Towards deep neural networks robust to adversarial examples by Matyasko, Alexander

“…These adversarially crafted modifications of the input, so-called adversarial examples, are neural network ``blind spots'' and is the main subject of this dissertation. In this thesis, we outline the problem of adversarial examples and show several partial solutions to it. …”
Get full text

Lane-aware image enhancement for lane detection in rain (part B - communications) by Tian, Zichen

“…Based on this dissertation’s work, many valuable improvements can be made in the future. 1) The PyTorch implementation’s main structure is finished, so future research only needs to refine the backpropagation, implement the 2-D Gaussian in Torch tensors. …”
Get full text

A feedforward neural network based indoor-climate control framework for thermal comfort and energy saving in buildings by Chaudhuri, Tanaya, Soh, Yeng Chai, Li, Hua, Xie, Lihua

“…The framework comprises two main components: a thermal-comfort prediction model, and an optimization algorithm termed as the optimal air temperature (OAT) algorithm; they collectively act as an intelligent mediator between the occupant and the ACMV system. …”
Get full text

Topology optimization of support structure for selective laser melting process by Wu, Lingyun

“…As such, this thesis mainly studies the support structure in SLM using topology optimization methods. …”
Get full text

P2P energy trading among isolated microgrids in the remote area by Huang, Rui

“…Generally speaking, the mode of P2P energy trading is a decentralized process, with each MG becoming the main individual. The optimization goal of P2P energy trading in this dissertation is to maximize the profit of the whole MMG system after ensuring the balance of power supply and optimizing the profits of each MG. …”
Get full text

Speeding up deep neural network training with decoupled and analytic learning by Zhuang, Huiping

“…The analytic learning trains neural networks in an exceedingly fast fashion as the training is completed within one single epoch. There are two main challenges in this area. The first challenge lies in the difficulty of finding analytical solutions for multilayer networks. …”
Get full text

Dynamic resource management to defend against advanced persistent threats in fog computing: a game theoretic approach by Feng, Shaohan, Xiong, Zehui, Niyato, Dusit, Wang, Ping

“…The system model under consideration consists of three main entities, i.e., the fog computing provider, attacker, and cyber-insurer. …”
Get full text

Mechanistic insights into the membrane fouling mechanism during ultrafiltration of high-concentration proteins via in-situ electrical impedance spectroscopy (EIS) by Tanudjaja, Henry Jonathan, Ng, Angie Qi Qi, Chew, Jia Wei

“…For 10,000 ppm of BSA, (i) for the PES100 and PSF100 membranes, the Nyquist plots shifted to the left initially, then to the right as the BSA deposition started, but (ii) for the tighter PES10 membrane, the shifts kept going leftwards due to NaCl accumulation out-pacing BSA build-up. The main takeaway for filtering feeds with high BSA concentrations is that higher steady-state flux could be achieved by averting internal fouling, either through using membranes with dense internal structures or lower MWCO.…”
Get full text

Mitigation of membrane biofouling via immobilizing Ag-MOFs on composite membrane surface for extractive membrane bioreactor by Yuan, Guoyu, Tian, Yuxiao, Wang, Bingxin, You, Xiaofei, Liao, Yuan

“…During the long-term EMBR operation, membrane biofouling is an inevitable phenomenon, which is one of the main obstacles impeding its wide applications. The excessive biofilm deposited on membrane surface could significantly reduce the organic mass transfer coefficient of composite membranes by more than 40%. …”
Get full text

Hydroxyapatite nanoparticles-cell interaction: new approaches to disclose the fate of membrane-bound and internalised nanoparticles by Bonany, Mar, Pérez-Berná, Ana Joaquina, Dučić, Tanja, Pereiro, Eva, Martin-Gómez, Helena, Mas-Moruno, Carlos, van Rijt, Sabine, Zhao, Zhitong, Espanol, Montserrat, Ginebra, Maria-Pau

“…Comparison between cells seeded on the discs and those supplemented with the nanoparticles allowed direct interaction of the cell membrane with the material to be ruled out as the main mechanism of toxicity. In addition, standard techniques such as flow cytometry were inconclusive when used to assess nanoparticles toxicity. …”
Get full text

From 2D to 3D: development of monolayer dopaminergic neuronal and midbrain organoid cultures for Parkinson's disease modeling and regenerative therapy by Yeap, Yee Jie, Tng, Teddy J. W., Lee, Mok Jung, Goh, Micaela, Lim, Kah-Leong

