Plant bioelectric potentials, signal analysis, and reliable machine learning—alongside earlier applied-AI collaborations.
Current work · Kanazawa University
Plant bioelectric potentials (BEP), from signal acquisition and preprocessing to time-series machine learning and reproducible evaluation.
My research examines how plant electrical recordings can be acquired, processed, and evaluated consistently. Biological signals are affected by electrode placement, experimental conditions, baseline drift, and noise; these issues shape what machine-learning results can reliably tell us.
Examine recording configurations, electrodes, reference placement, and experimental context.
Study drift and noise management, signal preparation, and representations for biological time series.
Investigate sequence models, transfer learning, and reproducible validation without overstating generalization.
Research scope: Multisite sensing, environmental context, and edge applications are directions under investigation—not claims of a completed multi-plant deployment or validated real-world system.
These ongoing research projects connect questions and methods to their associated publications. A paper link identifies a documented research output; it does not imply that every proposed system component has been built.
Research question: Which acquisition and signal-processing choices support trustworthy plant bioelectric measurements?
Approach: Review electrode placement, recording hardware, measurement environments, and edge-sensing considerations reported in the literature.
Research question: How can noisy plant bioelectric time-series be classified and evaluated reliably?
Approach: Study sequence classifiers, including Transformer-based models, with attention to preprocessing and validation.
My wider research has also involved applied machine learning in other domains. These are distinct from my current doctoral focus.
Applied computer-vision research combining GAN-based data augmentation and transfer learning for malaria classification from microscopy images.
Applied machine-learning research combining generative augmentation and radial-basis-function networks for landslide susceptibility mapping.
For a complete bibliography, see Publications. For appointments and earlier funded-project participation, see Experience. Get in touch →