Search

Multiple studies have investigated potential frequency-dependent dispersion measures (DM) in PSR B0329+54, with sensitivities at levels of $10^{-3} \, \text{pc} \, \text{cm}^{-3}$ or higher, using frequencies below 1 GHz. Utilizing the extensive bandwidth of the upgraded Giant Meterwave Radio Telescope, we conducted simultaneous observations of this pulsar across a frequency range of 300 to 1460 MHz. Our observations reveal a distinct point in the pulse profile of PSR B0329+54 that appears to align remarkably well with the cold-plasma dispersion law, resulting in a unique measured DM across the entire frequency range. In contrast, using times of arrival (ToAs) from widely adopted pulsar timing techniques (e.g., FFTFIT)-leads to frequency-dependent DMs. We investigated the potential causes of these frequency-dependent DMs in this pulsar and their relationship with the underlying magnetic field geometry corresponding to the radio emission. Our study reveals that all frequencies in the range 300-1460 MHz originate from a region no larger than 204 km, and the dipolar magnetic-field geometry model indicates that the emission region is centered at $\sim$800 km from the star. This is the tightest constraint on the size of the emission region reported so far for PSR B0329+54 at the given frequencies, and it is at least five times more stringent than the existing emission height constraints based on the dipolar geometry model., Comment: Accepted in the Astrophysical Journal (APJ) on 03-Jan-2025

The production mechanism of fast radio bursts (FRBs) remains elusive, and potential correlations between burst occurrence times and various burst properties may offer important clues. Among them, the spectral peak frequency is particularly important because it may encode direct information about the physical conditions and environment at the emission site. Analyzing over 4,000 bursts from the three most active sources -- FRB 20121102A, FRB 20201124A, and FRB 20220912A -- we measure the two-point correlation function $\xi(\Delta t, \Delta\nu_{\:\rm peak}\:)$ in the two-dimensional space of time separation $\Delta t$ and peak frequency shift $\Delta\nu_{\:\rm peak}\:$ between burst pairs. We find a universal trend of asymmetry about $\Delta\nu_{\:\rm peak}\:$ at high statistical significance; $\xi(\Delta\nu_{\:\rm peak}\:)$ decreases as $\Delta\nu_{\:\rm peak}\:$ increases from negative to positive values in the region of short time separation ($\Delta t < 0.3$ s), where physically correlated aftershock events produce a strong time correlation signal. This indicates that aftershocks tend to exhibit systematically lower peak frequencies than mainshocks, with this tendency becoming stronger at shorter $\Delta t$. We argue that the "sad trombone effect" -- the downward frequency drift observed among sub-pulses within a single event -- is not confined within a single event but manifests as a statistical nature that extends continuously to independent yet physically correlated aftershocks with time separations up to $\Delta t \sim 0.3$ s. This discovery provides new insights into underlying physical processes of repeater FRBs., Comment: 13 pages, 5 figures, submitted to ApJL, comments welcome

NinjaSat is an X-ray CubeSat designed for agile, long-term continuous observations of bright X-ray sources, with the size of 6U ($100\times200\times300$ mm$^3$) and a mass of 8 kg. NinjaSat is capable of pointing at X-ray sources with an accuracy of less than $0^{\circ}\hspace{-1.0mm}.1$ (2$\sigma$ confidence level) with 3-axis attitude control. The satellite bus is a commercially available NanoAvionics M6P, equipped with two non-imaging gas X-ray detectors covering an energy range of 2-50 keV. A total effective area of 32 cm$^2$ at 6 keV is capable of observing X-ray sources with a flux of approximately 10$^{-10}$ erg cm$^{-2}$ s$^{-1}$. The arrival time of each photon can be tagged with a time resolution of 61 $\mu$s. The two radiation belt monitors continuously measure the fluxes of protons above 5 MeV and electrons above 200 keV trapped in the geomagnetic field, alerting the X-ray detectors when the flux exceeds a threshold. The NinjaSat project started in 2020. Fabrication of the scientific payloads was completed in August 2022, and satellite integration and tests were completed in July 2023. NinjaSat was launched into a Sun-synchronous polar orbit at an altitude of about 530 km on 2023 November 11 by the SpaceX Transporter-9 mission. After about three months of satellite commissioning and payload verification, we observed the Crab Nebula on February 9, 2024, and successfully detected the 33.8262 ms pulsation from the neutron star. With this observation, NinjaSat met the minimum success criterion and stepped forward to scientific observations as initially planned. By the end of November 2024, we successfully observed 21 X-ray sources using NinjaSat. This achievement demonstrates that, with careful target selection, we can conduct scientific observations effectively using CubeSats, contributing to time-domain astronomy., Comment: 14 pages, 17 figures

