Predicting transport processes in environmental flows, including sediments, biota, and dissolved tracers, is computationally demanding. The quantum lattice Boltzmann method (QLBM) has emerged as a candidate acceleration strategy, but prior demonstrations rely on simulators and idealized benchmarks rather than environmental cases. We present a hybrid quantum–classical LBM model for advection–diffusion–settling transport in a shear-driven open channel, applied to invasive carp egg dispersion. In this study, the collision, streaming, and spatial indexing are computed classically, while the quantum resource encodes and samples the node-local post-collision probability distribution. The hardware experiments on an IBM quantum computer are snapshot validations of this quantum sampling step rather than a fully quantum time evolution. The results show that the emulated sampling error follows the theoretical Nshots−1/2 law, and the number of shots needed to match the LBM discretization accuracy grows as the inverse square of the discretization error. Snapshot circuits executed on the IBM Heron r2 processor ibm_pittsburgh reproduce the theoretical post-collision distributions with mean total-variation distances of 0.052 with 102 shots and 0.017 with 103 shots, close to the values of 0.048 and 0.015, which are produced on an error-free machine by statistical sampling noise only. We discuss what NISQ-era hardware can contribute today and longer-term scenarios for quantum-accelerated environmental simulation.
Driven by the rapid advancement of the Artificial Intelligence of Things (AIoT), new possibilities have emerged for predictive maintenance in safety-critical systems such as firefighting pumps, whose uninterrupted operation is essential for fire protection. This study presents a domain-specific, experimentally grounded AIoT maintenance framework for firefighting pumps, built around a carefully tuned hybrid model that combines Improved Particle Swarm Optimization with Long Short-Term Memory and Convolutional Neural Network (IPSO-LSTM-CNN) architectures for failure classification. Rather than proposing a fundamentally new algorithm, the contribution lies in (i) an end-to-end AIoT data-acquisition and maintenance pipeline that has been deployed in a real pump-manufacturing facility, and (ii) a systematically optimized IPSO-LSTM-CNN configuration in which IPSO automatically tunes the hyperparameters of the LSTM-CNN to capture both temporal and spatial fault signatures. Using a real firefighting-pump dataset collected at four sampling frequencies, the framework is evaluated with 20-fold cross-validation against RNN, GRU, LSTM, CNN, CNN-GRU, and CNN-LSTM baselines, and IPSO is compared with conventional PSO and random search. Across all sampling rates, the IPSO-LSTM-CNN model consistently attains the lowest error metrics and the most stable precision and recall, with the best results at the 20 s and 40 s rates and a practical accuracy-cost compromise at the 1-minute rate. These results demonstrate a reliable and deployable solution for real-time failure detection that strengthens the operational reliability of essential fire-protection infrastructure.
In this paper, we analyze the computational complexity of the problem of determining the best possible covenant under egalitarian social welfare. In particular, we consider the scenario in which there are several possible measures to be approved or rejected, and for each of them each agent has a value that it would obtain if it is approved (which will be negative if it is against the measure). If it is not approved, then it does not contribute any positive or negative value to any agent. Under these assumptions, and assuming egalitarian social welfare, we show that the problem of determining the best possible covenant is not only NP-complete, but also that it is not possible to guarantee good approximations in polynomial time. We also show that, when some agent wants to lie to benefit from the system, the problem of finding a lie providing that agent with some target profit is Σ2p-complete. In addition, we consider a realistic case study in which eight political parties must reach an agreement on 25 laws. On the one hand, we provide genetic algorithms to solve the problem. On the other hand, we analyze how to discourage political parties from lying about their preferences in order to benefit in the legislative agreement. In fact, we show that by using a simple restriction, we can discourage lying in practical terms.
The Ataxia telangiectasia mutated (ATM) protein plays a crucial role in impeding cancer progression and is traditionally recognized for its involvement in the DNA damage response (DDR). Recent studies reveal that ATM induces the expression of Snail1 and ZEB1, key drivers of epithelial-mesenchymal transition (EMT) and drug resistance. MicroRNA-34a (miR-34a) and microRNA-200 (miR-200), both p53-inducible miRNAs, directly target Snail1 and ZEB1, thereby inhibiting EMT and drug resistance. However, the dynamic regulatory interplay within the ATM signaling pathway remains unclear. To address this, we constructed a dynamic Boolean model informed by literature-based cancer studies and validated it using independent experimental evidence from PubMed. Subsequently, model predictions were evaluated using PubTator3, an AI-driven literature mining platform, to verify predictive coherence with recent experimental discoveries. Gain- and loss-of-function perturbation analyses further revealed dual regulatory roles of ATM and p53-inducible miRNAs in promoting or inhibiting DDR and apoptosis pathways. Notably, we identified a double-negative feedback loop between miR-34a and ZEB1, acting as a bistable switch that governs EMT, stemness, drug resistance, and apoptosis. Collectively, this study demonstrates how Boolean network modeling combined with AI-based literature assessment can elucidate emergent regulatory behaviors in cancer systems. The ATM/miR-34a–miR-200 axis uncovered here provides mechanistic and therapeutic insights into DDR modulation and cell-fate dynamics in breast cancer.
