Search

This paper provides a detailed survey of synthetic data techniques. We first discuss the expected goals of using synthetic data in data augmentation, which can be divided into four parts: 1) Improving Diversity, 2) Data Balancing, 3) Addressing Domain Shift, and 4) Resolving Edge Cases. Synthesizing data are closely related to the prevailing machine learning techniques at the time, therefore, we summarize the domain of synthetic data techniques into four categories: 1) Expert-knowledge, 2) Direct Training, 3) Pre-train then Fine-tune, and 4) Foundation Models without Fine-tuning. Next, we categorize the goals of synthetic data filtering into four types for discussion: 1) Basic Quality, 2) Label Consistency, and 3) Data Distribution. In section 5 of this paper, we also discuss the future directions of synthetic data and state three direction that we believe is important: 1) focus more on quality, 2) the evaluation of synthetic data, and 3) multi-model data augmentation.

In an axion haloscope, the weak photon signal, theoretically converted from axions, is captured by a detection cavity. The amplification chain assists the signal receiver to read the signal from the cavity and requires accurate calibration. Typically, the readout line is calibrated using the Y-factor method, involving a switch that directs the signal from either the detection line or the calibration line. However, this switch introduces several disadvantages. In response, we propose a calibration method that eliminates the need for a switch. In this approach, the cavity temperature is decoupled to its incoming noise source and can be controlled, resulting in excess or deficiency of the noise spectrum near its resonance frequency. The experimental result shows that the amplification chain can be calibrated directly using the temperature-varied cavity-emitted radiation., Comment: 6 journal pages, 5 figures, and 1 table in the main text

This Letter reports on the first results from the Taiwan Axion Search Experiment with Haloscope, a search for axions using a microwave cavity at frequencies between 4.70750 and 4.79815 GHz. Apart from the non-axion signals, no candidates with a significance more than 3.355 were found. The experiment excludes models with the axion-two-photon coupling $\left|g_{a\gamma\gamma}\right|\gtrsim 8.2\times 10^{-14}$ GeV$^{-1}$, a factor of eleven above the benchmark KSVZ model, reaching a sensitivity three orders of magnitude better than any existing limits in the mass range 19.4687 < $m_a$ < 19.8436 $\mu$eV. It is also the first time that a haloscope-type experiment places constraints on $g_{a\gamma\gamma}$ in this mass region., Comment: arXiv admin note: substantial text overlap with arXiv:2204.14265

We report on a holoscope axion search experiment near $19.6\ {\rm \mu eV}$ from the TASEH collaboration. The experiment is carried out via a frequency-tunable cavity detector with a volume $V = 0.234\ {\rm liter}$ in a magnetic field $B_0 = 8\ {\rm T}$. With a signal receiver that has a system noise temperature $T_{\rm sys} \cong 2.2\ {\rm K}$ and experiment time about 1 month, the search excludes values of the axion-photon coupling constant $g_{\rm a\gamma\gamma} \gtrsim 8.1 \times 10^{-14} \ {\rm GeV}^{-1}$, a factor of 11 above the KSVZ model, at the 95\% confidence level in the mass range of $19.4687-19.8436\ {\rm \mu eV}$. We present the experimental setup and procedures to accomplish this search., Comment: 16 pages, 14 figures, 3 tables

This paper presents the analysis of the data acquired during the first physics run of the Taiwan Axion Search Experiment with Haloscope (TASEH), a search for axions using a microwave cavity at frequencies between 4.70750 and 4.79815 GHz. The data were collected from October 13, 2021 to November 15, 2021, and are referred to as the CD102 data. The analysis of the TASEH CD102 data excludes models with the axion-two-photon coupling $|g_{a\gamma\gamma}| \gtrsim 8.2\times 10^{-14}$ GeV$^{-1}$, a factor of eleven above the benchmark KSVZ model for the mass range 19.4687 < ma < 19.8436 $\mu$eV., Comment: Submitted to Physical Review D

Millimeter wave (mmWave) is a key technology for fifth-generation (5G) and beyond communications. Hybrid beamforming has been proposed for large-scale antenna systems in mmWave communications. Existing hybrid beamforming designs based on infinite-resolution phase shifters (PSs) are impractical due to hardware cost and power consumption. In this paper, we propose an unsupervised-learning-based scheme to jointly design the analog precoder and combiner with low-resolution PSs for multiuser multiple-input multiple-output (MU-MIMO) systems. We transform the analog precoder and combiner design problem into a phase classification problem and propose a generic neural network architecture, termed the phase classification network (PCNet), capable of producing solutions of various PS resolutions. Simulation results demonstrate the superior sum-rate and complexity performance of the proposed scheme, as compared to state-of-the-art hybrid beamforming designs for the most commonly used low-resolution PS configurations., Comment: IEEE International Conference on Communications (ICC) 2022

Cycle-consistent generative adversarial networks (CycleGAN) were successfully applied to speech enhancement (SE) tasks with unpaired noisy-clean training data. The CycleGAN SE system adopted two generators and two discriminators trained with losses from noisy-to-clean and clean-to-noisy conversions. CycleGAN showed promising results for numerous SE tasks. Herein, we investigate a potential limitation of the clean-to-noisy conversion part and propose a novel noise-informed training (NIT) approach to improve the performance of the original CycleGAN SE system. The main idea of the NIT approach is to incorporate target domain information for clean-to-noisy conversion to facilitate a better training procedure. The experimental results confirmed that the proposed NIT approach improved the generalization capability of the original CycleGAN SE system with a notable margin.