Search

Harmonic-assisted phase amplification was achieved in a 300-micrometer-thick Nd:GdVO4 laser in the self-mixing interference scheme. The key event is the self-induced skew cosh Gaussian (e.g., skew-chG) mode oscillation in a thin-slice solid-state laser with wide-aperture laser-diode pumping. The skew-chG mode was proved to be formed by the phase-locking of nearly frequency-degenerate TEM00 and annular fields. The resultant modal-interference-induced gain modulation at the beat frequency between the two modal fields, which is far above the relaxation oscillation frequency, increased experimental self-mixing modulation bandwidth accordingly. Fifty-fold phase amplification was achieved in a strong optical feedback regime., Comment: 9 pages, 13 figures

Nonreciprocal charge transport in solids, where resistance is different between rightward and leftward currents, is a key function of rectifying devices in the modern electronics, as exemplified by $p$-$n$ semiconductor junctions. Recently, this was also demonstrated in noncentrosymmetric materials in magnetic fields, since their band structure exhibits spin polarization coupled to the position of momentum space due to the antisymmetric spin-orbit coupling. To enhance the magnitude of nonreciprocal effect, it is essential to tune such spin-momentum coupling, which has been hampered in the conventional materials owing to the difficulty in controlling the broken inversion symmetry built into the lattice and interfacial structures. Here we report large nonreciprocal resistivity in layered polar metal BaMn$X_2$ ($X$=Sb, Bi), where the spin-polarized Dirac dispersion depends on the in-plane polarization tunable by chemical substitution of the $X$ site. For $X$=Sb with a pair of single-type valleys, the nonreciprocal resistivity increases monotonically with decreasing temperature, while for $X$=Bi with multiple types of valleys it is reduced by about an order of magnitude and exhibits a peak at a low temperature. Theoretical calculations indicate that the nonreciprocal resistivity is sensitive not only to the spin-momentum (spin-valley) coupling, but also to the Fermi energy and the Dirac dispersion. The observed significant variation of nonreciprocal transport in the same series of materials might be of great use in the design of junction-free rectifying devices and circuits., Comment: 10 pages, 7 figures. This article is already published in Physical Review Research

In current mobile networks, optimizing which base station a mobile phone in a particular area connects to is crucial for ensuring good communication quality for each mobile phone but presents a challenging combinatorial optimization problem. In this study, we optimize the connection patterns to base stations using quantum annealing which is a generic solver using quantum fluctuation. However, since the number of qubits on a quantum annealer is limited, it is necessary to consider a formulation that efficiently utilizes qubits. By adopting a variable reduction formulation, we significantly reduce the qubit requirements compared to the naive formulation that is typically used when considering pattern-matching problems. Furthermore, experiments using quantum annealing revealed that the accuracy of the approximate solution obtained by the new formulation is superior to that of the conventional formulation. Additionally, we demonstrate that the new formulation provides better solutions than the conventional formulation as the problem size increases, even when using simulated annealing, the classical counterpart of quantum annealing., Comment: 10 pages, 7 figures, 1 table

This paper addresses the collision detection problem in population protocols. The network consists of state machines called agents. At each time step, exactly one pair of agents is chosen uniformly at random to have an interaction, changing the states of the two agents. The collision detection problem involves each agent starting with an input integer between $1$ and $n$, where $n$ is the number of agents, and requires those agents to determine whether there are any duplicate input values among all agents. Specifically, the goal is for all agents to output false if all input values are distinct, and true otherwise. In this paper, we present an algorithm that requires a polynomial number of states per agent and solves the collision detection problem with probability one in sub-linear parallel time, both with high probability and in expectation. To the best of our knowledge, this algorithm is the first to solve the collision detection problem using a polynomial number of states within sublinear parallel time, affirmatively answering the question raised by Burman, Chen, Chen, Doty, Nowak, Severson, and Xu [PODC 2021] for the first time.