Semi-supervised learning (SSL) can improve model performance by leveraging unlabeled images, which can be collected from public image sources with low costs. In recent years, synthetic images have become increasingly common in public image sources due to rapid advances in generative models. Therefore, it is becoming inevitable to include existing synthetic images in the unlabeled data for SSL. How this kind of contamination will affect SSL remains unexplored. In this paper, we introduce a new task, Real-Synthetic Hybrid SSL (RS-SSL), to investigate the impact of unlabeled data contaminated by synthetic images for SSL. First, we set up a new RS-SSL benchmark to evaluate current SSL methods and found they struggled to improve by unlabeled synthetic images, sometimes even negatively affected. To this end, we propose RSMatch, a novel SSL method specifically designed to handle the challenges of RS-SSL. RSMatch effectively identifies unlabeled synthetic data and further utilizes them for improvement. Extensive experimental results show that RSMatch can transfer synthetic unlabeled data from `obstacles' to `resources.' The effectiveness is further verified through ablation studies and visualization.

The recent success of denoising diffusion models has significantly advanced text-to-image generation. While these large-scale pretrained models show excellent performance in general image synthesis, downstream objectives often require fine-tuning to meet specific criteria such as aesthetics or human preference. Reward gradient-based strategies are promising in this context, yet existing methods are limited to single-reward tasks, restricting their applicability in real-world scenarios that demand adapting to multiple objectives introduced incrementally over time. In this paper, we first define this more realistic and unexplored problem, termed Reward Incremental Learning (RIL), where models are desired to adapt to multiple downstream objectives incrementally. Additionally, while the models adapt to the ever-emerging new objectives, we observe a unique form of catastrophic forgetting in diffusion model fine-tuning, affecting both metric-wise and visual structure-wise image quality. To address this catastrophic forgetting challenge, we propose Reward Incremental Distillation (RID), a method that mitigates forgetting with minimal computational overhead, enabling stable performance across sequential reward tasks. The experimental results demonstrate the efficacy of RID in achieving consistent, high-quality generation in RIL scenarios. The source code of our work will be publicly available upon acceptance., Comment: Under review

Image generation has shown remarkable results in generating high-fidelity realistic images, in particular with the advancement of diffusion-based models. However, the prevalence of AI-generated images may have side effects for the machine learning community that are not clearly identified. Meanwhile, the success of deep learning in computer vision is driven by the massive dataset collected on the Internet. The extensive quantity of synthetic data being added to the Internet would become an obstacle for future researchers to collect "clean" datasets without AI-generated content. Prior research has shown that using datasets contaminated by synthetic images may result in performance degradation when used for training. In this paper, we investigate the potential impact of contaminated datasets on Online Continual Learning (CL) research. We experimentally show that contaminated datasets might hinder the training of existing online CL methods. Also, we propose Entropy Selection with Real-synthetic similarity Maximization (ESRM), a method to alleviate the performance deterioration caused by synthetic images when training online CL models. Experiments show that our method can significantly alleviate performance deterioration, especially when the contamination is severe. For reproducibility, the source code of our work is available at https://github.com/maorong-wang/ESRM., Comment: Accepted to NeurIPS'24

In February and March 2024, a series of many Type I X-ray bursts from the accreting neutron star SRGA J144459.2$-$604207, which has been identified by multiple X-ray satellites, with the first reports coming from INTEGRAL and NinjaSat. These observations reveal that after exhibiting very regular behavior as a ``clocked'' burster, the peak luminosity of the SRGA J144459.2$-$604207 X-ray bursts shows a gradual decline. The observed light curves exhibit a short plateau feature, potentially with a double peak, followed by a rapid decay in the tail-features unlike those seen in previously observed clocked bursters. In this study, we calculate a series of multizone X-ray burst models with various compositions of accreted matter, specifically varying the mass fractions of hydrogen ($X$), helium ($Y$), and heavier CNO elements or metallicity ($Z_{\rm CNO}$). We demonstrate that a model with higher $Z_{\rm CNO}$ and/or lower $X/Y$ compared to the solar values can reproduce the observed behavior of SRGA J144459.2$-$604207. Therefore, we propose that this new X-ray burster is likely the first clocked burster with non-solar elemental compositions. Moreover, based on the X-ray burst light curve morphology in the decline phase observed by NinjaSat, a He-enhanced model with $X/Y \approx 1.5$ seems preferred over high-metallicity cases. We also give a brief discussion on the implications for the neutron star mass, binary star evolution, inclination angle, and the potential for a high-metallicity scenario, the last of which is closely related to the properties of the hot CNO cycle., Comment: 7 pages, 5 figures, accepted for publication in PASJ Letter

