About

Andreas Andreou is a professor of electrical and computer engineering at Johns Hopkins University, with secondary appointments in the Department of Computer Science and the Whitaker Biomedical Engineering Institute. He is the co-founder of the Johns Hopkins University Center for Language and Speech Processing. His research focuses on principles of computing and computing machinery, sensory information processing, and theoretical neuroscience, as well as pattern analysis and machine intelligence. Andreou's lab has achieved notable microsystems innovations over the past 25 years, including a contrast sensitive silicon retina, the first CMOS polarization-sensitive imager, silicon rods in standard foundry CMOS for single-photon detection, and a large-scale mixed analog/digital associative processor for character recognition. His significant algorithmic contributions include the vocal tract normalization technique for speech recognition and heteroscedastic linear discriminant analysis, a generalization of Fisher discriminants within the maximum likelihood framework. An IEEE Fellow, Andreou's work bridges hardware microsystems and algorithmic research, advancing understanding in sensory processing and machine intelligence.

Research topics

• Artificial Intelligence
• Computer Science
• Political Science
• Sociology
• Engineering ethics
• Virology
• Psychology
• Real-time computing
• Embedded system
• Computer hardware
• Algorithm
• Philosophy
• Medicine
• Law
• General surgery
• Telecommunications
• Social psychology
• Engineering
• Epistemology
• Internal medicine

Selected publications

• Higher-order neuromorphic Ising machines—autoencoders and Fowler-Nordheim annealers are all you need for scalability

  Nature Communications · 2026-04-16

  articleOpen access

  We report that an autoencoder-based neuromorphic architecture, combined with Fowler-Nordheim annealing, is sufficient to implement scalable higher-order Ising machines. We show that these machines can consistently produce state-of-the-art solutions with high reliability and with competitive time-to-solution metrics. The autoencoder captures higher-order interactions by decomposing Ising clauses and Ising spins into encoder-decoder layers of spiking neurons, thereby keeping the resource complexity independent of the interaction order for sparse problems. An annealing process based on the dynamics of Fowler-Nordheim quantum mechanical tunneling extrapolates between an $${{\mathcal{O}}}(1/t)$$ annealing schedule and an $${{\mathcal{O}}}(1/\log (t))$$ annealing schedule. This not only ensures fast convergence towards high-quality solutions but also guarantees asymptotic convergence to the Ising ground state. To demonstrate the advantages of the proposed higher-order neuromorphic Ising machine, we systematically solved benchmark combinatorial optimization problems such as MAX-CUT and MAX-SAT, comparing the results to those obtained using a second-order Ising machine employing the same annealing process. The authors demonstrate that an autoencoder-based neuromorphic architecture combined with Fowler-Nordheim annealing, is sufficient to implement scalable higher-order Ising machines. They show that these machines can consistently produce state-of-the-art solutions with high reliability and competitive time-to-solution metrics.

  PublisherOA PDFDOI

• In-Memory-Computing For Machine Cognition at the Edge

  IEEE Journal on Emerging and Selected Topics in Circuits and Systems · 2026-01-01

  article1st authorCorresponding

  In-Memory-Computing (IMC) is an alternative computer architecture to the standard von Neumann machine, in which compute functionality is incorporated with the state holding elements - memory. This idea is inspired from biology, where memory and processing are integrated in a single structure, the synapse. In this paper we explore the sometimes counterintuitive implications of integrating processing within the memory cells demonstrate how the latter can be leveraged to dramatically increase energy efficiency of computing hardware for machine cognition. We demonstrate the performance in real-world application examples on actual scaled up physical hardware (a 79 core IMC System On Chip (SOC). We finally show how recent developments allow IMC computers (as opposed to von Neumann computers with IMC accelerators) have the weights storage density necessary for embodied intelligence and machine cognition at the edge.

  PublisherDOI

• Author Correction: ON-OFF neuromorphic ISING machines using Fowler-Nordheim annealers

  Nature Communications · 2025-07-01 · 1 citations

  erratumOpen access
  PublisherOA PDFDOI

• Robotic-assisted Liver Resection in Elderly Patients: A Retrospective Single-center Analysis

  HPB · 2025-01-01

  article
  PublisherDOI

• Redo of Pancreatic Anastomosis after 19 Months due to Grade C-Fistula Following Whipple`s Procedure

  HPB · 2025-01-01

  articleOpen access
  PublisherOA PDFDOI

• Impact of parenchymal transection techniques on intraoperative blood loss during liver resection in a porcine model of elevated central venous pressure: A comparative study

  Annals of Hepatology · 2025-01-01 · 1 citations

  articleOpen access

  INTRODUCTION AND OBJECTIVES: Liver resection is the standard treatment for resectable liver tumors and metastases. However, mortality and morbidity remain significant concerns, particularly for patients with chronically elevated central venous pressure (CVP), which increases perioperative complication risks. The optimal parenchymal transection technique for these patients remains unclear, necessitating further research. MATERIALS AND METHODS: This study established an innovative porcine model for high-CVP liver resection. Animals were divided into two groups: a control group (CVP ≤ 5 mmHg, low-CVP) and an intervention group (CVP ≥ 10 mmHg, high-CVP). A left lateral liver resection was performed using three parenchymal transection techniques: clamp-crush (CC), harmonic scalpel (HS), and stapler (ST). The primary endpoint was intraoperative blood loss, while secondary endpoints included transection time and bile leakage. RESULTS: No differences were found for blood loss or transection time among the low-CVP subgroups. In the high-CVP group, the HS and ST techniques were associated with significantly reduced blood loss and faster transection times than the CC technique. While transection times for the HS and ST were similar between the low- and high-CVP groups, they were significantly longer with the CC technique in the high-CVP group. The incidence of bile leakage was comparable across all three techniques. CONCLUSIONS: This pilot study demonstrates superior outcomes for HS and ST techniques in high-CVP liver resections. Insights from this large animal model provide a basis for investigating optimal transection techniques for chronically elevated CVP, bridging preclinical research and clinical practice.

