Search

Generative Flow Networks (GFlowNets), a class of generative models over discrete and structured sample spaces, have been previously applied to the problem of inferring the marginal posterior distribution over the directed acyclic graph (DAG) of a Bayesian Network, given a dataset of observations. Based on recent advances extending this framework to non-discrete sample spaces, we propose in this paper to approximate the joint posterior over not only the structure of a Bayesian Network, but also the parameters of its conditional probability distributions. We use a single GFlowNet whose sampling policy follows a two-phase process: the DAG is first generated sequentially one edge at a time, and then the corresponding parameters are picked once the full structure is known. Since the parameters are included in the posterior distribution, this leaves more flexibility for the local probability models of the Bayesian Network, making our approach applicable even to non-linear models parametrized by neural networks. We show that our method, called JSP-GFN, offers an accurate approximation of the joint posterior, while comparing favorably against existing methods on both simulated and real data.

In end-to-end speech translation, speech and text pre-trained models improve translation quality. Recently proposed models simply connect the pre-trained models of speech and text as encoder and decoder. Therefore, only the information from the final layer of encoders is input to the decoder. Since it is clear that the speech pre-trained model outputs different information from each layer, the simple connection method cannot fully utilize the information that the speech pre-trained model has. In this study, we propose an inter-connection mechanism that aggregates the information from each layer of the speech pre-trained model by weighted sums and inputs into the decoder. This mechanism increased BLEU by approximately 2 points in en-de, en-ja, and en-zh by increasing parameters by 2K when the speech pre-trained model was frozen. Furthermore, we investigated the contribution of each layer for each language by visualizing layer weights and found that the contributions were different., Comment: Accepted at INTERSPEECH2023

Individual vibrational band spectroscopy presents an opportunity to examine exoplanet atmospheres in detail by distinguishing where the vibrational state populations of molecules differ from the current assumption of a Boltzmann distribution. Here, retrieving vibrational bands of OH in exoplanet atmospheres is explored using the hot Jupiter WASP-33b as an example. We simulate low-resolution spectroscopic data for observations with the JWST's NIRSpec instrument and use high resolution observational data obtained from the Subaru InfraRed Doppler instrument (IRD). Vibrational band-specific OH cross section sets are constructed and used in retrievals on the (simulated) low and (real) high resolution data. Low resolution observations are simulated for two WASP-33b emission scenarios: under the assumption of local thermal equilibrium (LTE) and a toy non-LTE model for vibrational excitation of selected bands. We show that mixing ratios for individual bands can be retrieved with sufficient precision to allow the vibrational population distributions of the forward models to be reconstructed. A simple fit for the Boltzmann distribution in the LTE case shows that the vibrational temperature is recoverable in this manner. For high resolution, cross-correlation applications, we apply the individual vibrational band analysis to an IRD spectrum of WASP-33b, applying an 'un-peeling' technique. Individual detection significances for the two strongest bands are shown to be in line with Boltzmann distributed vibrational state populations consistent with the effective temperature of the WASP-33b atmosphere reported previously. We show the viability of this approach for analysing the individual vibrational state populations behind observed and simulated spectra including reconstructing state population distributions., Comment: Submitted for publication in AJ

In this short note, we discuss the circumstances that can lead to a failure to observe the design order of discretization error convergence in accuracy verification when solving a time-dependent problem. In particular, we discuss the problem of failing to observe the design order of spatial accuracy with an extremely small time step. The same problem is encountered even if the time step is reduced with grid refinement. These can cause a serious problem because then one would wind up trying to find a coding error that does not exist. This short note clarifies the mechanism causing this failure and provides a guide for avoiding such pitfalls

In this paper, we propose an efficient quadratic interpolation formula utilizing solution gradients computed and stored at nodes and demonstrate its application to a third-order cell-centered finite-volume discretization on tetrahedral grids. The proposed quadratic formula is constructed based on an efficient formula of computing a projected derivative. It is efficient in that it completely eliminates the need to compute and store second derivatives of solution variables or any other quantities, which are typically required in upgrading a second-order cell-centered unstructured-grid finite-volume discretization to third-order accuracy. Moreover, a high-order flux quadrature formula, as required for third-order accuracy, can also be simplified by utilizing the efficient projected-derivative formula, resulting in a numerical flux at a face centroid plus a curvature correction not involving second derivatives of the flux. Similarly, a source term can be integrated over a cell to high-order in the form of the source term evaluated at the cell centroid plus a curvature correction, again, not requiring second derivatives of the source term. The discretization is defined as an approximation to an integral form of a conservation law but the numerical solution is defined as a point value at a cell center, leading to another feature that there is no need to compute and store geometric moments for a quadratic polynomial to preserve a cell average. Third-order accuracy and improved second-order accuracy are demonstrated and investigated for simple but illustrative test cases in three dimensions.

We report the mass determination of TOI-519 b, a transiting substellar object around a mid-M dwarf. We carried out radial velocity measurements using Subaru / InfraRed Doppler (IRD), revealing that TOI-519 b is a planet with a mass of $0.463^{+0.082}_{-0.088}~M_{\rm Jup}$. We also find that the host star is metal rich ($\rm [Fe/H] = 0.27 \pm 0.09$ dex) and has the lowest effective temperature ($T_{\rm eff}=3322 \pm 49$ K) among all stars hosting known close-in giant planets based on the IRD spectra and mid-resolution infrared spectra obtained with NASA Infrared Telescope Facility / SpeX. The core mass of TOI-519 b inferred from a thermal evolution model ranges from $0$ to $\sim30~M_\oplus$, which can be explained by both the core accretion and disk instability models as the formation origins of this planet. However, TOI-519 is in line with the emerging trend that M dwarfs with close-in giant planets tend to have high metallicity, which may indicate that they formed in the core accretion model. The system is also consistent with the potential trend that close-in giant planets around M dwarfs tend to be less massive than those around FGK dwarfs., Comment: 10 pages, 5 figures. Accepted for publication in PASJ

