IIT Guwahati’s SustainAI Lab develops SH²RFSSM, a brain-inspired AI model that processes long data sequences with substantially lower energy than conventional AI.
A brain-inspired solution for long data streams
Researchers at the Indian Institute of Technology Guwahati have developed a brain-inspired Artificial Intelligence model that can process long sequences of data with lower energy requirements than many conventional AI systems. The SustainAI Lab at IIT Guwahati’s Mehta Family School of Data Science and Artificial Intelligence carried out the research and presented it at the International Conference on Machine Learning (ICML) 2026 in Seoul, South Korea.
The model, named Spiking Heterogeneous Harmonic Resonate-and-Fire State Space Model (SH²RFSSM), combines spiking neural networks with state space modelling to handle long-range patterns while reducing computational demands. Titled ‘Spiking Heterogeneous Harmonic Resonate-and-Fire State Space Model’ (SH²RFSSM), the AI model combines principles inspired by the event-driven functioning of biological neurons with advanced state space modelling.
Why long-sequence AI needs to become more efficient
Modern AI systems increasingly rely on analysing long streams of sequential data such as health signals from wearable devices, environmental sensor readings, industrial monitoring data, and weather or 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.
To address this challenge, the IIT Guwahati team developed SH²RFSSM, a brain-inspired AI architecture that mimics the event-driven communication of biological neurons. The research has significant potential across sectors where continuous data analysis must be performed efficiently. These include wearable health monitoring systems, Internet of Things (IoT) sensors, smart manufacturing, environmental monitoring, autonomous systems, and long-term forecasting applications.
How SH²RFSSM works: spiking neurons and state space models
Unlike conventional neural networks that continuously process information, spiking neural networks activate only when meaningful events occur, enabling sparse and energy-efficient computation. The team 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 the model is neuronal heterogeneity, where individual artificial neurons are allowed to possess different characteristics rather than behaving identically. This design improves its ability to capture complex temporal patterns in real-world data. The architecture combines spiking neural networks (SNNs) with state space modelling, where SNNs process information through event-driven activation rather than continuously computing every input, enabling sparse computation.
Benchmarking across 17 datasets
The researchers evaluated the AI model across 17 benchmark datasets spanning 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, making it particularly promising for edge AI applications, the research team asserted.
The model was tested on 17 benchmark datasets covering long-range sequence classification, regression, human activity recognition and long-term forecasting. The researchers said SH²RFSSM achieved performance comparable to leading sequence models while recording substantially lower estimated energy consumption.
Team and presentation at ICML 2026
The research was co-authored by Kartikay Agrawal, Vaishnavi Nagabhushana, Abhijeet Vikram, Vedant Sharma, and Ayon Borthakur. The work was presented during the ICML 2026 poster session at the COEX Convention and Exhibition Centre in Seoul by Kartikay Agrawal, Vaishnavi Nagabhushana, and Ayon Borthakur on July 7.
Elaborating on the need for such research, Dr Ayon Borthakur, Assistant Professor, Mehta Family School of Data Science and AI, IIT Guwahati, said modern AI systems increasingly rely on analysing long streams of sequential data, but widely used AI architectures often become computationally expensive as the length of data increases. Highlighting the unique aspects of this model, Kartikay Agrawal, PhD Research Scholar, explained that spiking neural networks activate only when meaningful events occur, enabling sparse and energy-efficient computation.
Enabling AI on the edge
In these fields, reducing computational load can extend battery life and enable AI directly on devices without relying heavily on cloud computing. By combining spiking neural networks with state space modelling, SH²RFSSM offers a pathway to deploy capable sequence models on wearables, sensors and embedded systems that cannot afford continuous heavy computation.
The model’s lower energy footprint also reduces heat generation and infrastructure costs, which matters for large-scale deployments in smart cities, industrial plants and environmental monitoring networks.
Implications for wearables, IoT and autonomous systems
The research opens new possibilities for wearable devices, IoT sensors and autonomous systems that must process continuous streams of data under tight power budgets. In healthcare, such models can analyse heart-rate, ECG and activity signals on-device, preserving privacy and reducing latency.
In manufacturing and infrastructure monitoring, they can detect anomalies in vibration, temperature and pressure data in real time. For environmental applications, they can process long time series from air-quality, water-quality and weather sensors to support early warnings and adaptive control.
A step towards sustainable AI
As AI adoption grows, energy efficiency has become a critical constraint. SH²RFSSM demonstrates that brain-inspired designs can deliver strong performance on long-sequence tasks while using substantially less energy than conventional approaches. This aligns with broader efforts to make AI more sustainable and accessible on resource-constrained hardware.
By presenting the work at ICML 2026, one of the leading international conferences in artificial intelligence, the IIT Guwahati team positioned SH²RFSSM within the global conversation on efficient sequence modelling and neuromorphic computing.
Looking ahead
The researchers plan to extend SH²RFSSM to more complex multimodal tasks and explore hardware implementations that fully exploit its event-driven nature. As edge AI matures, such models can help shift intelligence from data centres to devices, enabling faster, cheaper and more private AI applications.
With SH²RFSSM, IIT Guwahati’s SustainAI Lab shows that brain-inspired AI can move beyond theory and deliver practical gains for long-range data processing in the real world.
#IITGuwahati #BrainInspiredAI #SH2RFSSM #SpikingNeuralNetworks #StateSpaceModels #EnergyEfficientAI #EdgeAI #ICML2026 #SustainAILab #LongSequenceModelling
Disclaimer
The information in this article is based on available public sources and official statements as of the time of publication. While we aim for accuracy, we do not guarantee completeness or correctness. We advise readers to verify key details from official sources before making any decisions. The website (iitiimsamvaad.com) is not liable for any loss or damage arising from the use of this content. The authors are also not responsible for any such loss or damage.