The CubeSat X-ray observatory NinjaSat was launched on 2023 November 11 and has provided opportunities for agile and flexible monitoring of bright X-ray sources. On 2024 February 23, the NinjaSat team started long-term observation of the new X-ray source SRGA J144459.2$-$604207 as the first scientific target, which was discovered on 2024 February 21 and recognized as the sixth clocked X-ray burster. Our 25-day observation covered almost the entire decay of this outburst from two days after the peak at $\sim$100 mCrab on February 23 until March 18 at a few mCrab level. The Gas Multiplier Counter onboard NinjaSat successfully detected 12 Type-I X-ray bursts with a typical burst duration of $\sim$20 s, shorter than other clocked burster systems. As the persistent X-ray emission declined by a factor of five, X-ray bursts showed a notable change in its morphology: the rise time became shorter from 4.4(7) s to 0.3(3) s (1$\sigma$ errors), and the peak amplitude increased by 44%. The burst recurrence time $\Delta t_{\rm rec}$ also became longer from 2 hr to 10 hr, following the relation of $\Delta t_{\rm rec} \propto F_{\rm per}^{-0.84}$, where $F_{\rm per}$ is the persistent X-ray flux, by applying a Markov chain Monte Carlo method. The short duration of bursts is explained by the He-enhanced composition of accretion matter and the relation between $\Delta t_{\textrm{rec}}$ and $F_{\rm per}$ by a massive neutron star. This study demonstrated that CubeSat pointing observations can provide valuable astronomical X-ray data., Comment: 7 pages, 5 figures, submitted to PASJ Letter

The Privacy Preserving Federated Learning Document VQA (PFL-DocVQA) competition challenged the community to develop provably private and communication-efficient solutions in a federated setting for a real-life use case: invoice processing. The competition introduced a dataset of real invoice documents, along with associated questions and answers requiring information extraction and reasoning over the document images. Thereby, it brings together researchers and expertise from the document analysis, privacy, and federated learning communities. Participants fine-tuned a pre-trained, state-of-the-art Document Visual Question Answering model provided by the organizers for this new domain, mimicking a typical federated invoice processing setup. The base model is a multi-modal generative language model, and sensitive information could be exposed through either the visual or textual input modality. Participants proposed elegant solutions to reduce communication costs while maintaining a minimum utility threshold in track 1 and to protect all information from each document provider using differential privacy in track 2. The competition served as a new testbed for developing and testing private federated learning methods, simultaneously raising awareness about privacy within the document image analysis and recognition community. Ultimately, the competition analysis provides best practices and recommendations for successfully running privacy-focused federated learning challenges in the future., Comment: 27 pages, 6 figures

A major challenge in fault-tolerant quantum computation (FTQC) is to reduce both space overhead -- the large number of physical qubits per logical qubit -- and time overhead -- the long physical gate sequences per logical gate. We prove that a protocol using non-vanishing-rate quantum low-density parity-check (LDPC) codes, combined with concatenated Steane codes, achieves constant space overhead and polylogarithmic time overhead, even when accounting for non-zero classical computation time. This protocol offers an improvement over existing constant-space-overhead protocols, which have polynomial time overhead using quantum LDPC codes and quasi-polylogarithmic time overhead using concatenated quantum Hamming codes. To ensure the completeness of this proof, we develop a technique called partial circuit reduction, which enables error analysis for the entire fault-tolerant circuit by examining smaller parts composed of a few gadgets. With this technique, we resolve a previously unaddressed logical gap in the existing arguments and complete the proof of the threshold theorem for the constant-space-overhead protocol with quantum LDPC codes. Our work highlights that the quantum-LDPC-code approach can realize FTQC with a negligibly small slowdown and a bounded overhead of physical qubits, similar to the code-concatenation approach, underscoring the importance of a comprehensive comparison of the future realizability of these two approaches., Comment: 63 pages, 19 figures

