arXiv subjects

Aaron Lee

Publications and source records attributed to Aaron Lee.

13 recordsLinked to original sources

arXiv2026

Autonomous Event-Driven Multi-Agent Orchestration for Enterprise AI at Scale

Enterprise AI aims to move toward continuous event monitoring, detection, and action across specialist agents, yet existing multi-agent systems largely assume discrete request-response workflows and remain underexplored at enterprise scale. We evaluate DAG Plan and Execute and ReAct across 208 production-derived enterprise scenarios spanning Persona (<10 agents), Department (20-80), and Enterprise (200) scales, and introduce a Task Manager for continuous operation via priority inference, related-event merging, and preemption. Results show that scale, not task complexity, dominates orchestration performance: both architectures perform well at small scale but degrade at enterprise scale as agent discovery noise becomes the primary bottleneck, with simple tasks degrading more sharply than complex ones. DAG Plan and Execute offers higher precision and structured parallelization at smaller scales, but its higher overhead worsens at enterprise scale; ReAct is more robust by handling failures incrementally. The Task Manager reduces high-priority queue latency by 14-75% and improves related-event correctness by over 20 percentage points at enterprise scale.

cs.AI↗

arXiv2026

Multi-Class Neurological Disorder Prediction with Tensor Network Feature Engineering

Accurate diagnosis of neurological disorders is contingent upon advanced imaging modalities such as Magnetic Resonance Imaging (MRI), which commonly utilize sparse imaging techniques to reconstruct images from limited data, thus reducing storage and acquisition time. However, challenges remain in managing noise and preserving critical diagnostic features for effective analysis. In this study, an ensemble classifier is enriched with PARAFAC CP tensor decompositions, drawing mathematical inspiration from quantum neural network architectures but implemented entirely classically. The model was evaluated on a large, balanced clinical dataset comprising 55,160 images across 8 diagnostic categories, employing both higher and lower PARAFAC rank configurations. Evaluated through 5-fold nested stratified cross-validation, both configurations achieved strong validation performance, demonstrating robustness to tensor network expressivity. Additionally, the proposed model achieved competitive performance relative to recent classical approaches, further underscoring the potential of quantum-inspired classical frameworks to enhance medical image analysis and support reliable clinical diagnosis. Future work will explore the integration of advanced encoding schemes, deployment on real quantum hardware, and the use of more diverse neurological datasets.

stat.AP↗

arXiv2025

A Reinforcement Learning Approach to Synthetic Data Generation

Synthetic data generation (SDG) is a promising approach for enabling data sharing in biomedical studies while preserving patient privacy. Yet, state-of-the-art generative models often require large datasets and complex training procedures, limiting their applicability in small-sample settings common in biomedical research. This study aims to develop a more principled and efficient approach to SDG and evaluate its efficacy for biomedical applications. In this work, we reframe SDG as a reinforcement learning (RL) problem and introduce RLSyn, a novel framework that models the data generator as a stochastic policy over patient records and optimizes it using Proximal Policy Optimization with discriminator-derived rewards. We evaluate RLSyn on two biomedical datasets--AI-READI and MIMIC-IV--and benchmark it against state-of-the-art generative adversarial networks (GANs) and diffusion-based methods across extensive privacy, utility, and fidelity evaluations. On MIMIC-IV, RLSyn achieves predictive utility comparable to diffusion models (S2R AUC 0.902 vs 0.906 respectively) while slightly outperforming them in fidelity (NMI 0.001 vs. 0.003; DWD 2.073 vs. 2.797) and achieving comparable, low privacy risk (~0.50 membership inference risk AUC). On the smaller AI-READI dataset, RLSyn again matches diffusion-based utility (S2R AUC 0.873 vs. 0.871), while achieving higher fidelity (NMI 0.001 vs. 0.002; DWD 13.352 vs. 16.441) and significantly lower vulnerability to membership inference attacks (AUC 0.544 vs. 0.601). Both RLSyn and diffusion-based models substantially outperform GANs across utility and fidelity on both datasets. Our results suggest that reinforcement learning provides a principled and effective approach for synthetic biomedical data generation, particularly in data-scarce regimes.

