DSpace community: 資訊工程研究所
http://ir.lib.ncu.edu.tw/handle/987654321/13
The community's search engineSearch the Channels
http://ir.lib.ncu.edu.tw/simple-search
基於深度學習之地震以及海平面溫度預測模型;Deep Learning-Based Models for Earthquake and Sea Surface Temperature Prediction
http://ir.lib.ncu.edu.tw/handle/987654321/93557
title: 基於深度學習之地震以及海平面溫度預測模型;Deep Learning-Based Models for Earthquake and Sea Surface Temperature Prediction abstract: 地震以及海平面溫度的預測對於地球科學和氣象研究至關重要。本研究提出了兩個獨立的預測模型,分別針對地震電離層前兆和海平面溫度,其中地震電離層前兆預測模型結合了深度神經網路(DNN)以及長短期記憶(LSTM)網路的比較,而海平面溫度預測則僅使用了LSTM網路。
在地震電離層前兆預測方面,本論文研究了深度神經網路(DNN)和LSTM網路的效能。透過比較這兩種模型的結果,能夠深入了解它們在捕捉地震電離層前兆模式和趨勢方面的優勢。實驗結果顯示,LSTM網路在地震電離層前兆預測中表現出色,相對於DNN模型有更好的泛化能力,特別是對於時間序列數據的建模。
在海平面溫度預測方面,本論文專注於LSTM網路的應用。這種網路的適應性和長期記憶特性使其成為捕捉溫度變化的理想工具。本文通過大量實驗證明,LSTM模型能夠有效地捕捉海平面溫度的季節性和趨勢,並在預測中表現出色。
總體而言,本研究提供了一個綜合性的地震電離層前兆和海平面溫度預測模型,結合了LSTM網路的優勢。
;The prediction of earthquakes and sea surface temperatures is crucial for Earth science and meteorological research. This study introduces two independent predictive models, focusing on earthquake and sea surface temperature forecasts. The earthquake prediction model integrates a comparison between Deep Neural Network (DNN) and Long Short-Term Memory (LSTM) network results, while the sea surface temperature prediction model exclusively utilizes the LSTM network.
In terms of earthquake prediction, we investigate the performance of Deep Neural Network (DNN) and LSTM network. By comparing the results of these two models, we gain insights into their advantages in capturing earthquake patterns and trends. Experimental results demonstrate that the LSTM network excels in earthquake prediction, exhibiting better generalization capabilities compared to the DNN model, particularly in modeling time-series data.
For sea surface temperature prediction, we focus on the application of the LSTM network. The adaptability and long-term memory characteristics of this network make it an ideal tool for capturing temperature variations. Through extensive experiments, we validate that the LSTM model effectively captures the seasonality and trends in sea surface temperatures, demonstrating outstanding performance in prediction.
In summary, this study provides a comprehensive earthquake and sea surface temperature prediction model, leveraging the advantages of the LSTM network.
<br>EAVPFunc:融合蛋白質語言模型的集成框架用於識別抗病毒肽及其功能分類;EAVPFunc: An Ensemble-Based Framework with the Protein Language Model for Antiviral Peptides Identification and Functional Classification
http://ir.lib.ncu.edu.tw/handle/987654321/93556
title: EAVPFunc:融合蛋白質語言模型的集成框架用於識別抗病毒肽及其功能分類;EAVPFunc: An Ensemble-Based Framework with the Protein Language Model for Antiviral Peptides Identification and Functional Classification abstract: 在21世紀,病毒性疾病的爆發對人類社會造成了重大影響。抗病毒肽(AVPs)作為對抗新興病毒疾病如SARS-CoV-2以及 HIV 和 HCV 等抗藥性菌株的重要治療藥物。然而,對於抗病毒肽的功能分類研究有限,以及不同病毒家族和物種之間的數據不均衡,對該領域構成了挑戰。為了克服這些挑戰,本研究引入了一個名為EAVPFunc的新型雙階段分類模型,旨在揭示抗病毒肽的功能特性。在第一階段,EAVPFunc將抗病毒肽從廣泛的肽譜中區分出來,將其與非抗微生物和非抗病毒的抗微生物肽區分。第二階段,EAVPFunc將抗病毒肽與特定病毒科和個別病毒進行精確對應。EAVPFunc結合了隨機森林演算法和卷積神經網絡,在一個集成模型中使用手工特徵和先進的蛋白質語言模型來提高解釋性和預測準確性。這種方法在兩個不同的數據集上達到了94.35%和99.46%的高準確率,超越了現有分類器在準確性和均衡分類任務上的表現。總之,我們提出EAVPFunc作為一個穩定且均衡的分類框架,代表了生物信息學方面的重大進步。;Viral outbreaks have had a significant impact on human society in the 21st century. Antiviral peptides (AVPs) are crucial therapeutic agents in the fight against emerging viral diseases such as SARS-CoV-2 and drug-resistant strains such as HIV and HCV. However, the limited research in functional classification poses a challenge for the field, along with data imbalance across different viral families and species. To overcome these challenges, the research introduces a new two-stage classification model named EAVPFunc, which aims to reveal the functional properties of AVPs. In the first stage, AVPs are distinguished from a wider range of peptides, including non-antimicrobial and non-antiviral peptides. In the second stage, EAVPFunc associates AVPs with specific virus families and individual viruses. EAVPFunc combines the Random Forest algorithm with Convolutional Neural Networks in an ensemble model, using handcrafted features and an advanced protein language model to improve interpretability and prediction accuracy. This approach resulted in high accuracy rates of 94.35% and 99.46% on two different datasets, outperforming existing classifiers in accuracy and balanced classification tasks. In conclusion, we propose EAVPFunc as a stable and balanced classification framework, which represents a significant advancement in bioinformatics.