We consider the population protocol model where indistinguishable state machines, referred to as agents, communicate in pairs. The communication graph specifies potential interactions (\ie communication) between agent pairs. This paper addresses the complete graph identification problem, requiring agents to determine if their communication graph is a clique or not. We evaluate various settings based on: (i) the fairness preserved by the adversarial scheduler -- either global fairness or weak fairness, and (ii) the knowledge provided to agents beforehand -- either the exact population size $n$, a common upper bound $P$ on $n$, or no prior information. Positively, we show that $O(n^2)$ states per agent suffice to solve the complete graph identification problem under global fairness without prior knowledge. With prior knowledge of $n$, agents can solve the problem using only $O(n)$ states under weak fairness. Negatively, we prove that complete graph identification remains unsolvable under weak fairness when only a common upper bound $P$ on the population size $n$ is known.

To achieve state-of-the-art jet energy resolution for Particle Flow, sophisticated energy clustering algorithms must be developed that can fully exploit available information to separate energy deposits from charged and neutral particles. Three published neural network-based shower separation models were applied to simulation and experimental data to measure the performance of the highly granular CALICE Analogue Hadronic Calorimeter (AHCAL) technological prototype in distinguishing the energy deposited by a single charged and single neutral hadron for Particle Flow. The performance of models trained using only standard spatial and energy and charged track position information from an event was compared to models trained using timing information available from AHCAL, which is expected to improve sensitivity to shower development and, therefore, aid in clustering. Both simulation and experimental data were used to train and test the models and their performances were compared. The best-performing neural network achieved significantly superior event reconstruction when timing information was utilised in training for the case where the charged hadron had more energy than the neutral one, motivating temporally sensitive calorimeters. All models under test were observed to tend to allocate energy deposited by the more energetic of the two showers to the less energetic one. Similar shower reconstruction performance was observed for a model trained on simulation and applied to data and a model trained and applied to data.

Contextualized end-to-end automatic speech recognition has been an active research area, with recent efforts focusing on the implicit learning of contextual phrases based on the final loss objective. However, these approaches ignore the useful contextual knowledge encoded in the intermediate layers. We hypothesize that employing explicit biasing loss as an auxiliary task in the encoder intermediate layers may better align text tokens or audio frames with the desired objectives. Our proposed intermediate biasing loss brings more regularization and contextualization to the network. Our method outperforms a conventional contextual biasing baseline on the LibriSpeech corpus, achieving a relative improvement of 22.5% in biased word error rate (B-WER) and up to 44% compared to the non-contextual baseline with a biasing list size of 100. Moreover, employing RNN-transducer-driven joint decoding further reduces the unbiased word error rate (U-WER), resulting in a more robust network., Comment: Accepted to INTERSPEECH 2024

End-to-end automatic speech recognition (E2E-ASR) can be classified by its decoder architectures, such as connectionist temporal classification (CTC), recurrent neural network transducer (RNN-T), attention-based encoder-decoder, and Mask-CTC models. Each decoder architecture has advantages and disadvantages, leading practitioners to switch between these different models depending on application requirements. Instead of building separate models, we propose a joint modeling scheme where four decoders (CTC, RNN-T, attention, and Mask-CTC) share the same encoder -- we refer to this as 4D modeling. The 4D model is trained jointly, which will bring model regularization and maximize the model robustness thanks to their complementary properties. To efficiently train the 4D model, we introduce a two-stage training strategy that stabilizes the joint training. In addition, we propose three novel joint beam search algorithms by combining three decoders (CTC, RNN-T, and attention) to further improve performance. These three beam search algorithms differ in which decoder is used as the primary decoder. We carefully evaluate the performance and computational tradeoffs associated with each algorithm. Experimental results demonstrate that the jointly trained 4D model outperforms the E2E-ASR models trained with only one individual decoder. Furthermore, we demonstrate that the proposed joint beam search algorithm outperforms the previously proposed CTC/attention decoding., Comment: accepted to IEEE/ACM Transactions on Audio Speech and Language Processing

