伊利诺伊大学东校区

When governments adopt private-sector techniques to reform public services, a tension arises between efforts to improve administrative efficiency and to secure democratic values. While scholars acknowledge this tradeoff, the specific institutional processes through which managers resolve these competing values in practice remain a puzzle. To bridge this gap, this study presents a longitudinal case study of Lean Government initiatives, based on 2.5 years of ethnographic fieldwork in an Ohio municipality. Using Constructive Grounded Theory, the analysis reveals that public managers are not passive executors of efficiency mandates; rather, they actively reconstruct public values through sophisticated social and institutional interactions. This research introduces a conceptual framework for the dynamic construction of good governance, demonstrating how value conflicts are mitigated via meso-level institutional processes. By illustrating how good government and equity are operationalized in managerial decision-making, the research findings offer a practical roadmap for maintaining democratic legitimacy amid administrative reform.

Reminiscence priming (priming from recall of the past) and semantic-to-autobiographical memory priming (priming from general information processing) have been shown to prime involuntary autobiographical memory production in the laboratory and everyday life. As these priming phenomena involve overt (reminiscence priming) or covert (semantic-to-autobiographical priming) autobiographical memory activations, questions naturally arise about which one would lead to stronger involuntary memory production. For various reasons, we hypothesized that despite this difference, both priming phenomena will lead to equivalent involuntary memory production. We tested this hypothesis by comparing reminiscence priming to semantic-to-autobiographical priming on the vigilance task (an involuntary memory task). Participants recalled memories from their past (the autobiographical group) or rated words on the familiarity of their meaning (the semantic group) and were then engaged in the vigilance task, where in between detecting the presence of vertical lines on numerous different slides, they were to report the experience of spontaneous thoughts or memories. The results showed that both groups produced equivalent primed involuntary memory production on the vigilance task. We argue that the results suggest that each of these priming phenomena have an equal likelihood of causing involuntary memory production in everyday life, and that covert memory activations can be as strong or overt activations.

Recent studies have questioned whether nonviolent tactics can be effective for ethnic minorities. However, they often overlook multiethnic coalitions, shifts in campaign composition, and ethnicity's parallel role in armed tactics. This paper re-evaluates the relationship between ethnicity, tactics, and outcomes in civil conflicts. To do so, we introduce new data on ethnic groups in contention that offer time-variant measures of the ethnic attributes of campaigns. We find that the effectiveness of nonviolent tactics for ethnic minorities depends on the point of comparison. Campaigns composed solely of excluded groups succeed less often than those made up entirely of privileged groups. However, minorities have still fared better when using nonviolent as compared to violent tactics. Additional analyses explore ethnic diversity, multiethnic coalitions, hybrid tactics, and alternative measures of success. Taken together, our findings complicate a prevailing assertion that nonviolent tactics are only effective for members of privileged groups.

As transistor technology advances and miniaturizes, single chips or multiple chips (chiplets) are now able to integrate hundreds to thousands of cores. Networks-on-Chips (NoCs) have become the standard on-chip communication fabric for multi/many-core systems on chip (MCSoC), offering advantages over traditional buses in terms of scalability, parallelism, and power efficiency. Due to these properties of NoC, computations and communications of different layers of deep neural networks (DNNs) can be performed efficiently on NoCs. However, traditional mapping strategies may not be suitable for running DNNs due to the different types of communication patterns, such as communication between layers of DNN and one-to-many and many-to-one communications in fully connected layers. Due to varying communication patterns, it is necessary to map computations for different layers of a DNN in a manner that mitigates the communication bottleneck within the NoC. The goal of this work is to accelerate DNNs running on NoCs while minimizing energy consumption by reducing the computation and communication loads on nodes (cores/routers) and links of NoC. We have mathematically modeled the task-resource co-allocation problem that maps neural networks onto NoC-based multicore systems using mixed integer linear programming (MILP). We propose a new mapping algorithm called neighbor-aware and, for the first time, introduce dragonfly algorithm (DA)-based mapping in NoCs in this work. Additionally, we adapt simulated annealing (SA), genetic algorithm (GA), and tabu search (TS) for faster and more energy-efficient mapping of DNNs onto NoCs. Simulations on our inhouse-developed platform, integrating gem5 with the Garnet NoC model and DSENT for energy estimation, and representative DNN workloads (trained on MNIST and Iris) show that the neighbor-aware algorithm consistently outperforms the SA, GA, DA, and TS algorithms in mapping performance score and energy consumption. However, the SA, GA, DA, and TS algorithms perform comparably on those metrics and occasionally surpass the neighbor-aware algorithm in latency and throughput.

This paper introduces a family of memory-lean data structures for associating unbounded label sets with graph entities while offering constant-time query mechanics and predictable storage overheads. The core design assigns a tuple index to each sorted label-set and maintains bidirectional mappings plus per-label posting lists, eliminating costly rehashing and tree rebalancing typical of conventional key-value containers. An inplace, single-vector linked organization connects tuple indices to entity identifiers, making total overhead effectively independent of the number of labels attached to any single node or edge. A lightweight index-recycling queue constrains auxiliary structure growth, and serialization enables compact persistence without format-specific coupling. Demonstrations on billion-scale graphs show linear memory scaling and sub-second multi-hop query latency on a single host, substantially improving over hashtable and balanced-tree baselines used for label association, and positioning the approach as a practical ADT for property-graph metadata at scale.