Skip to content
P Perceptronix.
Research

Active research, applied to live systems.

Perceptronix sits at the intersection of academic research and industrial AI delivery. Recent work spans macroeconomic forecasting with NIESR, neural ensemble methods for financial time-series, and applied GenAI architectures. ORCID 0000-0001-7339-0132 for the full publication record.

Themes

Three current research strands.

Multi-recurrent neural networks (MRNs)

A long-running research programme on deep recurrent architectures whose recurrent state is structured to capture multi-scale temporal dependencies. Applied across NLP, finance, and macroeconomics.

AI for macroeconomic forecasting

Published with NIESR and the University of Birmingham. UK CPI inflation forecasts to within ±0.2% across multi-month horizons; US inflation turning-point detection ahead of the SPF.

Cognitive science & neural computation

Work published in Trends in Cognitive Science and AI in Medicine on the computational foundations of recurrent learning systems.

Selected publications

Recent featured outputs.

  • 2025

    UK CPI inflation outlook — Winter 2025

    Tepper, J. A. et al. · NIESR Economic Outlook (Winter 2025)

    MASCET inflation forecasts and turning-point analysis.

  • 2025

    Multi-recurrent neural ensembles for financial time-series forecasting

    Tepper, J. A. et al. · Journal of Risk and Financial Management (JRFM)

  • 2024

    UK CPI inflation outlook — Spring 2024

    Tepper, J. A. et al. · NIESR Economic Outlook (Spring 2024)

    Forecasts within ±0.2% across multi-month horizons.

  • 2023

    Neural ensembles for cross-cutting economic trend detection

    Tepper, J. A. et al. · Procedia Computer Science

Full publication record on ORCID and Google Scholar.

Let's build something measurable

Have a forecasting or classification problem that needs to work in production?

Tell us about it. First conversations are confidential, no-obligation, and usually end with a clear view of feasibility, data needs, and time-to-value.