Computer vision models can be explained by attributing the output decision to the input pixels. While effective methods for explaining convolutional neural networks have been proposed, these methods often produce low-quality attributions when applied to vision transformers (ViTs). State-of-the-art methods for explaining ViTs capture the flow of patch information using transition matrices. However, we observe that transition matrices alone are not sufficiently expressive to accurately explain ViT models. In this paper, we define a theoretical approach to creating explanations for ViTs called InFlow. The framework models the patch-to-patch information flow using a combination of transition matrices and patch embeddings. Moreover, we define an algebra for updating the transition matrices of series connected components, diverging paths, and converging paths in the ViT model. This algebra allows the InFlow framework to produce high quality attributions which explain ViT decision making. In experimental evaluation on ImageNet, with three models, InFlow outperforms six ViT attribution methods in the standard insertion, deletion, SIC and AIC metrics by up to 18%. Qualitative results demonstrate InFlow produces more relevant and sharper explanations. Code is publicly available at \url{https://github.com/chasewalker26/InFlow-ViT-Explanation.}

Annotated aerial view images are often missing from fine-grained vehicle type classification datasets. This lack of data limits both the accuracy and robustness of models when applied to top-down views, which are essential for applications such as autonomous drones and aerial surveillance. Models trained only on street-level images often fail to generalize to aerial perspectives, requiring more time and multiple observations to recognize vehicles accurately. In contrast, models trained with both street-level and aerial views can perform more reliably and with faster inference in drone-based systems. However, collecting real aerial data at scale can be costly and logistically challenging. In this paper, we propose AVA (Automated Aerial View Augmentation), a framework for aerial data augmentation via 3D asset generation and contextual scene synthesis. Since standalone 3D vehicle models from 2D images are not directly usable for detection, we embed them in realistic backgrounds to enable learning of both object features and scene context. AVA first constructs 3D vehicle models from street-view images. To ensure data quality, we introduce a realism checker that discards incomplete or distorted assets. We then apply geometric transformations to generate aerial 2D views. The 2D views pass through a text-to-video generator that adds background context, mimicking typical drone imagery. We evaluate our data augmentation approach by fine-tuning several object detection backbones. Notably, the pretrained YOLOv11 model, when fine-tuned with AVA augmented data, achieves a significant [email protected] improvement from 0.06 to 0.51 in classifying previously unseen vehicles from aerial perspectives.

This paper introduces EdgeProfiler, a fast profiling framework designed for evaluating lightweight Large Language Models (LLMs) on edge systems. While LLMs offer remarkable capabilities in natural language understanding and generation, their high computational, memory, and power requirements often confine them to cloud environments. EdgeProfiler addresses these challenges by providing a systematic methodology for assessing LLM performance in resource-constrained edge settings. The framework profiles compact LLMs, including TinyLLaMA, Gemma3-1B, LLaMA3-1B, and DeepSeek-R1-1.5B, using aggressive quantization techniques and strict memory constraints. Analytical modeling is used to estimate latency, FLOPs, and energy consumption. The profiling reveals that 4-bit quantization reduces model memory usage by approximately 60–70%, while maintaining accuracy within 2–5% of full-precision baselines. Inference speeds are observed to improve by 2–3× compared to FP16 baselines across various edge devices. Power modeling estimates a 35–50% reduction in energy consumption for INT4 configurations, enabling practical deployment on hardware such as Raspberry Pi 4/5 and Jetson Orin Nano Super. Our findings emphasize the importance of efficient profiling tailored to lightweight LLMs in edge environments, balancing accuracy, energy efficiency, and computational feasibility.

Scaling laws dictate that the performance of AI models is proportional to the amount of available data. Data augmentation is a promising solution to expanding the dataset size. Traditional approaches focused on augmentation using rotation, translation, and resizing. Recent approaches use generative AI models to improve dataset diversity. However, the generative methods struggle with issues such as subject corruption and the introduction of irrelevant artifacts. In this paper, we propose the Automated Generative Data Augmentation (AGA). The framework combines the utility of large language models (LLMs), diffusion models, and segmentation models to augment data. AGA preserves foreground authenticity while ensuring background diversity. Specific contributions include: i) segment and superclass based object extraction, ii) prompt diversity with combinatorial complexity using prompt decomposition, and iii) affine subject manipulation. We evaluate AGA against state-of-the-art (SOTA) techniques on three representative datasets, ImageNet, CUB and iWildCam. The experimental evaluation demonstrates an accuracy improvement of 15.6% and 23.5% for in and out-of-distribution data compared to baseline models respectively. There is also 64.3% improvement in SIC score compared to the baselines.

Processing in-memory is projected to shatter the von Neumann bottleneck and enable acceleration of data-intensive applications. Flow-based computing is an efficient in-memory computing paradigm for accelerating the execution of Boolean logic. While recent synthesis algorithms can map complex functions into flow-based computing circuits, the functional correctness cannot be verified using state-of-the-art equivalence checking techniques. The challenge is that non-volatile memory devices are intrinsically bi-directional, which introduces cycles in the computational graph. These cycles break traditional equivalence checking methods that are based on SAT formulations. In this paper, we propose a framework for equivalence checking of flow-based computing circuits that is called FlowSAT. The framework captures each circuit using an undirected computational graph. The key idea of FlowSAT is to introduce helper variables, in the form of arrows, that dynamically convert the undirected graph into a directed graph. This facilitates equivalence checking to be performed using traditional SAT formulations. However, it is prohibitively expensive to ban all possible cycles using arrow variables. Therefore, we propose to eliminate cycles by iteratively adding constraints to the SAT formulation. Our experimental evaluation demonstrates that FlowSAT is up to an order of magnitude faster than state-of-the-art methods. The framework is capable of verifying all 20/20 benchmark circuits, while the previous state-of-the-art technique is only capable of verifying 12/20 circuits within a time limit of one hour.