<br>基於梯度的重構攻擊在隱私權保護聯合學習中的評估方法初探;A Preliminary Study on Evaluation Methods of Gradient-based Reconstruction Attacks in Privacy-Preserving Federated Learning
http://ir.lib.ncu.edu.tw/handle/987654321/93555
title: 基於梯度的重構攻擊在隱私權保護聯合學習中的評估方法初探;A Preliminary Study on Evaluation Methods of Gradient-based Reconstruction Attacks in Privacy-Preserving Federated Learning abstract: 在聯合學習(FL)中,參與者的模型更新可能會對隱私造成破壞性威脅,透過巧妙地充分利用共享更新,攻擊者可以重建參與者的訓練隱私數據,達到像素級別。 差分隱私(DP)作為資料匿名化的標準,就是為了應對這種新出現的威脅而提出的;在這種經過差分隱私改進的隱私保護FL(PPFL)設定中,傳輸的資訊會經過淨化(即 經過因子剪切和噪音擾動),以保護相關方的隱私。 儘管 DP 最初是用於集中學習和表格數據,但最近,它在處理多媒體數據(尤其是圖像)的 FL 方面獲得了越來越多的關注。
基於梯度的重構攻擊通常利用峰值信噪比(PSNR)、結構相似性指數(SSIM)和感知影像補丁相似性(LPIPS)等感知相似性指標作為主要評估方法,以暗示感知相似性與隱私洩露 之間的相關性。 基於深度神經網路(如AlexNet 和VGG)發明的Learned(LPIPS)等感知度量是為了模仿人類的感知,其設計目的是讓度量能夠捕捉兩張圖片之間細微的感知相似性和差異性,並解決 PSNR 和SSIM 等傳統測量無法超越影像像素值的問題。
然而,由於感知度量是建立在人的感知基礎上的,因此重構攻擊過程中造成的難以察覺的細微差別和損壞是否會影響這些度量還不得而知。 因此,作者認為這可能是需要填補的空白。
總而言之,據作者所知,在評估使用影像資料的聯邦學習框架的隱私洩漏時,對感知指標進行全面分析,以及隱私保護技術DP 在保護這種設定免受基於梯度的重構攻擊方面的效果如何 ,仍然是前所未聞的。
為此,本論文旨在研究1. 2.一種新型隱私評估方法的可行性,該方法可揭示SOTA 重構攻擊評估方法中廣泛使用的感知度量LPIPS 與PPFL 中分類任務準確性之間的關係 ;3.差分隱私保護技術對上述SOTA 基於梯度的重構攻擊的有效性。;In Federated Learning (FL), a participant’s model update can potentially be a devastating threat to privacy, by cleverly making full use of the shared updates, it is believed that an attacker can reconstruct the participant’s training private data to a pixel-level. Differential Privacy (DP), the norm in data anonymization, was proposed to deal with this emergent threat; in such a DP-fied Privacy-preserving FL (PPFL) setup, the transmitted information is sanitized (i.e. clipped by a factor and perturbed by noise) to protect the privacy of the parties involved. Though was originally intended to be used with centralized learning and tabular data, recently, DP has gained more and more attention in FL with multimedia data, especially images.
