. A second network, trained with optical flow from the first as pseudo-labels, takes disparities from the first network, estimates 3D rigid motion at every pixel, and reconstructs optical flow again. A final stage fuses the outputs from the two networks. In contrast with previous methods that only consider camera motion, our method also estimates the rigid motions of dynamic objects, which are of key interest in applications. This leads to better optical flow with visibly more detailed occlusions and object boundaries as a result. Our unsupervised pipeline achieves 7.36% optical flow error on the KITTI-2015 benchmark and outperforms the previous state-of-the-art 9.38% by a wide margin. It also achieves slightly better or comparable stereo depth results. Code will be made available.