Welding quality directly affects the welding structure's service performance and life. Hence, the effective monitoring welding defects is essential to ensure the quality of the weld structure. Owing to the non-uniformity of the shape, position and size of welding defects, it is a complicated task to analyze and evaluate the acquired welding defects images manually. Fortunately, deep learning has been successfully applied to image analysis and target recognition. However, the use of deep learning to identify welding defects is time-consuming and less accurate due to the lack of adequate training data samples, which easily cause redundancy into the classifier. In this situation, we proposed a new transfer learning model based on MobileNet as a welding defect feature extractor. By using the ImageNet dataset (non-welding defect data) to pre-train a MobileNet model, migrate the MobileNet model to the welding defects classification field. This article suggested a new TL-MobileNet structure by adding a new Full Connection layer (FC-128) and a Softmax classifier into a traditional model called MobileNet. The entire training process of TL-MobileNet model has been successfully optimized by the DropBlock technology and Global average pooling (GAP) method. They can effectively accelerate the convergence rate and improve the classification network generalization. By testing the proposed TL-MobileNet on the welding defects dataset, it turned out our model prediction accuracy has arrived at 97.69%. The experimental results show that in several aspects, TL-MobileNet have better performance than other transfer learning models and traditional neural network methods.
In this study, we investigate multi-scale features extracted from baseline structural magnetic resonance imaging (MRI) for classifying patients with mild cognitive impairment (MCI), who have either converted or not converted to Alzheimer's disease (AD) three years after their baseline visit. Total 549 subjects from the Alzheimer's disease Neuroimaging Initiative (ADNI) database are included, and there are 228 Normal controls (NC), 133 MCI patients (71 of them converted to AD within 3 years, referred as MCI converters, or MCIc) and 188 AD patients. The images are preprocessed with the standard voxel-based morphometry method with segmentation of grey matter, white matter and cerebrospinal fluid. Wavelet frame, a multi-scale image representation approach, is applied to extract features of different scales and directions from the processed grey matter image data. The features are extracted for both whole grey matter images and grey matter images of the hippocampus region. The support vector machine is adopted to construct classifiers for MCIc and MCI non-converters (MCInc). The accuracy using a leave-one-out procedure for classification of AD vs NC and MCIc vs MCInc is 84.13% and 76.69% respectively, both achieved by local hippocampus data. Our study shows that the proposed multi-scale method is capable of discriminating MCI converters and non-converters, and it can be potentially useful for MCI prognosis in clinical applications.
Radiotherapy (RT) is a major treatment method for non-small-cell lung cancer (NSCLC), and development of new treatment modality is now critical to amplify the negative effects of RT on tumors. In this study, we demonstrated a nanoparticle-loaded block copolymer micellar system for cancer hyperthermia treatment (HT) that can be used for synergistic therapy under alternating magnetic field (AMF) and radiation field. Block copolymer micelles (polyethylene glycol-block-polycaprolactone, or PEG-PCL) containing hyaluronic acid (HA) and Mn–Zn ferrite magnetic nanoparticles (MZF) were fabricated via a two-step preparation. HA-modified Mn–Zn ferrite magnetic nanoparticles (MZF-HA) can be enriched in CD44 highly expressing tumor cells, such as A549 (human lung adenocarcinoma cell line), through an active targeting mechanism via receptor–ligand binding of HA and CD44 (HA receptor). MZF can generate thermal energy under an AMF, leading to a local temperature increase to approximately 43 °C at tumor sites for mild HT, and the increased tumor oxygenation can enhance the therapeutic effect of RT. In vitro experiments show that MZF-HA is able to achieve excellent specific targeting performance toward A549 cells with excellent biocompatibility as well as enhanced therapy performance under HT and RT in vitro by apoptosis flow cytometry. In the A549 subcutaneous tumor xenografts model, MRI confirms the enrichment of MZF-HA in tumor, and hypoxia immunohistochemistry analysis (IHC) proved the increased tumor oxygenation after HT. Furthermore, the tumor volume decreases to 49.6% through the combination of HT and RT in comparison with the 58.8% increase of the untreated group. These results suggest that the application of MZF-HA is able to increase the therapeutic effect of RT on A549 and can be used for further clinical NSCLC treatment evaluation.
Although distributed learning has increasingly gained attention in terms of effectively utilizing local devices for data privacy enhancement, recent studies show that publicly shared gradients in the training process can reveal the private training data (gradient leakage) to a third party. However, so far there hasn't been any systematic study of the gradient leakage mechanism of the Transformer based language models. In this paper, as the first attempt, we formulate the gradient attack problem on the Transformer-based language models and propose a gradient attack algorithm, TAG, to recover the local training data. Experimental results on Transformer, TinyBERT 4 , TinyBERT 6 , BERT BASE , and BERT LARGE using GLUE benchmark show that compared with DLG (Zhu et al., 2019), TAG works well on more weight distributions in recovering private training data and achieves 1.5× Recover Rate and 2.5× ROUGE-2 over prior methods without the need of ground truth label. TAG can obtain up to 88.9% tokens and up to 0.93 cosine similarity in token embeddings from private training data by attacking gradients on CoLA dataset. In addition, TAG is stronger than previous approaches on larger models, smaller dictionary size, and smaller input length.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.