We study the liquid--hexatic transition of soft disks with massively parallel simulations and determine the equation of state as a function of system size. For systems with interactions decaying as the inverse $m$th power of the separation, the liquid--hexatic phase transition is continuous for $m = 12$ and $m=8$, while it is of first order for $m = 24$. The critical power $m$ for the transition between continuous and first-order behavior is larger than previously reported. The continuous transition for $ m=12 $ implies that the two-dimensional Lennard-Jones model has a continuous liquid--hexatic transition at high temperatures. We also study the Weeks--Chandler--Andersen model and find a continuous transition at high temperatures, that is consistent with the soft-disk case for $m=12$. Pressure data as well as our implementation are available from an open-source repository., Comment: 7 pages, 4 figures

The predominant source of nonlinearity in most existing accelerators are sextupoles, which introduce nonlinear resonances. When the horizontal tune of an accelerator is near such a resonance line ($n\nu_x$), stable fixed points (SFPs) may appear in the horizontal phase space to form a second closed orbit different from the "zero" closed orbit. The stable islands in phase space surrounding the SFPs are also referred as transverse resonance island "buckets" (TRIBs). However, the TRIBs are not always present near a resonance line in a storage ring. Necessary conditions for TRIBs formation will be discussed in this paper. A 6-GeV lattice with the horizontal tune near a 3rd-order resonance line at $3\nu_{x}$=50 is then designed for the Cornell Electron Storage Ring (CESR). After loading this lattice into CESR, the TRIBs are observed while adjusting the horizontal tune near the 3rd-order resonance, consistent with particle tracking simulations. TRIBs rotation in the horizontal phase space are also experimentally demonstrated with adjustment of a sextupole knob. These results are in good agreement with the theoretical calculation., Comment: 15 pages, 15 figures

We report on an Er-doped fiber (EDF)-laser-based dual-comb system that allows us to perform triggerless asynchronous optical sampling (ASOPS) pump-probe measurements of ultrafast demagnetization and spin precession in magnetic materials. Because the oscillation frequencies of the two frequency-comb light sources are highly stabilized, the pulse-to-pulse timing jitter is sufficiently suppressed and data accumulation without any trigger signals is possible. To effectively induce spin precession in ferromagnetic thin films, the spectral bandwidth of the output of one of the EDF frequency comb sources is broadened by a highly nonlinear fiber and then amplified at a wavelength of about 1030 nm by a Yb-doped fiber amplifier. The output of the other frequency comb source is converted to about 775 nm by second harmonic generation. We used this system to observe the ultrafast demagnetization and spin precession dynamics on the picosecond and nanosecond time scales in a permalloy thin film. This time-domain spectroscopy system is promising for the rapid characterization of spin-wave generation and propagation dynamics in magnetic materials., Comment: 20 pages, 8 figures

In this paper, we Fourier transform the Wightman function concerning energy and angular momentum on the $S^{D-1}$ spatial slice in radial quantization in $D=2,3$ dimensions. In each case, we use the conformal Ward Identities to solve systematically for the Fourier components. We then use these Fourier components to build conformal blocks for the four-point function in momentum space, giving a finite-volume version of the momentum-space conformal blocks. We check that this construction is consistent with the known result in infinite volume. Our construction may help to find bootstrap equations that can give nontrivial constraints that do not appear in analysis in infinite volume. We show some examples of bootstrap equations and their nontriviality., Comment: 83 pages, 14 figures

Robot manipulation tasks by natural language instructions need common understanding of the target object between human and the robot. However, the instructions often have an interpretation ambiguity, because the instruction lacks important information, or does not express the target object correctly to complete the task. To solve this ambiguity problem, we hypothesize that "naming" the target objects in advance will reduce the ambiguity of natural language instructions. We propose a robot system and method that incorporates naming with appearance of the objects in advance, so that in the later manipulation task, instruction can be performed with its unique name to disambiguate the objects easily. To demonstrate the effectiveness of our approach, we build a system that can memorize the target objects, and show that naming the objects facilitates detection of the target objects and improves the success rate of manipulation instructions. With this method, the success rate of object manipulation task increases by 31% in ambiguous instructions., Comment: 8 pages, 7 figures

The popularity of cryptocurrencies has grown significantly in recent years, and they have become an important asset for internet trading. One of the main drawbacks of cryptocurrencies is the high volatility and fluctuation in value. The value of cryptocurrencies can change rapidly and dramatically, making them a risky investment. Cryptocurrencies are largely unregulated, which can exacerbate their volatility. The high volatility of cryptocurrencies has also led to a speculative bubble, with many investors buying and selling cryptocurrencies based on short-term price fluctuations rather than their underlying values. Therefore, how to reduce the fluctuation risk introduced by exchanges, transform uncertain prices to deterministic value, and promote the benefits of decentralized finance are critical for the future development of cryptos and Web 3.0. To address the issues, this paper proposes a novel theory as Automatic Increase Market Systems (AIMS) for cryptos, which could potentially be designed to automatically adjust the value of a cryptocurrency helping to stabilize the price and increase its value over time in a deterministic manner. We build a crypto, WISH (https://wishbank.wtf), based on AIMS in order to demonstrate how the automatic increase market system would work in practice, and how it would influence the supply of the cryptocurrency in response to market demand and finally make itself to be a stable medium of exchange, ensuring that the AIMS is fair and transparent.