cs.LG↗

arXiv2025

Optimas: Optimizing Compound AI Systems with Globally Aligned Local Rewards

Compound AI systems integrating multiple components, such as Large Language Models, specialized tools, and traditional machine learning models, are increasingly deployed to solve complex real-world tasks. However, optimizing compound systems remains challenging due to their non-differentiable structures and diverse configuration types across components, including prompts, hyperparameters, and model parameters. To address this challenge, we propose Optimas, a unified framework for effective optimization of compound systems. The core idea of Optimas is to maintain one Local Reward Function (LRF) per component, each satisfying a local-global alignment property, i.e., each component's local reward correlates with the global system performance. In each iteration, Optimas efficiently adapts the LRFs to maintain this property while simultaneously maximizing each component's local reward. This approach enables independent updates of heterogeneous configurations using the designated optimization method, while ensuring that local improvements consistently lead to performance gains. We present extensive evaluations across five real-world compound systems to demonstrate that Optimas outperforms strong baselines by an average improvement of 11.92%, offering a general and effective approach for improving compound systems. Our website is at https://optimas.stanford.edu.

cs.LG↗

arXiv2021

Unsupervised cross domain learning with applications to 7 layer segmentation of OCTs

Unsupervised cross domain adaptation for OCT 7 layer segmentation and other medical applications where labeled training data is only available in a source domain and unavailable in the target domain. Our proposed method helps generalize of deep learning to many areas in the medical field where labeled training data are expensive and time consuming to acquire or where target domains are too novel to have had labelling.

eess.IV↗

arXiv2021

Generalized Permutation Framework for Testing Model Variable Significance

A common problem in machine learning is determining if a variable significantly contributes to a model's prediction performance. This problem is aggravated for datasets, such as gene expression datasets, that suffer the worst case of dimensionality: a low number of observations along with a high number of possible explanatory variables. In such scenarios, traditional methods for testing variable statistical significance or constructing variable confidence intervals do not apply. To address these problems, we developed a novel permutation framework for testing the significance of variables in supervised models. Our permutation framework has three main advantages. First, it is non-parametric and does not rely on distributional assumptions or asymptotic results. Second, it not only ranks model variables in terms of relative importance, but also tests for statistical significance of each variable. Third, it can test for the significance of the interaction between model variables. We applied this permutation framework to multi-class classification of the Iris flower dataset and of brain regions in RNA expression data, and using this framework showed variable-level statistical significance and interactions.

stat.ME↗

arXiv2018

Real-time Prediction of Segmentation Quality

Recent advances in deep learning based image segmentation methods have enabled real-time performance with human-level accuracy. However, occasionally even the best method fails due to low image quality, artifacts or unexpected behaviour of black box algorithms. Being able to predict segmentation quality in the absence of ground truth is of paramount importance in clinical practice, but also in large-scale studies to avoid the inclusion of invalid data in subsequent analysis. In this work, we propose two approaches of real-time automated quality control for cardiovascular MR segmentations using deep learning. First, we train a neural network on 12,880 samples to predict Dice Similarity Coefficients (DSC) on a per-case basis. We report a mean average error (MAE) of 0.03 on 1,610 test samples and 97% binary classification accuracy for separating low and high quality segmentations. Secondly, in the scenario where no manually annotated data is available, we train a network to predict DSC scores from estimated quality obtained via a reverse testing strategy. We report an MAE=0.14 and 91% binary classification accuracy for this case. Predictions are obtained in real-time which, when combined with real-time segmentation methods, enables instant feedback on whether an acquired scan is analysable while the patient is still in the scanner. This further enables new applications of optimising image acquisition towards best possible analysis results.

cs.CV↗

arXiv2015

Many-body localization in a quantum simulator with programmable random disorder

When a system thermalizes it loses all local memory of its initial conditions. This is a general feature of open systems and is well described by equilibrium statistical mechanics. Even within a closed (or reversible) quantum system, where unitary time evolution retains all information about its initial state, subsystems can still thermalize using the rest of the system as an effective heat bath. Exceptions to quantum thermalization have been predicted and observed, but typically require inherent symmetries or noninteracting particles in the presence of static disorder. The prediction of many-body localization (MBL), in which disordered quantum systems can fail to thermalize in spite of strong interactions and high excitation energy, was therefore surprising and has attracted considerable theoretical attention. Here we experimentally generate MBL states by applying an Ising Hamiltonian with long-range interactions and programmably random disorder to ten spins initialized far from equilibrium. We observe the essential signatures of MBL: memory retention of the initial state, a Poissonian distribution of energy level spacings, and entanglement growth in the system at long times. Our platform can be scaled to higher numbers of spins, where detailed modeling of MBL becomes impossible due to the complexity of representing such entangled quantum states. Moreover, the high degree of control in our experiment may guide the use of MBL states as potential quantum memories in naturally disordered quantum systems.