“…Here, we have distilled the essence of the different approaches and summarized the main factors driving DA neuronal differentiation, with the view to provide a useful guide to newcomers who are interested in developing iPSC-based models of PD.…”
Get full text

Quantifying uncertainties in tropical cyclone wind hazard assessment due to synthetic track stochastic variability for Southeast Asia by Jian, Wei, Lo, Edmond Yat-Man, Stojanovski, Pane, Pan, Tso-Chien

“…Large uncertainties mainly concentrate in high TC activity zones where the peak gusts from individual synthetic event set can be up to 20% higher than the ensemble outputs. …”
Get full text

Noise matters in image restoration: from synthetic degradation to real-world challenges by Guo, Lanqing

“…Finally, we introduce a diffusion-based method that jointly generates a shadow-free image and refines incorrectly detected shadow masks, effectively addressing noisy annotation problems. The main contributions of this thesis are three folds: We thoroughly study the noise matters that occurred in various real-world degradation scenarios; we propose various approaches to solve different noise matters image restoration problems from different angles; to verify the effectiveness of the proposed methods, we conduct extensive experiments on multiple datasets. …”
Get full text

Transferring a deep learning model from healthy subjects to stroke patients in a motor imagery brain-computer interface by Nagarajan, Aarthy, Robinson, Neethu, Ang, Kai Keng, Chua, Karen Sui Geok, Chew, Effie, Guan, Cuntai

“…Experiments were conducted using Deep ConvNet and state-of-the-art subject-specific machine learning MI classifiers, evaluated on OpenBMI two-class MI-EEG data from healthy subjects and two-class MI versus rest data from stroke patients. Main results. Results of our study indicate that through domain adaptation of a model pre-trained using healthy subjects’ data, an average MI detection accuracy of 71.15% ( ± 12.46 % ) can be achieved across 71 stroke patients. …”
Get full text

Self-supervised Self2Self denoising strategy for OCT speckle reduction with a single noisy image by Ge, Chenkun, Yu, Xiaojun, Yuan, Miao, Fan, Zeming, Chen, Jinna, Shum, Perry Ping, Liu, Linbo

“…Specifically, in this study, the main deep learning architecture is the Self2Self network, with its partial convolution being updated with a gated convolution layer. …”
Get full text

Migrants in between: rural migrant families' practices of settlement in a small but rapidly urbanizing Chinese city by Zhao, Haoyu

“…This study employs a multi-sided ethnography in urban Jiaonan and migrants’ villages of origin to not only investigate rural migrant families’ housing strategies in the city but also trace their relations with rural villages. The main research question is answered from three interrelated dimensions: migrants’ negotiation with changing institutional barriers of accessing housing; the dynamic relations between young migrants’ settlement intention and intra-family relationships; migration patterns and living arrangements across rural-urban divide after owning housing in the city. …”
Get full text

Further studies of extreme learning machine and compressed signal detection by Cao, Jiuwen

“…We first propose a composite function wavelet neural network (WNN) learning with the recent ELM algorithm in Chapter 2. The main contributions of the proposed WNN comparing with traditional ones are using composite functions at the hidden nodes and applying ELM algorithm to WNN as a learning algorithm. …”
Get full text

新马华语量词“粒”的泛化研究 = A case study on the generalization of Chinese noun classifier “li” in Singapore and Malaysia by 司徒奕龙 Situ, Yilong

“…Furthermore, through the comparison between the three Chinese-speaking regions, namely Singapore, Mainland China and Taiwan, we had proven that dialect influence is merely the catalyst for the generalization of li, the main driving force is the analogy of the grammaticalization of the measure word ge(“个”). …”
Get full text

Global response of Vibrio cholerae to predation by heterotrophic protists by Hu, Jie

“…Previous studies show protozoan predation is the main biological factor constraining bacterial growth in the environment. …”
Get full text

聊斋志异的魅力 ：原著小说与《聊斋电视系列片》的比较 = Charm of strange tales from a Chinese studio : a comparison between original novel and 1986 TV series by Pau, Florence Wei Lin, 包玮琳 Pau, Florence Wei Lin

“…From a macro view, I discovered that the TV series and the novel are largely similar in terms of the main characters, plot and ideological content. However, from a micro view, the differences between them are obvious and it shows the author’s own creative spirit.…”
Get full text