Researchers at the Indian Institute of Technology Guwahati (IIT Guwahati) are developing a novel brain-inspired Artificial Intelligence (AI) model designed to process long sequences of data efficiently while consuming significantly less energy than many conventional AI approaches.
The research was carried out by researchers from the Mehta Family School of Data Science and Artificial Intelligence at IIT Guwahati and presented at the International Conference on Machine Learning (ICML) 2026 in Seoul, South Korea. ICML is a top-tier, CORE A*-ranked international conference in AI.
The model, titled Spiking Heterogeneous Harmonic Resonate-and-Fire State Space Model (SH²RFSSM), was presented as a poster at ICML 2026.
The research was co-authored by Kartikay Agrawal, Vaishnavi Nagabhushana, Abhijeet Vikram, Vedant Sharma and Ayon Borthakur. The work was presented at the ICML 2026 poster session at the COEX Convention and Exhibition Center in Seoul by Kartikay Agrawal, Vaishnavi Nagabhushana and Ayon Borthakur on July 7, 2026.
Addressing energy challenges
The research focuses on applications involving continuous and long-range data processing, where computational requirements can increase significantly as the length of data grows.
IIT Guwahati, Assistant Professor, Mehta Family School of Data Science and AI, Dr Ayon Borthakur, 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 or traffic forecasts.
However, he said widely used AI architectures can become computationally expensive as data length increases, making them less suitable for battery-powered and resource-constrained devices.
To address this challenge, the researchers developed SH²RFSSM, a brain-inspired AI architecture that mimics the event-driven communication of biological neurons.
Spiking neural networks
IIT Guwahati, PhD Research Scholar, Mehta Family School of Data Science and AI, Kartikay Agrawal, said conventional neural networks continuously process information, whereas spiking neural networks activate only when meaningful events occur, enabling sparse and energy-efficient computation.
The researchers combined this principle with advanced state space modelling, allowing the system to learn long-range patterns without the heavy computational cost associated with traditional sequence models.
A distinctive feature of SH²RFSSM is neuronal heterogeneity, under which individual artificial neurons can possess different characteristics rather than behaving identically. According to the researchers, this diversity improves the model’s ability to capture complex temporal patterns in real-world sequential data.
Potential for edge AI
The researchers evaluated the model across 17 benchmark datasets covering long-range sequence classification, regression, human activity recognition and long-term forecasting.
The model delivered performance comparable to state-of-the-art sequence models while demonstrating substantially lower estimated energy consumption, according to the research findings.
The researchers said the approach could have applications in wearable health monitoring systems, Internet of Things (IoT) sensors, smart manufacturing, environmental monitoring, autonomous systems and long-term forecasting.
IIT Guwahati, PhD Research Scholar, Mehta Family School of Data Science and AI, Vaishnavi Nagabhushana, said the team plans to further explore the model’s potential in real-world applications involving continuous and long-range data processing.
The focus, she said, will be on improving its efficiency and adaptability so that such AI systems can be deployed more effectively on resource-constrained devices and support practical edge AI applications.
By reducing computational requirements, the researchers believe the technology could help extend battery life and enable AI processing directly on devices, reducing reliance on cloud computing.