Guwahati: Researchers from the Indian Institute of Technology Guwahati have developed a brain-inspired Artificial Intelligence (AI) model designed to process long sequences of data while using less energy than conventional AI approaches.
The model, developed by researchers at IIT Guwahati’s Mehta Family School of Data Science and Artificial Intelligence, was presented at the International Conference on Machine Learning (ICML) 2026 in Seoul, South Korea. ICML is a major international conference in the field of AI and machine learning.
The research could have applications in wearable health monitoring systems, Internet of Things (IoT) sensors, smart manufacturing, environmental monitoring, autonomous systems and long-term forecasting, particularly where data needs to be processed continuously on battery-powered or resource-constrained devices.
Titled ‘Spiking Heterogeneous Harmonic Resonate-and-Fire State Space Model’ (SH²RFSSM), the model was developed by researchers from IIT Guwahati’s SustainAI Lab and presented as a poster at ICML 2026.
The research was co-authored by Kartikay Agrawal, Vaishnavi Nagabhushana, Abhijeet Vikram, Vedant Sharma and Ayon Borthakur. Agrawal, Nagabhushana and Borthakur presented the work during the ICML 2026 poster session at the COEX Convention and Exhibition Center in Seoul on July 7.
Dr Ayon Borthakur, Assistant Professor at the Mehta Family School of Data Science and AI, said modern AI systems increasingly rely on analysing long streams of sequential data, including health signals from wearable devices, environmental sensor readings, industrial monitoring data, and weather and traffic forecasts.
“However, widely used AI architectures often become computationally expensive as the length of data increases, making them less suitable for battery-powered and resource-constrained devices,” he said.
The IIT Guwahati team developed SH²RFSSM to address this challenge by combining spiking neural networks with state space modelling. Spiking neural networks are designed to mimic the event-driven communication of biological neurons, activating when meaningful changes or events occur rather than continuously processing information.
Kartikay Agrawal, a PhD research scholar at the Mehta Family School of Data Science and AI, said the approach allows for sparse and energy-efficient computation while enabling the model to learn long-range patterns.
“Unlike conventional neural networks that continuously process information, spiking neural networks activate only when meaningful events occur, enabling sparse and energy-efficient computation,” Agrawal said.
A key feature of the model is neuronal heterogeneity, in which individual artificial neurons are allowed to have different characteristics rather than behaving identically. According to the researchers, this helps the model capture complex temporal patterns found in real-world sequential data.
The researchers evaluated the model on 17 benchmark datasets covering long-range sequence classification, regression, human activity recognition and long-term forecasting.
They reported that the model achieved performance comparable to state-of-the-art sequence models while showing substantially lower estimated energy consumption.
Vaishnavi Nagabhushana, a PhD research scholar at the Mehta Family School of Data Science and AI, said the team plans to further examine the model’s potential in real-world applications involving continuous and long-range data processing.
The researchers said their focus will be on improving the model’s efficiency and adaptability for deployment on resource-constrained devices, with potential applications in edge AI.
Also Read: Explained: Why Eastern Nagaland needed its own governance model