quant-ph↗

arXiv2014

ACO Implementation for Sequence Alignment with Genetic Algorithms

In this paper, we implement Ant Colony Optimization (ACO) for sequence alignment. ACO is a meta-heuristic recently developed for nearest neighbor approximations in large, NP-hard search spaces. Here we use a genetic algorithm approach to evolve the best parameters for an ACO designed to align two sequences. We then used the best parameters found to interpolate approximate optimal parameters for a given string length within a range. The basis of our comparison is the alignment given by the Needleman-Wunsch algorithm. We found that ACO can indeed be applied to sequence alignment. While it is computationally expensive compared to other equivalent algorithms, it is a promising algorithm that can be readily applied to a variety of other biological problems.

cs.CE↗

arXiv2014

Non-local propagation of correlations in long-range interacting quantum systems

The maximum speed with which information can propagate in a quantum many-body system directly affects how quickly disparate parts of the system can become correlated and how difficult the system will be to describe numerically. For systems with only short-range interactions, Lieb and Robinson derived a constant-velocity bound that limits correlations to within a linear effective light cone. However, little is known about the propagation speed in systems with long-range interactions, since the best long-range bound is too loose to give the correct light-cone shape for any known spin model and since analytic solutions rarely exist. In this work, we experimentally determine the spatial and time-dependent correlations of a far-from-equilibrium quantum many-body system evolving under a long-range Ising- or XY-model Hamiltonian. For several different interaction ranges, we extract the shape of the light cone and measure the velocity with which correlations propagate through the system. In many cases we find increasing propagation velocities, which violate the Lieb-Robinson prediction, and in one instance cannot be explained by any existing theory. Our results demonstrate that even modestly-sized quantum simulators are well-poised for studying complicated many-body systems that are intractable to classical computation.

quant-ph↗

arXiv2013

Quantum Catalysis of Magnetic Phase Transitions in a Quantum Simulator

We control quantum fluctuations to create the ground state magnetic phases of a classical Ising model with a tunable longitudinal magnetic field using a system of 6 to 10 atomic ion spins. Due to the long-range Ising interactions, the various ground state spin configurations are separated by multiple first-order phase transitions, which in our zero temperature system cannot be driven by thermal fluctuations. We instead use a transverse magnetic field as a quantum catalyst to observe the first steps of the complete fractal devil's staircase, which emerges in the thermodynamic limit and can be mapped to a large number of many-body and energy-optimization problems.

quant-ph↗

arXiv2013

Experimental Performance of a Quantum Simulator: Optimizing Adiabatic Evolution and Identifying Many-Body Ground States

We use local adiabatic evolution to experimentally create and determine the ground state spin ordering of a fully-connected Ising model with up to 14 spins. Local adiabatic evolution -- in which the system evolution rate is a function of the instantaneous energy gap -- is found to maximize the ground state probability compared with other adiabatic methods while only requiring knowledge of the lowest $\sim N$ of the $2^N$ Hamiltonian eigenvalues. We also demonstrate that the ground state ordering can be experimentally identified as the most probable of all possible spin configurations, even when the evolution is highly non-adiabatic.

quant-ph↗

arXiv2008

High Voltage CMOS Control Interface for Astronomy - Grade Charged Coupled Devices

The Pan-STARRS telescope consists of an array of smaller mirrors viewed by a Gigapixel arrays of CCDs. These focal planes employ Orthogonal Transfer CCDs (OTCCDs) to allow on-chip image stabilization. Each OTCCD has advanced logic features that are controlled externally. A CMOS Interface Device for High Voltage has been developed to provide the appropiate voltage signal levels from a readout and control system designated STARGRASP. OTCCD chip output levels range from -3.3V to 16.7V, with two different output drive strenghts required depending on load capacitance (50pF and 1000pF), with 24mA of drive and a rise time on the order of 100ns. Additional testing ADC structures have been included in this chip to evaluate future functional additions for a next version of the chip.

physics.ins-det↗