Adversarial examples have raised several open questions, such as why they can deceive classifiers and transfer between different models. A prevailing hypothesis to explain these phenomena suggests that adversarial perturbations appear as random noise but contain class-specific features. This hypothesis is supported by the success of perturbation learning, where classifiers trained solely on adversarial examples and the corresponding incorrect labels generalize well to correctly labeled test data. Although this hypothesis and perturbation learning are effective in explaining intriguing properties of adversarial examples, their solid theoretical foundation is limited. In this study, we theoretically explain the counterintuitive success of perturbation learning. We assume wide two-layer networks and the results hold for any data distribution. We prove that adversarial perturbations contain sufficient class-specific features for networks to generalize from them. Moreover, the predictions of classifiers trained on mislabeled adversarial examples coincide with those of classifiers trained on correctly labeled clean samples. The code is available at https://github.com/s-kumano/perturbation-learning., Comment: NeurIPS24

Electromagnetic interference (EMI) filters are used to reduce electromagnetic noise. It is well known that the performance of an EMI filter in reducing electromagnetic noise largely depends on its conductor layout. Therefore, if a conductor layout optimization method with a high degree of freedom is realized, a drastic performance improvement is expected. Although there are a few design methods based on topology optimization for this purpose, these methods have some difficulties originating from topology optimization. In this paper, we therefore propose a conductor layout design method for EMI filters on the basis of data-driven topology design (DDTD), which is a high degree of freedom structural design methodology incorporating a deep generative model and data-driven approach. DDTD was proposed to overcome the intrinsic difficulties of topology optimization, and we consider it suitable for the conductor layout design problem of EMI filters. One significant challenge in applying DDTD to the conductor layout design problem is maintaining the topology of the circuit diagram during the solution search. For this purpose, we propose a simple yet efficient constraint. We further provide numerical examples to confirm the usefulness of the proposed method., Comment: 9 pages, 16 figures

A cognitive function of tracking multiple objects, needed in autonomous mobile vehicles, comprises object detection and their temporal association. While great progress owing to machine learning has been recently seen for elaborating the similarity matrix between the objects that have been recognized and the objects detected in a current video frame, less for the assignment problem that finally determines the temporal association, which is a combinatorial optimization problem. Here we show an in-vehicle multiple object tracking system with a flexible assignment function for tracking through multiple long-term occlusion events. To solve the flexible assignment problem formulated as a nondeterministic polynomial time-hard problem, the system relies on an embeddable Ising machine based on a quantum-inspired algorithm called simulated bifurcation. Using a vehicle-mountable computing platform, we demonstrate a realtime system-wide throughput (23 frames per second on average) with the enhanced functionality., Comment: 18 pages, 7 figures, 2 tables

The quantum error-correcting code in the continuous-variable (CV) system attracts much attention due to its flexibility and high resistance against specific noise. However, the theory of fault tolerance in CV systems is premature and lacks the general strategy to translate the noise in CV systems into the noise in logical qubits, leading to severe restrictions on the correctable noise models. In this paper, we show that the Markovian-type noise in CV systems is translated into the Markovian-type noise in the logical qubits through the Gottesman-Kitaev-Preskill code with an explicit bound on the noise strength. Combined with the established threshold theorem of the concatenated code against Markovian-type noise, we show that CV quantum computation has a fault-tolerant threshold against general Markovian-type noise, closing the existing crucial gap in CV quantum computation. We also give a new insight into the fact that careful management of the energy of the state is required to achieve fault tolerance in the CV system., Comment: 12 pages with 4 figures for main text, 34 pages with 1 figure for supplementary information

Open-set semi-supervised learning (OSSL) leverages practical open-set unlabeled data, comprising both in-distribution (ID) samples from seen classes and out-of-distribution (OOD) samples from unseen classes, for semi-supervised learning (SSL). Prior OSSL methods initially learned the decision boundary between ID and OOD with labeled ID data, subsequently employing self-training to refine this boundary. These methods, however, suffer from the tendency to overtrust the labeled ID data: the scarcity of labeled data caused the distribution bias between the labeled samples and the entire ID data, which misleads the decision boundary to overfit. The subsequent self-training process, based on the overfitted result, fails to rectify this problem. In this paper, we address the overtrusting issue by treating OOD samples as an additional class, forming a new SSL process. Specifically, we propose SCOMatch, a novel OSSL method that 1) selects reliable OOD samples as new labeled data with an OOD memory queue and a corresponding update strategy and 2) integrates the new SSL process into the original task through our Simultaneous Close-set and Open-set self-training. SCOMatch refines the decision boundary of ID and OOD classes across the entire dataset, thereby leading to improved results. Extensive experimental results show that SCOMatch significantly outperforms the state-of-the-art methods on various benchmarks. The effectiveness is further verified through ablation studies and visualization., Comment: ECCV 2024 accepted

