While modern text-to-image models excel at prompt-based generation, they often lack the fine-grained control necessary for specific user requirements like spatial layouts or subject appearances. Multi-condition control addresses this, yet its integration into Diffusion Transformers (DiTs) is bottlenecked by the conventional “concatenate-and-attend” strategy, which suffers from quadratic computational and memory overhead as the number of conditions scales. Our analysis reveals that much of this cross-modal interaction is spatially or semantically redundant. To this end, we propose Position-aligned and Keyword-scoped Attention (PKA), a highly efficient framework designed to eliminate these redundancies. Specifically, Position-Aligned Attention (PAA) linearizes spatial control by enforcing localized patch alignment, while Keyword-Scoped Attention (KSA) prunes irrelevant subject-driven interactions via semantic-aware masking. To facilitate efficient learning, we further introduce a Conditional Sensitivity-Aware Sampling (CSAS) strategy that reweights the training objective towards critical denoising phases, drastically accelerating convergence and enhancing conditional fidelity. Empirically, PKA delivers a 10.0× inference speedup and a 5.1× VRAM saving, providing a scalable and resource-friendly solution for high-fidelity multi-conditioned generation.

Although pretrained joint audio-visual diffusion models offer rich control over what to generate, they provide no explicit control over when an utterance should occur. To address this, we study inference-time speech scheduling, a novel task that places coupled speech and visual articulation within user-specified begin–end intervals without finetuning the backbone model. We uncover two intrinsic properties of the denoising process that enable this task. First, a timing-sensitive text-to-audio cross-attention head exposes each utterance's model-implied source span along the latent timeline. Second, the predicted clean latent already organizes coupled speech and visual articulation, allowing their temporal placement to be edited without regenerating the content. Building on these discoveries, we propose TimeSteer, a training-free framework that localizes each utterance's source span through Source Span Localization and transfers the associated audio-visual latent content from the source interval to the specified target interval through Region-Aware Latent Remapping. We further introduce SpeechShift, the first benchmark for interval-level speech scheduling in joint audio-visual generation. Experiments across two representative backbones show that TimeSteer substantially improves interval controllability over training-free baselines while maintaining competitive overall generation quality.

This article examines the transformation of urban referencing in Taipei City after democratization, using policy mobility theories and case study research methods to 'follow policy changes' over three decades. It argues that geopolitical considerations in the 1990s prompted Taipei to adopt the global city discourse as a political strategy, subsequently leading to the implementation of neoliberal urban policies. These policies encouraged property-led development and housing speculation, exacerbating housing affordability issues. Amidst these challenges, a progressive bottom-up social housing movement emerged in 2010, drawing inspiration from the Dutch model. However, this process of policy mobility is selective, with Amsterdam serving more as an inspirational benchmark than a direct model for replication. Neoliberalism hinders both Dutch and Taiwanese social housing policies from achieving housing justice. This case study of Taiwan's post-democratization urban development will also contribute to the literature on urban developmentalism.

idea of "becoming a global city" has strongly influenced the urban policies in Taipei, since the 1990s. The Taipei City Government has implemented several mega projects in the city, claiming to improve Taipei's global status, such as building the highest building in the world and creating a new financial district. Meanwhile, the squatter settlements, which used to be a part of Taipei's landscape after 1949, have rapidly disappeared and are displaced by luxury buildings and parks. Globalizing Taipei has become the fertile ground of housing speculation and has led to serious problems with housing affordability. Recently, the post-2005 housing boom has triggered a strong social rental housing movement. This article will first examine how "global city discourse" has influenced the urban projects in Taipei, since the 1990s. Then, it will explore the status of low-income housing in Taipei's urban policies. This article will draw on several theoretical concepts, including policy mobility, global cities, the right to the city, neoliberalization in East Asian cities, and worlding cities, to discuss the problem of low-income housing in globalizing Taipei.

The literature on East Asian developmental states has mostly focused on economic policies; few studies have investigated how developmentalism affects urbanisation. This study addresses this gap by analysing the formation of a housing system by the developmental state and the role of the housing sector in the transition from an economic miracle in the 1960s to stagnation after the 1990s. Wang, Lee, and Chen (2017) argued that Taiwan’s state bureaucrats played a gradually diminishing role in directing the market. This article takes a different approach by arguing that the government’s interventions on housing intensified after the 1990s; housing finance sectors were formalised, squatters were removed, and interventions based on market mechanisms were employed to stimulate the housing market. However, the lack of an effective state intervention on housing made speculation difficult to control and created an obstacle to economic development. Such housing speculation increased social inequality.

Deep learning-based techniques for facial anonymization strive to protect identity details while preserving image usability. We present Antiano, the first work to reveal possible security risks associated with deep learning-based facial anonymization algorithms. By evaluating the entire anonymization process, we introduce a framework comprising four attack types: Anonymization Deactivation Attack, ID Destruction Attack, Downstream Task Attack, and Attribute Attack. These attacks aim to undermine anonymization effectiveness, mislead anonymized outputs and downstream tasks, or modify facial attributes in subtle ways. For each scenario, we develop novel strategies that craft perturbations tailored to a user’s face and transferable across different keys and models. Experimental findings reveal that Antiano achieves significant attack effectiveness against various reversible anonymization models, highlighting security vulnerabilities in current deep learning anonymization algorithms. This paper also distinguishes Antiano from traditional adversarial attacks and provides security recommendations for the future development of facial anonymization technologies.

