Generative AI (AI) has become ubiquitous in both daily and professional life, with emerging research demonstrating its potential as a tool for accessibility. Neurodivergent people, often left out by existing accessibility technologies, develop their own ways of navigating normative expectations. GAI offers new opportunities for access, but it is important to understand how neurodivergent "power users"-successful early adopters-engage with it and the challenges they face. Further, we must understand how marginalization and intersectional identities influence their interactions with GAI. Our autoethnography, enhanced by privacy-preserving GAI-based diaries and interviews, reveals the intricacies of using GAI to navigate normative environments and expectations. Our findings demonstrate how GAI can both support and complicate tasks like code-switching, emotional regulation, and accessing information. We show that GAI can help neurodivergent users to reclaim their agency in systems that diminish their autonomy and self-determination. However, challenges such as balancing authentic self-expression with societal conformity, alongside other risks, create barriers to realizing GAI's full potential for accessibility.

In this paper, we present a method for synthesizing optimal policies for stochastic systems under high-level mission specifications while maintaining information security. The stochastic systems are modeled as probabilistic labeling Markov decision processes (PLMDPs), and the high-level mission specifications are represented by linear temporal logic (LTL) formulas, which must be satisfied repeatedly, infinitely often. Meanwhile, an intruder is modeled as an observer who knows the exact structure of the system and passively monitors its behavior, aiming to infer the high-level mission specifications. Our objective is to synthesize an optimal policy that minimizes the mean payoff cost while preventing the intruder from definitively determining the given mission specifications. We begin by extending the PLMDP model by incorporating an observation function and the corresponding probability function, and we introduce the concept of LTL opacity for stochastic systems. Next, by proposing a new structure called the synchronous accepting maximally end component, we identify the subset of the accepting end components that satisfy the LTL opacity requirement. Finally, we develop a modified linear program with information security constraints to synthesize an optimal policy that ensures both formal correctness and LTL opacity.

A hyperspectral image classification method based on probabilistic weighted fusion of multiple spectral-spatial features is proposed in this letter. First, dimensionality reduction and feature extraction of hyperspectral images are conducted by minimum noise fraction. Then, two spectral-spatial features are composed through a combination of texture features and multiscale morphological features with characteristic images obtained by minimum noise fraction. Next, a support vector machine classifier is employed to classify each spectral-spatial feature. Finally, a probabilistic weighted fusion model is established and applied for probabilistic fusion of the classification output of every single feature. The proposed method is validated by classification experiments using Reflective Optics Spectrographic Imaging System (ROSIS) and Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) images, which achieves the overall accuracy of 96.32% and 96.49%, respectively. The results indicate that the accuracy of hyperspectral image classification is improved effectively by the proposed method and the method has superior classification performance to other conventional multifeature fusion methods.