End-to-end (E2E) automatic speech recognition (ASR) can operate in two modes: streaming and non-streaming, each with its pros and cons. Streaming ASR processes the speech frames in real-time as it is being received, while non-streaming ASR waits for the entire speech utterance; thus, professionals may have to operate in either mode to satisfy their application. In this work, we present joint optimization of streaming and non-streaming ASR based on multi-decoder and knowledge distillation. Primarily, we study 1) the encoder integration of these ASR modules, followed by 2) separate decoders to make the switching mode flexible, and enhancing performance by 3) incorporating similarity-preserving knowledge distillation between the two modular encoders and decoders. Evaluation results show 2.6%-5.3% relative character error rate reductions (CERR) on CSJ for streaming ASR, and 8.3%-9.7% relative CERRs for non-streaming ASR within a single model compared to multiple standalone modules., Comment: Accepted to IEEE ICASSP 2024 workshop Hands-free Speech Communication and Microphone Arrays (HSCMA 2024)

Deep biasing (DB) enhances the performance of end-to-end automatic speech recognition (E2E-ASR) models for rare words or contextual phrases using a bias list. However, most existing methods treat bias phrases as sequences of subwords in a predefined static vocabulary. This naive sequence decomposition produces unnatural token patterns, significantly lowering their occurrence probability. More advanced techniques address this problem by expanding the vocabulary with additional modules, including the external language model shallow fusion or rescoring. However, they result in increasing the workload due to the additional modules. This paper proposes a dynamic vocabulary where bias tokens can be added during inference. Each entry in a bias list is represented as a single token, unlike a sequence of existing subword tokens. This approach eliminates the need to learn subword dependencies within the bias phrases. This method is easily applied to various architectures because it only expands the embedding and output layers in common E2E-ASR architectures. Experimental results demonstrate that the proposed method improves the bias phrase WER on English and Japanese datasets by 3.1 -- 4.9 points compared with the conventional DB method.

A neural network for software compensation was developed for the highly granular CALICE Analogue Hadronic Calorimeter (AHCAL). The neural network uses spatial and temporal event information from the AHCAL and energy information, which is expected to improve sensitivity to shower development and the neutron fraction of the hadron shower. The neural network method produced a depth-dependent energy weighting and a time-dependent threshold for enhancing energy deposits consistent with the timescale of evaporation neutrons. Additionally, it was observed to learn an energy-weighting indicative of longitudinal leakage correction. In addition, the method produced a linear detector response and outperformed a published control method regarding resolution for every particle energy studied.

Amphiphilic molecules spontaneously form self-assembled structures of various shapes depending on their molecular structures, the temperature, and other physical conditions. The functionalities of these structures are dictated by their formations and their properties must be evaluated for reproduction using molecular simulations. However, the assessment of such intricate structures involves many procedural steps. This study investigates the potential of machine-learning models to extract structural features from mesoscale non-ordered self-assembled structures, and suggests a methodology in which machine-learning models for the structural analysis of self-assembled structures are trained on particle types and coordinate data. In the proposed approach, graph neural networks are utilised to extract local structural data for analysis. In simulations using several hundred self-assembled structures of up to 4050 coarse-grained particles, local structures are successfully extracted and classified with up to 78.35 % accuracy. As the machine-learning models learn structural characteristics without the need for human-made feature engineering, the proposed method has important potential applications in the field of materials science., Comment: 26 pages, 13 figures

There has been an increasing interest in large speech models that can perform multiple tasks in a single model. Such models usually adopt an encoder-decoder or decoder-only architecture due to their popularity and good performance in many domains. However, autoregressive models can be slower during inference compared to non-autoregressive models and also have potential risks of hallucination. Though prior studies observed promising results of non-autoregressive models for certain tasks at small scales, it remains unclear if they can be scaled to speech-to-text generation in diverse languages and tasks. Inspired by the Open Whisper-style Speech Model (OWSM) project, we propose OWSM-CTC, a novel encoder-only speech foundation model based on Connectionist Temporal Classification (CTC). It is trained on 180k hours of public audio data for multilingual automatic speech recognition (ASR), speech translation (ST), and language identification (LID). Compared to encoder-decoder OWSM, our OWSM-CTC achieves competitive results on ASR and up to 24% relative improvement on ST, while it is more robust and 3 to 4 times faster for inference. OWSM-CTC also improves the long-form ASR result with 20x speed-up. We will publicly release our code, pre-trained model, and training logs to promote open science in speech foundation models., Comment: Accepted at ACL 2024 main conference