The transfer of development rights (TDR) has been widely used in the preservation of historic districts. The Dihua Street TDR (DS-TDR) in Taipei, Taiwan, successfully preserves the exteriors of historic buildings and traditional landscape in Dihua Street, without significant displacement of previous residents or increases in rents. This study describes this process as 'alternative gentrification', which facilitates the coexistence of traditional and new industries in historic districts, unlike typical gentrification in other cities. Although new shops gradually replace existing shops, the rent level remains relatively affordable compared with other shopping streets in the Taipei city centre. These aspects enable the coexistence of a clustering of new creative-industrial stores and existing stores within the buildings restored and landscaped by the DS-TDR.

Generative Adversarial Networks (GAN) is a popular machine learning method that possesses powerful image generation ability, which is useful for different multimedia applications (e.g., photographic filters, image editing). However, typical GAN models have a large memory footprint that limits their practical applications for resource-constrained devices (e.g., smartphones). To deploy GAN models on devices with various hardware constraints, we propose our method, AdjustableGAN, which can compress a pretrained GAN model to different compression ratios. Our method compresses GAN by performing filter-wise pruning that follows these objectives: (1) deactivate convolutional filters for minimal performance decrease, (2) reactivate convolutional filters for maximal performance increase. We implement multiple Genetic Algorithms (GA) to perform each of these objectives— Downsize GA for best filter deactivations, while Upsize GA searches for best filter reactivations. By selective utilization of Upsize/Downsize GA, we could explicitly control the compression ratio of the model. For finalization, we fine-tune the compressed output model using the training dataset of the original input model. Our experimental results show that our method can reliably compress generative networks with minimal accuracy drop compared to other state-of-the-art compression algorithms.

In cloud storage applications, data owners' original images are usually encrypted before being outsourced to the cloud for preserving data owners' privacy. However, in deep learning model-based image encryption methods, an adversary can conduct the model extraction attack to reveal the model parameters and thus restore the privacy information by obtaining numerous encrypted images. In this paper, we propose an image translation-based deniable encryption (ITDE) scheme to achieve encryption deniability and defend against model extraction attacks. Differing from traditional encryption methods in which encrypted images are visually meaningless, ITDE applies image translation to generate encrypted images in the form of human faces. Moreover, ITDE provides deniability for data owners to keep the encryption parameters private. To defend against model extraction attacks, the defense mechanism is introduced in our proposed ITDE to preserve deep learning models. Experimental results demonstrate the superiority of our proposed methods in terms of encryption deniability and privacy preservation.

Using miniscope recordings of calcium fluorescence signals in the CA1 region of the hippocampus of mice, we monitor the neural activity of hippocampal regions while the animals are freely moving in an open chamber. Using a data-driven statistical modeling approach, the statistical properties of the recorded data are mapped to spin-glass models with pairwise interactions. Considering the parameter space of the model, the observed system is generally near a critical state between two distinct phases. The close proximity to the criticality is found to be robust against different ways of sampling and segmentation of the measured data. By independently altering the coupling distribution and the network structure of the statistical model, the network structures are found to be vital to maintain the proximity to the critical state. We further find the observed assignment of the coupling strengths makes the net coupling at each site more balanced with slight variation, which likely helps the maintenance of the critical state. Network analysis on the connectivity obtained by thresholding the coupling strengths find the connectivity of the networks to be well described by a random network model. These results are consistent across different experiments, sampling and segmentation choices in our analysis. A new result of our analysis is that the proximity to critical state and all the network properties are largely maintained even if random subsamples with a fraction of neurons are selected from the dataset as long as the number of neurons is more than 30 to 40. Thus the relevant degrees of freedom of CA1 region in the collective state we studied is not as large as one would expect.

This article deals with the path-dependent features of financialization of housing in Taiwan, an East Asian developmental state. The levels of foreign capital and securitization in Taiwan's housing market remain relatively low, meaning domestic capital, of which there is an abundance, is the major financial source of such speculation. The process does not include the retrenchment of the welfare state, because Taiwan has been a homeowner society. After financial liberalization in the 1980s, Taiwan's state intervention in the housing and financial sectors has actually intensified via the enactment of more regulations to decrease the role of the informal financial and housing sectors. As a result of neoliberalization giving precedence to market mechanisms, various low-interest mortgage programs in the 1990s, all subsidized by public funding, have increased the rate of homeownership and sustained housing prices. Even though this varied the development of housing financialization, housing in Taiwan has largely become a tool of speculation, and housing affordability has become a serious problem.

