Search

This study investigates differential item functioning (DIF) detection in computerized adaptive testing (CAT) using multilevel modeling. We argue that traditional DIF methods have proven ineffective in CAT due to the hierarchical nature of the data. Our proposed two-level model accounts for dependencies between items via provisional ability estimates. Simulations revealed that our model outperformed others in Type-I error control and power, particularly in scenarios with high exposure rates and longer tests. Expanding item pools, incorporating item parameters, and exploring Bayesian estimation are recommended for future research to further enhance DIF detection in CAT. Balancing model complexity with convergence remains a key challenge for robust outcomes., Comment: 38 pages, preprint

Speech dysfluency modeling is a task to detect dysfluencies in speech, such as repetition, block, insertion, replacement, and deletion. Most recent advancements treat this problem as a time-based object detection problem. In this work, we revisit this problem from a new perspective: tokenizing dysfluencies and modeling the detection problem as a token-based automatic speech recognition (ASR) problem. We propose rule-based speech and text dysfluency simulators and develop VCTK-token, and then develop a Whisper-like seq2seq architecture to build a new benchmark with decent performance. We also systematically compare our proposed token-based methods with time-based methods, and propose a unified benchmark to facilitate future research endeavors. We open-source these resources for the broader scientific community. The project page is available at https://rorizzz.github.io/

Few health-related constructs or measures have received critical evaluation in terms of measurement equivalence, such as self-reported health survey data. Differential item functioning (DIF) analysis is crucial for evaluating measurement equivalence in self-reported health surveys, which are often hierarchical in structure. While traditional DIF methods rely on single-level models, multilevel models offer a more suitable alternative for analyzing such data. In this article, we highlight the advantages of multilevel modeling in DIF analysis and demonstrate how to apply the DIF framework to self-reported health survey data using multilevel models. For demonstration, we analyze DIF associated with population density on the probability to answer "Yes" to a survey question on depression and reveal that multilevel models achieve better fit and account for more variance compared to single-level models. This article is expected to increase awareness of the usefulness of multilevel modeling for DIF analysis and assist healthcare researchers and practitioners in improving the understanding of self-reported health survey data validity., Comment: preprint, 11 pages (excluding references)

This study evaluated four multi-group differential item functioning (DIF) methods (the root mean square deviation approach, Wald-1, generalized logistic regression procedure, and generalized Mantel-Haenszel method) via Monte Carlo simulation of controlled testing conditions. These conditions varied in the number of groups, the ability and sample size of the DIF-contaminated group, the parameter associated with DIF, and the proportion of DIF items. When comparing Type-I error rates and powers of the methods, we showed that the RMSD approach yielded the best Type-I error rates when it was used with model-predicted cutoff values. Also, this approach was found to be overly conservative when used with the commonly used cutoff value of 0.1. Implications for future research for educational researchers and practitioners were discussed., Comment: preprint, 16 pages (excluding figures, references, and title page)

The upcoming Chinese Space Station Telescope (CSST) slitless spectroscopic survey poses a challenge of flux calibration, which requires a large number of flux-standard stars. In this work, we design an uncertainty-aware residual attention network, the UaRA-net, to derive the CSST SEDs with a resolution of R = 200 over the wavelength range of 2500-10000 \AA using LAMOST normalized spectra with a resolution of R = 2000 over the wavelength range of 4000-7000 \AA. With the special structure and training strategy, the proposed model can not only provide accurate predictions of SEDs but also their corresponding errors. The precision of the predicted SEDs depends on effective temperature (Teff), wavelength, and the LAMOST spectral signal-to-noise ratios (SNRs), particularly in the GU band. For stars with Teff = 6000 K, the typical SED precisions in the GU band are 4.2%, 2.1%, and 1.5% at SNR values of 20, 40, and 80, respectively. As Teff increases to 8000 K, the precision increases to 1.2%, 0.6%, and 0.5%, respectively. The precision is higher at redder wavelengths. In the GI band, the typical SED precisions for stars with Teff = 6000 K increase to 0.3%, 0.1%, and 0.1% at SNR values of 20, 40, and 80, respectively. We further verify our model using the empirical spectra of the MILES and find good performance. The proposed method will open up new possibilities for optimal utilization of slitless spectra of the CSST and other surveys., Comment: 20 pages, 15 figures, accepted by ApJS