We have conducted Q-band (30 GHz $-$ 50 GHz) line survey observations toward a carbon-chain emission peak in the Serpens South cluster-forming region with the extended Q-band (eQ) receiver installed on the Nobeyama 45 m radio telescope. Approximately 180 lines have been detected including tentative detection, and these lines are attributed to 52 molecules including isotopologues. It has been found that this position is rich in carbon-chain species as much as Cyanopolyyne Peak in Taurus Molecular Cloud-1 (TMC-1 CP), suggesting chemical youth. Not only carbon-chain species, but several complex organic molecules (CH$_3$OH, CH$_3$CHO, HCCCHO, CH$_3$CN, and tentatively C$_2$H$_3$CN) have also been detected, which is similar to the chemical complexity found in evolved prestellar cores. The HDCS/H$_2$CS ratio has been derived to be $11.3 \pm 0.5$ %, and this value is similar to the prestellar core L1544. The chemically young features that are similar to the less-dense starless core TMC-1 CP ($10^4$ cm$^{-3}$ $-$ $10^5$ cm$^{-3}$) and chemically evolved characters which resemble the dense prestellar core L1544 ($\sim 10^6$ cm$^{-3}$) mean that the clump including the observed position is a pre-cluster clump without any current star formation activity., Comment: Accepted for publication in Publications of the Astronomical Society of Japan (PASJ), 32 pages, 9 figures, 6 tables

Topology optimization is a structural design methodology widely utilized to address engineering challenges. However, sensitivity-based topology optimization methods struggle to solve optimization problems characterized by strong non-linearity. Leveraging the sensitivity-free nature and high capacity of deep generative models, data-driven topology design (DDTD) methodology is considered an effective solution to this problem. Despite this, the training effectiveness of deep generative models diminishes when input size exceeds a threshold while maintaining high degrees of freedom is crucial for accurately characterizing complex structures. To resolve the conflict between the both, we propose DDTD based on principal component analysis (PCA). Its core idea is to replace the direct training of deep generative models with material distributions by using a principal component score matrix obtained from PCA computation and to obtain the generated material distributions with new features through the restoration process. We apply the proposed PCA-based DDTD to the problem of minimizing the maximum stress in 3D structural mechanics and demonstrate it can effectively address the current challenges faced by DDTD that fail to handle 3D structural design problems. Various experiments are conducted to demonstrate the effectiveness and practicability of the proposed PCA-based DDTD., Comment: 19 pages, 18 figures

Magic state distillation is a crucial yet resource-intensive process in fault-tolerant quantum computation. The protocol's overhead, defined as the number of input magic states required per output magic state with an error rate below $\epsilon$, typically grows as $\mathcal{O}(\log^\gamma(1/\epsilon))$. Achieving smaller overheads, i.e., smaller exponents $\gamma$, is highly desirable; however, all existing protocols require polylogarithmically growing overheads with some $\gamma > 0$, and identifying the smallest achievable exponent $\gamma$ for distilling magic states of qubits has remained challenging. To address this issue, we develop magic state distillation protocols for qubits with efficient, polynomial-time decoding that achieve an $\mathcal{O}(1)$ overhead, meaning the optimal exponent $\gamma = 0$; this improves over the previous best of $\gamma \approx 0.678$ due to Hastings and Haah. In our construction, we employ algebraic geometry codes to explicitly present asymptotically good quantum codes for $2^{10}$-dimensional qudits that support transversally implementable logical gates in the third level of the Clifford hierarchy. The use of asymptotically good codes with non-vanishing rate and relative distance leads to the constant overhead. These codes can be realised by representing each $2^{10}$-dimensional qudit as a set of $10$ qubits, using stabiliser operations on qubits. The $10$-qubit magic states distilled with these codes can be converted to and from conventional magic states for the controlled-controlled-$Z$ ($CCZ$) and $T$ gates on qubits with only a constant overhead loss, making it possible to achieve constant-overhead distillation of such standard magic states for qubits. These results resolve the fundamental open problem in quantum information theory concerning the construction of magic state distillation protocols with the optimal exponent., Comment: 55 pages, Added comments on independent work

