Predicting future walking joint kinematics is crucial for assistive device control, especially in variable walking environments. Traditional optical motion capture systems provide kinematics data but require laborious post-processing, whereas IMU based systems provide direct calculations but add delays due to data collection and algorithmic processes. Predicting future kinematics helps to compensate for these delays, enabling the system real-time. Furthermore, these predicted kinematics could serve as target trajectories for assistive devices such as exoskeletal robots and lower limb prostheses. However, given the complexity of human mobility and environmental factors, this prediction remains to be challenging. To address this challenge, we propose the Dual-ED-Attention-FAM-Net, a deep learning model utilizing two encoders, two decoders, a temporal attention module, and a feature attention module. Our model outperforms the state-of-the-art LSTM model. Specifically, for Dataset A, using IMUs and a combination of IMUs and videos, RMSE values decrease from 4.45° to 4.22° and from 4.52° to 4.15°, respectively. For Dataset B, IMUs and IMUs combined with pressure insoles result in RMSE reductions from 7.09° to 6.66° and from 7.20° to 6.77°, respectively. Additionally, incorporating other modalities alongside IMUs helps improve the performance of the model.
}, author={Hossain, Md Sanzid Bin and Guo, Zhishan and Sui, Ning and Choi, Hwan}, year={2023}, month={Oct} } @inproceedings{li_sui_guo_guo_2023, title={Real-Time Deep Learning Framework for Dermatology Image Classification on Low-Power Embedded Devices}, url={http://dx.doi.org/10.1109/icist59754.2023.10367148}, DOI={10.1109/icist59754.2023.10367148}, abstractNote={The utilization of deep learning (DL) in medical research and industry has witnessed substantial growth in recent years. A pivotal application involves employing DL for dermatology image classification tasks. However, the major challenges in such tasks, in terms of the scarcity and bias of high-quality labeled data, significantly hinder further advancement in this domain. Such data insufficiency gives rise to concerns regarding accuracy disparities across different demographic groups, which may ultimately lead to unfair outcomes. Additionally, complex and effective DL models are often unsuitable with low-power embedded devices, which hinders their usability in resource-limited environments. In this paper, we propose a DL framework to address these issues. Our major approach involves augmenting data with Gaussian white noise to generate synthetic data samples and employing knowledge distillation techniques to transfer valuable knowledge from a larger and more complex model to a smaller and more efficient counterpart. Through comprehensive experimentation on an open-access skin disease classification dataset, we demonstrate that our proposed framework significantly enhances the performance of DL models on low-power embedded devices, thereby optimizing the trade-offs among overall accuracy, fairness for different demographic groups, and inference latency on low-power embedded devices.11Our code and experiments can be reproduced by utilizing the details provided in the Methodology section on image preprocessing and augmentation, model architecture, and training configurations.(https://github.com/yixinli19/Dermatology-image-classification). The MobileNet V3 is available at(https://pytorch.org/vision/stable/models/mobilenetv3.html). The Swin Transformer V2 is available at (http-s://pytorch.org/vision/stable/models/swin_transformer.html). The dermatology image dataset is available upon request from ESFair 2023. They can be reached at: https://esfair2023.github.io/ESFair/.}, booktitle={2023 13th International Conference on Information Science and Technology (ICIST)}, author={Li, Yixin and Sui, Ning and Guo, Chengan and Guo, Zhishan}, year={2023}, month={Dec} } @article{kumar_wang_zhou_dillard_li_sciandra_sui_zentella_zahn_shabanowitz_et al._2023, title={Structure and dynamics of the Arabidopsis O-fucosyltransferase SPINDLY}, url={http://dx.doi.org/10.1038/s41467-023-37279-1}, DOI={10.1038/s41467-023-37279-1}, abstractNote={Abstract}, journal={Nature Communications}, author={Kumar, Shivesh and Wang, Yan and Zhou, Ye and Dillard, Lucas and Li, Fay-Wei and Sciandra, Carly A. and Sui, Ning and Zentella, Rodolfo and Zahn, Emily and Shabanowitz, Jeffrey and et al.