Point-cloud-based 3D perception has attracted great attention in various applications including robotics, autonomous driving and AR/VR. In particular, the 3D sparse convolution (SpConv) network has emerged as one of the most popular backbones due to its excellent performance. However, it poses severe challenges to real-time perception on general-purpose platforms, such as lengthy map search latency, high computation cost, and enormous memory footprint. In this paper, we propose SpOctA, a SpConv accelerator that enables high-speed and energy-efficient point cloud processing. SpOctA parallelizes the map search by utilizing algorithm-architecture co-optimization based on octree encoding, thereby achieving 8.8-21.2x search speedup. It also attenuates the heavy computational workload by exploiting inherent sparsity of each voxel, which eliminates computation redundancy and saves 44.4-79.1% processing latency. To optimize on-chip memory management, a SpConv-oriented non-uniform caching strategy is introduced to reduce external memory access energy by 57.6% on average. Implemented on a 40nm technology and extensively evaluated on representative benchmarks, SpOctA rivals the state-of-the-art SpConv accelerators by 1.1-6.9x speedup with 1.5-3.1x energy efficiency improvement., Comment: Accepted to ICCAD 2023

Domain generalization (DG) tends to alleviate the poor generalization capability of deep neural networks by learning model with multiple source domains. A classical solution to DG is domain augmentation, the common belief of which is that diversifying source domains will be conducive to the out-of-distribution generalization. However, these claims are understood intuitively, rather than mathematically. Our explorations empirically reveal that the correlation between model generalization and the diversity of domains may be not strictly positive, which limits the effectiveness of domain augmentation. This work therefore aim to guarantee and further enhance the validity of this strand. To this end, we propose a new perspective on DG that recasts it as a convex game between domains. We first encourage each diversified domain to enhance model generalization by elaborately designing a regularization term based on supermodularity. Meanwhile, a sample filter is constructed to eliminate low-quality samples, thereby avoiding the impact of potentially harmful information. Our framework presents a new avenue for the formal analysis of DG, heuristic analysis and extensive experiments demonstrate the rationality and effectiveness., Comment: accepted by CVPR 2023

Accurate and comprehensive material databases extracted from research papers are crucial for materials science and engineering, but their development requires significant human effort. With large language models (LLMs) transforming the way humans interact with text, LLMs provide an opportunity to revolutionize data extraction. In this study, we demonstrate a simple and efficient method for extracting materials data from full-text research papers leveraging the capabilities of LLMs combined with human supervision. This approach is particularly suitable for mid-sized databases and requires minimal to no coding or prior knowledge about the extracted property. It offers high recall and nearly perfect precision in the resulting database. The method is easily adaptable to new and superior language models, ensuring continued utility. We show this by evaluating and comparing its performance on GPT-3 and GPT-3.5/4 (which underlie ChatGPT), as well as free alternatives such as BART and DeBERTaV3. We provide a detailed analysis of the method's performance in extracting sentences containing bulk modulus data, achieving up to 90% precision at 96% recall, depending on the amount of human effort involved. We further demonstrate the method's broader effectiveness by developing a database of critical cooling rates for metallic glasses over twice the size of previous human curated databases., Comment: 13 pages, 4 figures

Extensive studies on Unsupervised Domain Adaptation (UDA) have propelled the deployment of deep learning from limited experimental datasets into real-world unconstrained domains. Most UDA approaches align features within a common embedding space and apply a shared classifier for target prediction. However, since a perfectly aligned feature space may not exist when the domain discrepancy is large, these methods suffer from two limitations. First, the coercive domain alignment deteriorates target domain discriminability due to lacking target label supervision. Second, the source-supervised classifier is inevitably biased to source data, thus it may underperform in target domain. To alleviate these issues, we propose to simultaneously conduct feature alignment in two individual spaces focusing on different domains, and create for each space a domain-oriented classifier tailored specifically for that domain. Specifically, we design a Domain-Oriented Transformer (DOT) that has two individual classification tokens to learn different domain-oriented representations, and two classifiers to preserve domain-wise discriminability. Theoretical guaranteed contrastive-based alignment and the source-guided pseudo-label refinement strategy are utilized to explore both domain-invariant and specific information. Comprehensive experiments validate that our method achieves state-of-the-art on several benchmarks., Comment: Accepted at ACMMM 2022

In this letter, we consider hybrid beamforming for millimeter wave (mmWave) MIMO integrated sensing and communications (ISAC). We design the transmit beam of a dual-functional radar-communication (DFRC) base station (BS), aiming at approaching the objective radar beam pattern, subject to the constraints of the signal to interference-plus-noise ratio (SINR) of communication users and total transmission power of the DFRC BS. To provide additional degree of freedom for the beam design problem, we introduce a phase vector to the objective beam pattern and propose an alternating minimization method to iteratively optimize the transmit beam and the phase vector, which involves second-order cone programming and constrained least squared estimation, respectively. Then based on the designed transmit beam, we determine the analog beamformer and digital beamformer subject to the constant envelop constraint of phase shifter network in mmWave MIMO, still using the alternating minimization method. Simulation results show that under the same SINR constraint of communication users, larger antenna array can achieve better radar beam quality.