The second law of thermodynamics is the cornerstone of physics, characterizing the convertibility between thermodynamic states through a single function, entropy. Given the universal applicability of thermodynamics, a fundamental question in quantum information theory is whether an analogous second law can be formulated to characterize the convertibility of resources for quantum information processing by a single function. In 2008, a promising formulation was proposed, linking resource convertibility to the optimal performance of a variant of the quantum version of hypothesis testing. Central to this formulation was the generalized quantum Stein's lemma, which aimed to characterize this optimal performance by a measure of quantum resources, the regularized relative entropy of resource. If proven valid, the generalized quantum Stein's lemma would lead to the second law for quantum resources, with the regularized relative entropy of resource taking the role of entropy in thermodynamics. However, in 2023, a logical gap was found in the original proof of this lemma, casting doubt on the possibility of such a formulation of the second law. In this work, we address this problem by developing alternative techniques to successfully prove the generalized quantum Stein's lemma under a smaller set of assumptions than the original analysis. Based on our proof, we reestablish and extend the second law of quantum resource theories, applicable to both static resources of quantum states and a fundamental class of dynamical resources represented by classical-quantum (CQ) channels. These results resolve the fundamental problem of bridging the analogy between thermodynamics and quantum information theory., Comment: 33 pages, 1 figure

Detecting out-of-distribution (OOD) samples is crucial for ensuring the safety of machine learning systems and has shaped the field of OOD detection. Meanwhile, several other problems are closely related to OOD detection, including anomaly detection (AD), novelty detection (ND), open set recognition (OSR), and outlier detection (OD). To unify these problems, a generalized OOD detection framework was proposed, taxonomically categorizing these five problems. However, Vision Language Models (VLMs) such as CLIP have significantly changed the paradigm and blurred the boundaries between these fields, again confusing researchers. In this survey, we first present a generalized OOD detection v2, encapsulating the evolution of AD, ND, OSR, OOD detection, and OD in the VLM era. Our framework reveals that, with some field inactivity and integration, the demanding challenges have become OOD detection and AD. In addition, we also highlight the significant shift in the definition, problem settings, and benchmarks; we thus feature a comprehensive review of the methodology for OOD detection, including the discussion over other related tasks to clarify their relationship to OOD detection. Finally, we explore the advancements in the emerging Large Vision Language Model (LVLM) era, such as GPT-4V. We conclude this survey with open challenges and future directions., Comment: survey paper. We welcome questions, issues, and paper requests via https://github.com/AtsuMiyai/Awesome-OOD-VLM

Neutral atoms are among the leading platforms toward realizing fault-tolerant quantum computation (FTQC). However, scaling up a single neutral-atom device beyond $\sim 10^4$ atoms to meet the demands of FTQC for practical applications remains a challenge. To overcome this challenge, we clarify the criteria and technological requirements for further scaling based on multiple neutral atom quantum processing units (QPUs) connected through photonic networking links. Our quantitative analysis shows that nanofiber optical cavities have the potential as an efficient atom-photon interface to enable fast entanglement generation between atoms in distinct neutral-atom modules, allowing multiple neutral-atom QPUs to operate cooperatively without sacrificing computational speed. Using state-of-the-art millimeter-scale nanofiber cavities with the finesse of thousands, over a hundred atoms can be coupled to the cavity mode with an optical tweezer array, with expected single-atom cooperativity exceeding 100 for telecom-band transition of ytterbium atoms. This enables efficient time-multiplexed entanglement generation with a predicted Bell pair generation rate of 100 kHz while maintaining a small footprint for channel multiplexing. These proposals and results indicate a promising pathway for building large-scale multiprocessor fault-tolerant quantum computers using neutral atoms, nanofiber optical cavities, and fiber-optic networks.

Quantum state preparation (QSP) is a fundamental task in quantum computation to prepare a quantum state for a given classical description of the quantum state. The classical description of an $n$-qubit quantum state may have $\exp(O(n))$ parameters in general, which are inherently inefficient to deal with in the worst case; however, in many practical cases, we may be able to employ suitable data structures to represent such large-scale data in a compressed way, e.g., by using a free binary decision diagram (FBDD), a rooted directed acyclic graph with two terminal nodes to concisely represent a Boolean function. We here construct a quantum algorithm for QSP when the classical description of a quantum state is given by an FBDD with weighted edges, and analyze the space, and time complexity of QSP in this setting. We provide a nontrivial example of an $n$-qubit state that can be represented by a weighted FBDD with $N=O(\mathrm{poly}(n))$ nodes rather than $\mathrm{exp}(O(n))$. We show that any quantum state represented by the weighted FBDD with $N$ nodes can be prepared by an $O(N)$-sized quantum circuit using $N$ ancillary qubits, exponentially improving the required circuit size for QSP compared to other BDD-based QSPs. We also provide another example of an $n$-qubit state that can be represented by a weighted FBDD with $N=O(n^2)$ nodes, and $O(n^2)$ ancillary qubits, but cannot be prepared efficiently by a QSP based on the amplitude amplification. These results provide techniques to employ FBDDs as a tool for broadening the possibility of efficient QSP., Comment: 19 pages with 1 table and 9 figures

