**Abstract:** This research proposes a novel approach to temporal pattern recognition by integrating reservoir computing (RC) with quantum-enhanced stochastic dynamics to emulate non-Markovian processes observed in biological neural networks. Existing RC systems typically suffer from limitations in handling complex, memory-dependent temporal signals. Our system, Quantum-Reservoir Temporal Amplifier (QRTA), leverages adiabatic quantum computation (AQC) to dynamically tune the reservoir’s response, exponentiating pattern recognition capabilities and demonstrating increased accuracy in classifying chaotic time series representative of neural spiking activity.

Researchers have developed a new computational method that accurately predicts how long molecular spins can maintain information by directly linking a molecule’s structure to the disruptive influence of surrounding atomic nuclei at extremely low temperatures.

YouTube video by Institute for Pure & Applied Mathematics (IPAM)

We introduce SpinPulse, an open-source python package for simulating spin qubit-based quantum computers at the pulse-level. SpinPulse models the specific physics of spin qubits, particularly through t...

**Abstract:** This research proposes a novel, scalable, and robust method for integrating Microbial Electrolysis Cells (MECs) with lunar habitat waste management systems. Leveraging established microbial fuel cell technology and recent advances in Markov Decision Processes (MDPs), we introduce a dynamically controlled, closed-loop system capable of organic waste decomposition, hydrogen generation, and supporting in-situ resource utilization (ISRU) on the lunar surface. Our framework prioritizes reliability and adaptability, crucial for long-term, autonomous operation in a resource-constrained extraterrestrial environment.

arXiv abstract link

**Abstract:** This paper proposes a novel approach to analyzing non-Markovian deviations in chaotic time series, leveraging a dynamically weighted hybrid Bayesian filtering network (DWHBFN). Traditional methods often struggle to accurately model the evolving statistical properties inherent in these systems. Our system combines Kalman filtering, particle filtering, and Gaussian process regression within a hierarchical Bayesian framework, enabling adaptive learning of complex dependencies and robust prediction of future states.

**Abstract:** This paper introduces a novel methodology, Dynamically Adaptive Markovian Approximation (DAMA), for characterizing and predicting behavior within stochastic dynamic systems exhibiting non-Markovian time series. Traditional approaches struggle with the inherent memory effects present in these systems, often requiring computationally expensive simulations. DAMA provides a computationally efficient alternative by dynamically adjusting the order of a Markov chain approximation based on localized statistical divergence between observed trajectories and higher-order predictions.

arXiv abstract link

arXiv abstract link

arXiv abstract link

Fault-tolerant quantum computing requires extremely precise knowledge and control of qubit dynamics during the application of a gate. We develop a data-driven learning protocol for characterizing quan...

arXiv abstract link

Markovian Pre-Trained Transformer for Next-Item Recommendation - BDML-lab/MPT