Recent studies have highlighted the importance of fully open foundation models. The Open Whisper-style Speech Model (OWSM) is an initial step towards reproducing OpenAI Whisper using public data and open-source toolkits. However, previous versions of OWSM (v1 to v3) are still based on standard Transformer, which might lead to inferior performance compared to state-of-the-art speech encoder architectures. This work aims to improve the performance and efficiency of OWSM without additional data. We present a series of E-Branchformer-based models named OWSM v3.1, ranging from 100M to 1B parameters. OWSM v3.1 outperforms its predecessor, OWSM v3, in most evaluation benchmarks, while showing an improved inference speed of up to 25%. We further reveal the emergent ability of OWSM v3.1 in zero-shot contextual biasing speech recognition. We also provide a model trained on a subset of data with low license restrictions. We will publicly release the code, pre-trained models, and training logs., Comment: Accepted at INTERSPEECH 2024. Webpage: https://www.wavlab.org/activities/2024/owsm/

End-to-end (E2E) automatic speech recognition (ASR) methods exhibit remarkable performance. However, since the performance of such methods is intrinsically linked to the context present in the training data, E2E-ASR methods do not perform as desired for unseen user contexts (e.g., technical terms, personal names, and playlists). Thus, E2E-ASR methods must be easily contextualized by the user or developer. This paper proposes an attention-based contextual biasing method that can be customized using an editable phrase list (referred to as a bias list). The proposed method can be trained effectively by combining a bias phrase index loss and special tokens to detect the bias phrases in the input speech data. In addition, to improve the contextualization performance during inference further, we propose a bias phrase boosted (BPB) beam search algorithm based on the bias phrase index probability. Experimental results demonstrate that the proposed method consistently improves the word error rate and the character error rate of the target phrases in the bias list on both the Librispeech-960 (English) and our in-house (Japanese) dataset, respectively., Comment: accepted by ICASSP20224

Cellular ingredient concentrations can be stabilized by adjusting generation and consumption rates through multiple pathways. To explore the portion of cellular metabolism equipped with multiple pathways, we categorize individual metabolic reactions and compounds as viable or inviable: A compound is viable if processed by two or more reactions, and a reaction is viable if all of its substrates and products are viable. Using this classification, we identify the maximal subnetwork of viable nodes, referred to as the {\it viable core}, in bipartite metabolic networks across thousands of species. The obtained viable cores are remarkably larger than those in degree-preserving randomized networks, while their broad degree distributions commonly enable the viable cores to shrink gradually as reaction nodes are deleted. We demonstrate that the positive degree-degree correlations of the empirical networks may underlie the enlarged viable cores compared to the randomized networks. By investigating the relation between degree and cross-species frequency of metabolic compounds and reactions, we elucidate the evolutionary origin of the correlations., Comment: 8 pages, 4 figures

Optimization of circuits is an essential task for both quantum and classical computers to improve their efficiency. In contrast, classical logic optimization is known to be difficult, and a lot of heuristic approaches have been developed so far. In this study, we define and construct a quantum algorithmic primitive called quantum circuit unoptimization, which makes a given quantum circuit complex by introducing some redundancies while preserving circuit equivalence, i.e., the inverse operation of circuit optimization. Using quantum circuit unoptimization, we propose the quantum circuit equivalence test, a decision problem contained both in the NP and BQP classes but is not trivially included in the P class. Furthermore, as a practical application, we construct concrete unoptimization recipes to generate compiler benchmarks and evaluate circuit optimization performance using Qiskit and Pytket. Our numerical simulations demonstrate that quantum circuit unoptimizer systematically generates redundant circuits that are challenging for compilers to optimize, which can be used to compare the performance of different compilers and improve them. We also offer potential applications of quantum circuit unoptimization, such as generating quantum advantageous machine learning datasets and quantum computer fidelity benchmarks., Comment: 10 pages, 5 figures