In classical information theory, channel capacity quantifies the maximum number of messages that can be reliably transmitted using shared information. An equivalent concept, termed uncommon information, represents the number of messages required to be exchanged to completely share all information in common. However, this equivalence does not extend to quantum information theory. Specifically, quantum uncommon information is operationally defined as the minimal amount of entanglement required for the quantum communication task of quantum state exchange, where two parties exchange quantum states to share all quantum messages in common. Currently, an analytical closed-form expression for the quantum uncommon information remains undetermined. In this work, by investigating underlying characterization of the quantum uncommon information, we derive improved bounds on it. To obtain these bounds, we develop a subspace exchange strategy that leverages a common subspace of two parties to identify the unnecessary qubits for exchange. We also consider a referee-assisted exchange, wherein a referee aids two parties in efficiently performing the quantum state exchange. Our bounds provide more precise estimations for the quantum uncommon information. Furthermore, we demonstrate that the subspace technique is a versatile tool for characterizing uncommon information not only in the bipartite scenario but also in various multi-partite ones., Comment: 24 pages, 7 figures, typos were corrected and readability was improved

Dataset bias is a significant problem in training fair classifiers. When attributes unrelated to classification exhibit strong biases towards certain classes, classifiers trained on such dataset may overfit to these bias attributes, substantially reducing the accuracy for minority groups. Mitigation techniques can be categorized according to the availability of bias information (\ie, prior knowledge). Although scenarios with unknown biases are better suited for real-world settings, previous work in this field often suffers from a lack of interpretability regarding biases and lower performance. In this study, we propose a framework to identify potential biases as keywords without prior knowledge based on the partial occurrence in the captions. We further propose two debiasing methods: (a) handing over to an existing debiasing approach which requires prior knowledge by assigning pseudo-labels, and (b) employing data augmentation via text-to-image generative models, using acquired bias keywords as prompts. Despite its simplicity, experimental results show that our framework not only outperforms existing methods without prior knowledge, but also is even comparable with a method that assumes prior knowledge.

Semi-supervised learning (SSL) can improve model performance by leveraging unlabeled images, which can be collected from public image sources with low costs. In recent years, synthetic images have become increasingly common in public image sources due to rapid advances in generative models. Therefore, it is becoming inevitable to include existing synthetic images in the unlabeled data for SSL. How this kind of contamination will affect SSL remains unexplored. In this paper, we introduce a new task, Real-Synthetic Hybrid SSL (RS-SSL), to investigate the impact of unlabeled data contaminated by synthetic images for SSL. First, we set up a new RS-SSL benchmark to evaluate current SSL methods and found they struggled to improve by unlabeled synthetic images, sometimes even negatively affected. To this end, we propose RSMatch, a novel SSL method specifically designed to handle the challenges of RS-SSL. RSMatch effectively identifies unlabeled synthetic data and further utilizes them for improvement. Extensive experimental results show that RSMatch can transfer synthetic unlabeled data from `obstacles' to `resources.' The effectiveness is further verified through ablation studies and visualization.

Ordinal regression (OR) is classification of ordinal data in which the underlying categorical target variable has a natural ordinal relation for the underlying explanatory variable. For $K$-class OR tasks, threshold methods learn a one-dimensional transformation (1DT) of the explanatory variable so that 1DT values for observations of the explanatory variable preserve the order of label values $1,\ldots,K$ for corresponding observations of the target variable well, and then assign a label prediction to the learned 1DT through threshold labeling, namely, according to the rank of an interval to which the 1DT belongs among intervals on the real line separated by $(K-1)$ threshold parameters. In this study, we propose a parallelizable algorithm to find the optimal threshold labeling, which was developed in previous research, and derive sufficient conditions for that algorithm to successfully output the optimal threshold labeling. In a numerical experiment we performed, the computation time taken for the whole learning process of a threshold method with the optimal threshold labeling could be reduced to approximately 60\,\% by using the proposed algorithm with parallel processing compared to using an existing algorithm based on dynamic programming.

Threshold methods are popular for ordinal regression problems, which are classification problems for data with a natural ordinal relation. They learn a one-dimensional transformation (1DT) of observations of the explanatory variable, and then assign label predictions to the observations by thresholding their 1DT values. In this paper, we study the influence of the underlying data distribution and of the learning procedure of the 1DT on the classification performance of the threshold method via theoretical considerations and numerical experiments. Consequently, for example, we found that threshold methods based on typical learning procedures may perform poorly when the probability distribution of the target variable conditioned on an observation of the explanatory variable tends to be non-unimodal. Another instance of our findings is that learned 1DT values are concentrated at a few points under the learning procedure based on a piecewise-linear loss function, which can make difficult to classify data well.