}, year={2023}, month={Mar} } @misc{farhangi_sui_hua_bai_huang_guo_2022, title={Protoformer: Embedding Prototypes for Transformers}, ISBN={9783031059322 9783031059339}, ISSN={0302-9743 1611-3349}, url={http://dx.doi.org/10.1007/978-3-031-05933-9_35}, DOI={10.1007/978-3-031-05933-9_35}, abstractNote={Transformers have been widely applied in text classification. Unfortunately, real-world data contain anomalies and noisy labels that cause challenges for state-of-art Transformers. This paper proposes Protoformer, a novel self-learning framework for Transformers that can leverage problematic samples for text classification. Protoformer features a selection mechanism for embedding samples that allows us to efficiently extract and utilize anomalies prototypes and difficult class prototypes. We demonstrated such capabilities on datasets with diverse textual structures (e.g., Twitter, IMDB, ArXiv). We also applied the framework to several models. The results indicate that Protoformer can improve current Transformers in various empirical settings.}, journal={Advances in Knowledge Discovery and Data Mining}, publisher={Springer International Publishing}, author={Farhangi, Ashkan and Sui, Ning and Hua, Nan and Bai, Haiyan and Huang, Arthur and Guo, Zhishan}, year={2022}, pages={447–458} } @article{zentella_sui_barnhill_hsieh_hu_shabanowitz_boyce_olszewski_zhou_hunt_et al._2017, title={The Arabidopsis O-fucosyltransferase SPINDLY activates nuclear growth repressor DELLA}, url={http://dx.doi.org/10.1038/nchembio.2320}, DOI={10.1038/nchembio.2320}, abstractNote={Mass spectrometry analysis combined with in vitro assays reveals that SPINDLY is an O-fucosyltransferase that modifies the growth repressor DELLA and consequently enhances its activity to regulate transcription of target genes. Plant development requires coordination among complex signaling networks to enhance the plant's adaptation to changing environments. DELLAs, transcription regulators originally identified as repressors of phytohormone gibberellin signaling, play a central role in integrating multiple signaling activities via direct protein interactions with key transcription factors. Here, we found that DELLA is mono-O-fucosylated by the novel O-fucosyltransferase SPINDLY (SPY) in Arabidopsis thaliana. O-fucosylation activates DELLA by promoting its interaction with key regulators in brassinosteroid- and light-signaling pathways, including BRASSINAZOLE-RESISTANT1 (BZR1), PHYTOCHROME-INTERACTING-FACTOR3 (PIF3) and PIF4. Moreover, spy mutants displayed elevated responses to gibberellin and brassinosteroid, and increased expression of common target genes of DELLAs, BZR1 and PIFs. Our study revealed that SPY-dependent protein O-fucosylation plays a key role in regulating plant development. This finding may have broader importance because SPY orthologs are conserved in prokaryotes and eukaryotes, thus suggesting that intracellular O-fucosylation may regulate a wide range of biological processes in diverse organisms.}, journal={Nature Chemical Biology}, author={Zentella, Rodolfo and Sui, Ning and Barnhill, Benjamin and Hsieh, Wen-Ping and Hu, Jianhong and Shabanowitz, Jeffrey and Boyce, Michael and Olszewski, Neil E and Zhou, Pei and Hunt, Donald F and et al.}, year={2017}, month={May} } @article{li_kang_sui_liu_2012, title={ROP11 GTPase is a Negative Regulator of Multiple ABA Responses in Arabidopsis}, url={http://dx.doi.org/10.1111/j.1744-7909.2012.01100.x}, DOI={10.1111/j.1744-7909.2012.01100.x}, abstractNote={Abstract}, journal={Journal of Integrative Plant Biology}, author={Li, Zixing and Kang, Jun and Sui, Ning and Liu, Dong}, year={2012}, month={Mar} } @article{xu_sui_tang_xie_lai_liu_2010, title={One‐step, zero‐background ligation‐independent cloning intron‐containing hairpin RNA constructs for RNAi in plants}, url={http://dx.doi.org/10.1111/j.1469-8137.2010.03253.x}, DOI={10.1111/j.1469-8137.2010.03253.x}, abstractNote={*The hairpin-based RNA interference (RNAi) technique plays an important role in exploring gene function in plants. Although there are several methods for making hairpin RNA (hpRNA) constructs, these methods usually need multiple relatively laborious, time-consuming or high-cost cloning steps. Here we describe a one-step, zero-background ligation-independent cloning (OZ-LIC) method for making intron-containing hpRNA (ihpRNA) constructs by our vector pRNAi-LIC. *To generate the ihpRNA constructs with zero-background, this method only requires treating two PCR products of target gene flanked with different LIC sequences and SmaI-linearized pRNAi-LIC vector by T4 DNA polymerase respectively, and then transforming these treated DNA mixture into Escherichia coli. *The ihpRNA constructs generated with our OZ-LIC RNAi vector can efficiently induce not only transient silencing of the exogenous marker genes and the endogenous resistance-related Nicotiana benthamiana SGT1 gene, but also stable transgenic suppression of Arabidopsis SGT1b gene. *Our new OZ-LIC method and RNAi vector will represent a powerful tool for gene knockdown in plants and may facilitate high-throughput determination of plant gene function.}, journal={New Phytologist}, author={Xu, Guoyong and Sui, Ning and Tang, Yang and Xie, Ke and Lai, Yizhen and Liu, Yule}, year={2010}, month={Jul} }