Emotions are a compass that guide our social lives, relationships, decision-making, and overall well-being. Yet, many of the hidden, individualized brain processes that shape how people experience and regulate emotions remain poorly understood.
Similar challenges arise in neurological and psychiatric conditions, where individuals with the same diagnosis may experience very different symptoms and patterns of brain activity.
To help us better understand, a UWinnipeg interdisciplinary team is looking at how brain activity relates to emotion, cognition, behaviour, and mental health that includes psychology, neuroscience, and artificial intelligence.
In this study, we explored a logic-based approach that substantially reduced computing requirements while maintaining competitive performance.
Shyamal Dharia
Dr. Stephen D. Smith in Psychology and Dr. Camilo E. Valderrama in Applied Computer Science are working with Shyamal Dharia (MSc 25), a senior researcher, who received the UWinnipeg 2025 Graduate Student of Highest Distinction Award in applied computer science, will help connect the dots between psychology, neuroscience, and artificial intelligence.
By bringing together expertise in psychology, neuroscience, signal processing, and artificial intelligence, Dr. Smith’s research examines how emotion, cognition, personality, and mental health are represented in the brain; while Dr. Valderrama contributes expertise in signal processing, statistical modelling, machine learning, and artificial intelligence.
This interdisciplinary team recently presented their research at the Canadian AI 2026 conference at Simon Fraser University in British Columbia. Their paper earned the Best Paper Award for their work exploring an efficient artificial intelligence method for analyzing brain activity titled Differentiable Logic Gate Networks for Low-Latency EEG Classification on Edge Devices.
The study explored the use of Differentiable Logic Gate Networks to classify electroencephalography, or EEG, signals—recordings of the brain’s electrical activity collected using sensors placed on the scalp—on small computing devices. Unlike many conventional artificial intelligence models, which rely on computationally demanding mathematical operations, this approach processes information through simple on-and-off decisions performed by units called logic gates.
“The model we used works differently,” shared Dharia. “Our model makes many simple yes-or-no decisions using units called ‘logic gates’. You can think of each decision as being like a light switch, where 0 means ‘off’ or ‘no’ and 1 means ‘on’ or ‘yes’. This model combines many of these small yes-or-no decisions to identify patterns in the EEG signals and make a final prediction. Because these operations are simple, the model can run on smaller computing devices with limited computing power.”
Large artificial intelligence models can be very effective at analyzing brain signals, but their computational demands may limit their use on smaller, power-constrained devices.
“In this study, we explored a logic-based approach that substantially reduced computing requirements while maintaining competitive performance,” explained Dharia. “We also found that its processing time remained nearly constant as the model became considerably larger across the model sizes we tested.”
By applying the method to EEG data, the researchers showed that it could substantially reduce computing demands while maintaining performance comparable to, and in some cases better than, conventional models. In the future, this improved efficiency could allow for the processing of EEG data directly on compact devices in a physician’s office rather than requiring patients to go for testing in specialized hospital wards.
“Combining these areas allows us to develop computational methods that are technically strong while remaining connected to meaningful questions about brain function and behaviour,” explained Dr. Valderrama.
The team plans to continue exploring efficient and interpretable methods for analyzing complex brain signals, with the broader aim of developing practical research tools for neuroscience and brain health.