Research on distributed computing by a team of identical mobile computational entities, called robots, operating in a Euclidean space in $\mathit{Look}$-$\mathit{Compute}$-$\mathit{Move}$ ($\mathit{LCM}$) cycles, has recently focused on better understanding how the computational power of robots depends on the interplay between their internal capabilities (i.e., persistent memory, communication), captured by the four standard computational models (OBLOT, LUMI, FSTA, and FCOM) and the conditions imposed by the external environment, controlling the activation of the robots and their synchronization of their activities, perceived and modeled as an adversarial scheduler. We consider a set of adversarial asynchronous schedulers ranging from the classical semi-synchronous (SSYNCH) and fully asynchronous (ASYNCH) settings, including schedulers (emerging when studying the atomicity of the combination of operations in the $\mathit{LCM}$ cycles) whose adversarial power is in between those two. We ask the question: what is the computational relationship between a model $M_1$ under adversarial scheduler $K_1$ ($M_1(K_1)$) and a model $M_2$ under scheduler $K_2$ ($M_2(K_2)$)? For example, are the robots in $M_1(K_1)$ more powerful (i.e., they can solve more problems) than those in $M_2(K_2)$? We answer all these questions by providing, through cross-model analysis, a complete characterization of the computational relationship between the power of the four models of robots under the considered asynchronous schedulers. In this process, we also provide qualified answers to several open questions, including the outstanding one on the proper dominance of SSYNCH over ASYNCH in the case of unrestricted visibility.

We present a quantum-classical hybrid algorithm for calculating the ground state and its energy of the quantum many-body Hamiltonian by proposing an adaptive construction of a quantum state for the quantum-selected configuration interaction (QSCI) method. QSCI allows us to select important electronic configurations in the system to perform CI calculation (subspace diagonalization of the Hamiltonian) by sampling measurement for a proper input quantum state on a quantum computer, but how we prepare a desirable input state has remained a challenge. We propose an adaptive construction of the input state for QSCI in which we run QSCI repeatedly to grow the input state iteratively. We numerically illustrate that our method, dubbed ADAPT-QSCI, can yield accurate ground-state energies for small molecules, including a noisy situation for eight qubits where error rates of two-qubit gates and the measurement are both as large as 1%. ADAPT-QSCI serves as a promising method to take advantage of current noisy quantum devices and pushes forward its application to quantum chemistry., Comment: 14 pages

Consider that there are $k\le n$ agents in a simple, connected, and undirected graph $G=(V,E)$ with $n$ nodes and $m$ edges. The goal of the dispersion problem is to move these $k$ agents to mutually distinct nodes. Agents can communicate only when they are at the same node, and no other communication means, such as whiteboards, are available. We assume that the agents operate synchronously. We consider two scenarios: when all agents are initially located at a single node (rooted setting) and when they are initially distributed over one or more nodes (general setting). Kshemkalyani and Sharma presented a dispersion algorithm for the general setting, which uses $O(m_k)$ time and $\log(k + \Delta)$ bits of memory per agent [OPODIS 2021], where $m_k$ is the maximum number of edges in any induced subgraph of $G$ with $k$ nodes, and $\Delta$ is the maximum degree of $G$. This algorithm is currently the fastest in the literature, as no $o(m_k)$-time algorithm has been discovered, even for the rooted setting. In this paper, we present significantly faster algorithms for both the rooted and the general settings. First, we present an algorithm for the rooted setting that solves the dispersion problem in $O(k\log \min(k,\Delta))=O(k\log k)$ time using $O(\log (k+\Delta))$ bits of memory per agent. Next, we propose an algorithm for the general setting that achieves dispersion in $O(k \log k \cdot \log \min(k,\Delta))=O(k \log^2 k)$ time using $O(\log (k+\Delta))$ bits. Finally, for the rooted setting, we give a time-optimal (i.e.,~$O(k)$-time) algorithm with $O(\Delta+\log k)$ bits of space per agent. All algorithms presented in this paper work only in the synchronous setting, while several algorithms in the literature, including the one given by Kshemkalyani and Sharma at OPODIS 2021, work in the asynchronous setting.