We report on the development and commissioning of a new Q-band receiver for the Nobeyama 45-m telescope, covering 30--50 GHz with a receiver noise temperature of about 15 K. We name it eQ (extended Q-band) receiver. The system noise temperatures for observations are measured to be $\sim$ 30 K at 33 GHz and $\sim$ 75 K at 45 GHz. The Half-Power-Beam-Width (HPBW) is around 38\arcsec at 43 GHz. To enhance the observation capability, we tested the smoothed bandpass calibration technique and demonstrated the observation time can be significantly reduced compared to the standard position switch technique. The wide-bandwidth capability of this receiver provides precise determination of rest frequencies for molecular transitions with an accuracy of a few kHz through simultaneous observations of multiple transitions. Particularly, we determined the rest frequency of SO ($J_N$ = $1_0$--$0_1$) to be 30.001542 GHz, along with the rest frequency of CCS ($J_N$ = $4_3$--$3_2$) being 45.379033 GHz, adopting CCS ($J_N$ = $3_2$--$2_1$) at 33.751370 GHz as a reference line. The SO profile shows a double peak shape at the Cyanopolyyne Peak (CP) position of the Taurus Molecular Cloud-1 (TMC-1). The SO peaks coincide well with the CCS sub-components located near the outer parts of the TMC-1 filament. We interpret that the gravitational infall of TMC-1 generates shocks which enhance the SO abundance. The TMC-1 map shows that carbon-chain molecules are more abundant in the southern part of the filament, whereas SO is more abundant in the northern part. The eQ's excellent sensitivity allowed us to detect faint CO ($J$ = 1--0) spectra from the high-redshift object at a redshift of 2.442. Our receiver is expected to open new avenues for high-sensitivity molecular line observations in the Q-band., Comment: 20 pages, 18 figures, PASJ in press

Current video summarization methods rely heavily on supervised computer vision techniques, which demands time-consuming and subjective manual annotations. To overcome these limitations, we investigated self-supervised video summarization. Inspired by the success of Large Language Models (LLMs), we explored the feasibility in transforming the video summarization task into a Natural Language Processing (NLP) task. By leveraging the advantages of LLMs in context understanding, we aim to enhance the effectiveness of self-supervised video summarization. Our method begins by generating captions for individual video frames, which are then synthesized into text summaries by LLMs. Subsequently, we measure semantic distance between the captions and the text summary. Notably, we propose a novel loss function to optimize our model according to the diversity of the video. Finally, the summarized video can be generated by selecting the frames with captions similar to the text summary. Our method achieves state-of-the-art performance on the SumMe dataset in rank correlation coefficients. In addition, our method has a novel feature of being able to achieve personalized summarization.

The second "Mineral Detection of Neutrinos and Dark Matter" (MDvDM'24) meeting was held January 8-11, 2024 in Arlington, VA, USA, hosted by Virginia Tech's Center for Neutrino Physics. This document collects contributions from this workshop, providing an overview of activities in the field. MDvDM'24 was the second topical workshop dedicated to the emerging field of mineral detection of neutrinos and dark matter, following a meeting hosted by IFPU in Trieste, Italy in October 2022. Mineral detectors have been proposed for a wide variety of applications, including searching for dark matter, measuring various fluxes of astrophysical neutrinos over gigayear timescales, monitoring nuclear reactors, and nuclear disarmament protocols; both as paleo-detectors using natural minerals that could have recorded the traces of nuclear recoils for timescales as long as a billion years and as detectors recording nuclear recoil events on laboratory timescales using natural or artificial minerals. Contributions to this proceedings discuss the vast physics potential, the progress in experimental studies, and the numerous challenges lying ahead on the path towards mineral detection. These include a better understanding of the formation and annealing of recoil defects in crystals; identifying the best classes of minerals and, for paleo-detectors, understanding their geology; modeling and control of the relevant backgrounds; developing, combining, and scaling up imaging and data analysis techniques; and many others. During the last years, MDvDM has grown rapidly and gained attention. Small-scale experimental efforts focused on establishing various microscopic readout techniques are underway at institutions in North America, Europe and Asia. We are looking ahead to an exciting future full of challenges to overcome, surprises to be encountered, and discoveries lying ahead of us., Comment: Summary and proceedings of the MDvDM'24 conference, Jan 8-11 2024