Gradient-based reconstruction attacks typically utilized perceptual similarity metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Perceptual Image Patch Similarity (LPIPS) as the main evaluation method to imply the correlation between perceptual similarity and privacy leakage. Perceptual metrics such as Learned (LPIPS) were invented to mimic human perception, based on deep neural networks (such as AlexNet and VGG), the design is intended to allow the metric to capture the subtle perceptual similarity and differences between 2 pictures, and solve the incapability to look beyond the image pixel’s value of the traditional metrics like PSNR and SSIM.
However, since the perceptual metrics are built upon human perception, it is unknown whether the imperceptible nuances and corruptions caused by the reconstruction attack process could influence those metrics. Therefore, the author sees this could potentially be a gap that needs to be filled.
To summarize, according to the author′s best knowledge, a comprehensive analysis of perceptual metrics in evaluating privacy leakages of a Federated Learning framework with image data, and how effectively the privacy-preserving technique DP works in protecting such a setting against gradient-based reconstruction attacks is still unheard of.
For that matter, this dissertation is intended to study: 1. The reliability of perceptual metrics, which are employed by reconstruction attacks literature in a realistic Federated Learning framework; 2. The feasibility of a novel privacy evaluation method that can reveal the relationship between the widely used perceptual metric LPIPS in the SOTA reconstruction attack′s evaluation method and the accuracy of a classification task in PPFL; 3. The effectiveness of differential privacy against the aforementioned SOTA gradient-based reconstruction attack.
<br>無需人工標記資料的中國古代文獻事件對齊模型:以《清實錄》及《滿文老檔》為例;An Event Alignment Model for Ancient Chinese Literature without Requirement of Manually Labeled Data: A Case Study of the Qing Shi-Lu and Manchu Old Archives
http://ir.lib.ncu.edu.tw/handle/987654321/93554
title: 無需人工標記資料的中國古代文獻事件對齊模型:以《清實錄》及《滿文老檔》為例;An Event Alignment Model for Ancient Chinese Literature without Requirement of Manually Labeled Data: A Case Study of the Qing Shi-Lu and Manchu Old Archives abstract: 在應用於中國古代文獻的數位人文領域中,已有些研究探討如何 實現文本對齊技術來幫助歷史學者比較不同的文獻,不過這些研究並 沒有以「相同語意」的觀點來對齊文本。故本研究將引入自然語言處 理中釋義識別任務的概念,來找出不同文本中擁有相同語意的段落, 並應用於後漢書、三國志和資治通鑑以作為範例。然而如果要採用釋 義識別任務中最先進的自然語言處理技術,則會有一些限制需要去考 量:(1)訓練資料不足(2)基於注意力方法的文本長度限制。為了 解決這些問題,本研究提出了應用二階段訓練於中國古代文獻釋義識 別的弱監督學習架構(SPITAC)。此方法有兩個主要部分:偽標籤訓 練集生成和二階段訓練。在偽標籤訓練集生成中,本研究使用基於規 則的方法來自動產生訓練資料集以解決訓練資料不足的問題。而為了 解決文本長度限制,則採用句子過濾器的方法來刪減不重要的句子, 將句子長度縮減到最大長度的範圍內。在二階段訓練的設計中,此方 法可以使分類器更好的識別出硬負樣本來提升模型性能。從實驗結果 表明,本研究的弱監督學習方法可以達到接近監督式學習的效果,而 在消融實驗中,句子過濾器和二階段訓練可以有效提升性能,能提高 4.14 F1 分數並超越基線模型。最後本研究將從實際的文本中演示並分 析此方法的成果,並從成效中探討這項任務的困難及未來改進方向。;Implementing text alignment on ancient Chinese literature offers signif- icant assistance to academics investigating historical events, particularly as variations may occur in the descriptions of an event across different texts. These variations represent valuable research materials. However, the current studies rarely align text from the perspective of the ”same event”. In order to develop a tool that better aligns with the practical application conditions of text alignment in ancient Chinese literature, we adopted the predecessors’ ideas. We have redefined the ”Paraphrase” definition of Paraphrase Identi- fication task (a Natural Language Processing task determining whether two texts convey the same meaning) to facilitate the task of text alignment for ancient Chinese literature.
This work encounters two primary challenges: 1) the deficiency of train- ing data and 2) the limitations in input length of the attention-based method. To address these issues, we proposed the Event Alignment Model for Ancient Chinese Literature without Requirement of Manually Labeled Data. In this framework, we utilize ChatGPT to generate a training set, thereby overcom- ing the lack of training data. Furthermore, we resolve the issue of text length limitation by employing a data slicing method to reduce paragraph size within a maximum length. Additionally, the GujiBERT model is also implemented for paraphrase identification. Experimental results show that our proposed EAMAC outperforms significantly more than the baseline and exhibits con- siderable stability and applicability when applied to other texts.
<br>