  PublisherDOI

• The neurobench framework for benchmarking neuromorphic computing algorithms and systems

  Nature Communications · 2025-02-11 · 63 citations

  reviewOpen access

  Neuromorphic computing shows promise for advancing computing efficiency and capabilities of AI applications using brain-inspired principles. However, the neuromorphic research field currently lacks standardized benchmarks, making it difficult to accurately measure technological advancements, compare performance with conventional methods, and identify promising future research directions. This article presents NeuroBench, a benchmark framework for neuromorphic algorithms and systems, which is collaboratively designed from an open community of researchers across industry and academia. NeuroBench introduces a common set of tools and systematic methodology for inclusive benchmark measurement, delivering an objective reference framework for quantifying neuromorphic approaches in both hardware-independent and hardware-dependent settings. For latest project updates, visit the project website ( neurobench.ai ).

  PublisherOA PDFDOI

• ON-OFF neuromorphic ISING machines using Fowler-Nordheim annealers

  Nature Communications · 2025-03-31 · 9 citations

  articleOpen access

  We introduce NeuroSA, a neuromorphic architecture specifically designed to ensure asymptotic convergence to the ground state of an Ising problem using a Fowler-Nordheim quantum mechanical tunneling based threshold-annealing process. The core component of NeuroSA consists of a pair of asynchronous ON-OFF neurons, which effectively map classical simulated annealing dynamics onto a network of integrate-and-fire neurons. The threshold of each ON-OFF neuron pair is adaptively adjusted by an FN annealer and the resulting spiking dynamics replicates the optimal escape mechanism and convergence of SA, particularly at low-temperatures. To validate the effectiveness of our neuromorphic Ising machine, we systematically solved benchmark combinatorial optimization problems such as MAX-CUT and Max Independent Set. Across multiple runs, NeuroSA consistently generates distribution of solutions that are concentrated around the state-of-the-art results (within 99%) or surpass the current state-of-the-art solutions for Max Independent Set benchmarks. Furthermore, NeuroSA is able to achieve these superior distributions without any graph-specific hyperparameter tuning. For practical illustration, we present results from an implementation of NeuroSA on the SpiNNaker2 platform, highlighting the feasibility of mapping our proposed architecture onto a standard neuromorphic accelerator platform. Combinatorial Optimization problems can be solved by investigating the ground states of particular Ising models. Here, the authors developed a neuromorphic architecture to ensure asymptotic convergence to the ground state of an Ising problem and to consistently produce high-quality solutions.

  PublisherOA PDFDOI

• From Lipid Regulation to Neuroprotection: Multitarget (Benzo)thiazine Derivatives as Promising Leads

  Molecules · 2025-11-25 · 1 citations

  articleOpen access

  Neurodegenerative and cardiovascular disorders share multifactorial origins, including oxidative stress, (neuro)inflammation, and lipid dysregulation—factors often addressed independently by single-target therapies. In this study, we report a rational multitarget approach through the design and synthesis of novel (benzo)thiazine derivatives that integrate antioxidant, anti-inflammatory, and antihyperlipidemic functionalities within a single molecular framework. The compounds were obtained in good yields via 3–7 step synthetic routes and evaluated through complementary in vitro and in vivo assays. Several derivatives displayed potent inhibition of lipoxygenase (IC50 < 100 μM), significant reduction in carrageenan-induced edema (up to 60%), strong free radical scavenging and lipid peroxidation inhibition, as well as effective iron chelation. In vivo, most derivatives enhanced total antioxidant capacity (by 50–800%) and significantly improved plasma lipid profiles in mouse, while almost all compounds increased the plasma antiatherogenic index by more than 100% with selected compounds exceeding 600%. Notably, several molecules also showed moderate acetylcholinesterase inhibition, suggesting preliminary neuroprotective potential. Altogether, these multifunctional (benzo)thiazine derivatives represent promising lead structures for the development of agents targeting the complex interplay of oxidative, inflammatory, and metabolic pathways underlying neurodegenerative and cardiovascular diseases.

  PublisherOA PDFDOI

• Natural Language to Verilog: Design of a Recurrent Spiking Neural Network using Large Language Models and ChatGPT

  arXiv (Cornell University) · 2024-05-02

  preprintOpen accessSenior author

  This paper investigates the use of Large Language Models (LLMs) and natural language prompts to generate hardware description code, namely Verilog. Building on our prior work, we employ OpenAI's ChatGPT4 and natural language prompts to synthesize an RTL Verilog module of a programmable recurrent spiking neural network, while also generating test benches to assess the system's correctness. The resultant design was validated in three simple machine learning tasks, the exclusive OR, the IRIS flower classification and the MNIST hand-written digit classification. Furthermore, the design was validated on a Field-Programmable Gate Array (FPGA) and subsequently synthesized in the SkyWater 130 nm technology by using an open-source electronic design automation flow. The design was submitted to Efabless Tiny Tapeout 6.

  PublisherOA PDFDOI

Frequent coauthors

• Philippe O. Pouliquen

  Johns Hopkins University

  71 shared

• P. Julián

  Universidad Nacional del Sur

  58 shared

• Teresa Serrano‐Gotarredona

  Instituto de Microelectrónica de Sevilla

  35 shared

• B. Linares-Barranco

  34 shared

• Julius Georgiou

  University of Cyprus

  28 shared

• Guillaume Garreau

  24 shared

• Eugenio Culurciello

  22 shared

• Andrew Cassidy

  21 shared

Awards & honors

• IEEE Fellow