Explainable artificial intelligence (XAI) seeks to enhance the transparency, interpretability, and trustworthiness of AI models. One solution strategy for explaining complex AI models for high-dimensional tabular data is to approximate them locally using surrogate models. Surrogate models such as linear regression and decision trees are inherently interpretable and can be used as an explanation. However, it is challenging for linear regression and decision trees to provide meaningful explanations for data points far from the decision boundary. In this paper, we propose a framework that provides Context-aware Local Explanations for high-dimensional tabular data called CLE. We observe that the quality of explanations from different local models varies depending on the data point. The CLE framework uses the context around a data point to select the type of symbolic explanation. Moreover, we propose to utilize feature attributions to explain data points that are far from the decision boundary. The proposed method is evaluated using high-dimensional tabular datasets from the domains of power systems, breast cancer detection, heart disease detection, and website phishing detection. The experimental results show that CLE can provide meaningful local explanations for data points far from the decision boundary. The framework explains data points using three different types of local models and demonstrates a smooth trade-off between explanation accuracy and interpretability. It can be observed that a relatively simple decision tree can explain a data point with 92.31 % accuracy.

Modeling system-level behaviors of intricate System-on-Chip (SoC) designs is crucial for design analysis, testing, and validation. This paper presents an approach, AutoModel, to automatically inferring concise and abstract models from SoC communication traces, capturing the system-level protocols that govern co-ordinations among design blocks for various system functions. In this approach, a causality graph with annotations obtained from the SoC traces is constructed first. The annotated causality graph represents all potential causality relations among messages under consideration. Next, a constraint satisfaction problem is formulated from the causality graph, which is then solved by a satisfiability modulo theories (SMT) solver to find satisfying solutions. Finally, finite state models are extracted from the generated solutions, which can be used to explain and understand the input traces. The proposed approach is validated through experiments using synthetic traces obtained from simulating a transaction-level model of a multicore SoC design and traces collected from running real programs on a realistic multicore SoC modeled in gem5.

Hospital capacity expansion planning is critical for a healthcare authority, especially in regions with a growing diverse population. Policymaking to this end often requires satisfying two conflicting objectives, minimizing capacity expansion cost and minimizing the number of denial of service (DoS) for patients seeking hospital admission. The uncertainty in hospital demand, especially considering a pandemic event, makes expansion planning even more challenging. This work presents a multi-objective reinforcement learning (MORL) based solution for healthcare expansion planning to optimize expansion cost and DoS simultaneously for pandemic and non-pandemic scenarios. Importantly, our model provides a simple and intuitive way to set the balance between these two objectives by only determining their priority percentages, making it suitable across policymakers with different capabilities, preferences, and needs. Specifically, we propose a multi-objective adaptation of the popular Advantage Actor-Critic (A2C) algorithm to avoid forced conversion of DoS discomfort cost to a monetary cost. Our case study for the state of Florida illustrates the success of our MORL based approach compared to the existing benchmark policies, including a state-of-the-art deep RL policy that converts DoS to economic cost to optimize a single objective.

This article proposes a specification mining framework,FlowMiner, that automatically mines patterns from highly concurrent communication traces for system-on-chip (SoC) designs. It addresses the problem of the lack of comprehensive, accurate, and up-to-date specifications necessary to perform rigorous and thorough validation of complex SoC designs. The extracted patterns characterize how components of an SoC design communicate and coordinate with each other to realize various system functions. InFlowMiner, a set of inference rules and optimization techniques are presented to reduce mining complexity. Evaluation of this framework in several experiments shows promising results.

Concise and abstract models of system-level behaviors are invaluable in design analysis, testing, and validation. In this paper, we consider the problem of inferring models from communication traces of system-on-chip (SoC) designs. The traces capture communications among different blocks of a system design in terms of messages exchanged. The extracted models characterize the system-level communication protocols governing how blocks exchange messages, and coordinate with each other to realize various system functions. In this paper, the above problem is formulated as a constraint satisfaction problem, which is then fed to a satisfiability modulo theories (SMT) solver. The solutions returned by the SMT solver are used to extract the models that accept the input traces. In the experiments, we demonstrate the proposed approach with traces collected from a transaction-level simulation model of a multicore SoC design and a trace of a more detailed multicore SoC modeled in GEM5.

The stochastic nature of hospital bed demands and population growth rate in high migration areas poses significant challenges for the authorities to devise an appropriate hospital augmentation scheme. In this study, we propose a deep reinforcement learning (DRL) based model that can identify an appropriate hospital expansion plan for a particular geographical region of interest. Our proposed model analyzes the cost-benefit over a range of geographic regions and recommends the best capacity expansion area. We consider hospital bed numbers as a capacity determiner and population demographics for analyzing future demands economics in our approach. We divide a concerned geographic region into several sub-regions based on the local administrative body to recommend a sub-region where augmentation is necessary. The RL agent then works based on the age group, population growth, and current bed capacity utilizing the Advantage Actor-Critic (A2C) algorithm to minimize the cumulative cost. We also implemented our proposed approach for a case study in the Tampa Bay region, Florida, USA, to identify a hospital augmentation plan. The results from the case study verify this approach's superiority over traditional per capita-based and complaint-based policies.