Despite the wide applications of dry anaerobic digestion (AD), a number of fundamental issues, such as composition-oriented estimation of biogas production and CO2 reduction potential, were not well understood yet. The objective of this study was to establish composition-oriented models for prediction of biogas production and the associated shift of microbial communities. Three important factors regarding feedstock, including loading, carbon-to-nitrogen ratio, and solid-to-liquid ratio, were found to significantly affect the biogas production. The biogas production and digestion kinetics were evaluated with the response surface methodology. The major contribution to biogas production was found to be hydrogenotrophic methanogens (82.6 +/- 0.4%). The net CO2 reduction potential was assessed from the life-cycle approach, and a substantial amount of CO2 generation (i.e., 2.8-6.7 tonne/tonne-VS) could be reduced by AD, compared to incineration, revealing that dry AD for food waste treatment should be one of the essential practices in the portfolio of global CO2 mitigation.

From DeepBlue to AlphaGo, computer game is the drosophila of Artificial Intelligence. For the AI services of assisting human learning, we believe computer board games can also play the role of the drosophila. From the viewpoint of social development, after the rise of AI, human need more ability of logical thinking and judgement than before. Advocating computer games is an excellent tool for the training of logical concepts and hence produces positive impact in our society. Since human Go players have a different process of reasoning compared to Go programs today, we need to develop learning methods that more closely match how humans think. Deep Learning takes inspiration from human cognitive processes and is similar to human intuition. As a result, Go programs developed with Deep Learning generate plays that feel more human. We use Deep Learning and Reinforcement learning to develop scaffolding learning system for Go. The system contains human-like Go programs with various strengths, which allows novice players to learn the game progressively. We also introduced a simplified variant of Go, named Jungo. The game could help the beginners to learning the game of Go.

由于情感语料问题、情感与声学特征之间关联问题、语音情感识别建模问题等因素,语音情感识别一直充满挑战性.针对传统基于上下文的语音情感识别系统仅局限于特征层造成标签层上下文细节丢失以及两层级差异性被忽略的缺陷,本文提出嵌入注意力机制并结合层级上下文学习的双向长短时记忆(BLSTM)网络模型.模型分3个阶段完成语音情感识别任务,第1阶段提取情感语音特征全集后采用SVM-RFE特征排序算法降维得到最优特征子集,并对其进行注意力加权;第2阶段将加权后的特征子集输入BLSTM网络学习特征层上下文获得最初情感预测结果;第3阶段利用情感标签值对另一独立BLSTM网络训练学习标签层上下文信息并据此在第2阶段输出结果基础上完成最终预测.模型嵌入注意力机制使其自动学习调整对输入特征子集的关注度,引入标签层上下文使其联合特征层上下文实现层级上下文信息融合提高鲁棒性,提升了模型对情感语音的建模能力,在SEMAINE和RECOLA数据集上实验结果表明:与基线模型相比RMSE和CCC均得到较好改善.

Since the late 1980s, the process of democratization and increased public participation has pressed for the expansion of social welfare in Taiwan, while neoliberalization has affected housing policies to enhance the operation of market mechanisms for housing provision. Nowadays, the Taiwanese state primarily facilitates the growth of housing market and homeownership, exercising little control over speculation. Escalating housing prices have led to a strong social rental housing movement. This chapter explores the diminishing role of the state in Taiwan's housing system and how housing has been understood by the state, placing the discussions in the political and economic contexts after 1949. The chapter also examines how the social housing movement since 2010 has gradually transformed the role of the state in its provision of housing and what obstacles the movement has to confront.

Considering Asian cities, this collection discusses the socio-political processes of how neoliberalization entwines with local political economies and legacies of ‘developmental’ or ‘socialist’ statism to produce urban contestations centered on housing.

This paper proposes a novel AC-DC power supply with features of power-factor correction (PFC) and cost-effectiveness for indoor LED lighting applications. The proposed AC-DC power supply combines an inverse buck-boost converter with a lossless snubber, and also includes input-current shaping. An 1SW(60V/0.3A)-rated prototype power supply has been successfully developed and implemented for operating with an input utility-line voltage of 110 V. Satisfactory results have demonstrated the functionality of the presented AC-DC power supply and its suitability for indoor LED lighting applications.

The chapter discusses how important housing questions are for Asian cities and citizens in the contemporary era of neoliberalization. It highlights the importance of understanding the path-dependent nature of neoliberalization, especially in the context of Asia’s condensed development under the developmentalist states. It also shows how housing has been at the heart of urban contestations in Asian cities, especially given its significance for wealth generation and upward class mobility when the formation of middle classes would have been the state’s legitimacy project. The chapter concludes by providing a summary of each contributing chapter.

Considering Asian cities ranging from Taipei, Hong Kong and Bangkok to Hanoi, Nanjing and Seoul, this collection discusses the socio-political processes of how neoliberalization entwines with local political economies and legacies of ‘developmental’ or ‘socialist’ statism to produce urban contestations centered on housing. The book takes housing as a key entry point, given its prime position in the making of social and economic policies as well as the political legitimacy of Asian states. It examines urban policies related to housing in Asian economies in order to explore their continuing alterations and mutations, as they come into conflict and coalesce with neoliberal policies. In discussing the experience of each city, it takes into consideration the variegated relations between the state, the market and the society, and explores how the global pressure of neoliberalization has manifested in each country and has influenced the shaping of national housing questions.