Publications

Project Goal: Develop a novel neuromorphic hardware platform utilizing combined photonic and electronic spiking neurons for ultra-low energy and high-speed artificial intelligence applications.

Challenges Addressed:

• Current AI models require increasingly high computational resources, leading to limitations in conventional digital hardware.
• Existing spiking neural networks (SNNs) based on electronics face limitations in speed and energy efficiency.

Proposed Solution:

SPIKEPro will create a first-of-its-kind integrated circuit (IC) platform combining:

• All-optical spiking neurons: Utilizing micron-sized lasers for ultrafast and energy-efficient spike generation.
• Electronic spiking neurons: Employing resonant-tunneling diode (RTD) devices for low-energy electrical spiking.
• Non-volatile synaptic weights: memristive devices offering fast switching and compatibility with the electronic-photonic integration platform.

Benefits:

• Ultra-low energy consumption: Targeting sub picojoule range for photonic and picojoule range for electronic spiking events.
• High-speed processing: Enabling operation at picosecond to nanosecond timescales.
• Multimodal data processing: Handling both electrical and optical input signals.

Potentiaal Applications:

• Energy-efficient AI systems for Internet-of-Things (IoT), smart buildings, and autonomous systems.
• High-performance computing with SNN co-processing alongside digital neuromorphic hardware.

Expected Impact:

• Transform AI systems into environmentally sustainable technologies.
• Advance neuromorphic hardware with significant societal and economic benefits.

Overall, SPIKEPro presents a groundbreaking approach to ultra-low energy and high-speed SNN hardware, paving the way for a new era of efficient and powerful AI systems