Intel announces a series of new developments to advance neuromorphic computing application development

Intel Labs launched the Kapoho Point development board, updated the Lava framework, and announced new Intel neuromorphic research community support projects.

Intel Labs is leading the development of neuromorphic computing through the Loihi 2 second-generation research chip and open source Lava software framework released last year. As part of Intel's goal of commercializing neuromorphic technology, Intel Labs is providing developers with new tools to advance the development process to the next stage. For example, the 8-chip Loihi 2 development board Kapoho Point can meet the needs of large-scale workloads through stacking and can achieve direct interconnection with low-latency event cameras (event-based vision sensors).

Loihi 2 is Intel's second-generation neuromorphic research chip. It supports new brain-like algorithms and applications, provides faster processing speed and higher resource density, while improving energy efficiency. (Image source: Intel Corporation)

In addition, Intel Research has updated the open source Lava framework to support programmable neurons, integer spike neurons, convolutional networks, and continuous learning. Starting with the latest version of Lava (v0.5), these new features enable Kapoho Point to run deep learning applications 12 times faster and with 15 times less energy than the same workload on the Loihi 1 system1. In addition, Intel has launched eight Intel-sponsored university projects through the Intel Neuromorphic Research Community (INRC).

Delivering the next generation of neuromorphic systems to community members 

Kapoho Point, a Loihi 2-based development board, is a compact system that is ideal for a variety of small-sized devices and applications, from drones to satellites and smart cars. Kapoho Point can run AI models with up to 1 billion parameters and solve optimization problems covering up to 8 million variables. Compared with advanced solvers running on CPUs, it increases the speed by more than 10 times and reduces energy consumption by 1,000 times. In addition, Kapoho Point can be expanded by stacking multiple development boards to solve larger-scale problems. 

The Loihi 2-based development board Kapoho Point is a compact system ideal for a variety of small-form-factor devices and applications, from drones to satellites and smart cars. (Image source: Intel Corporation)

The U.S. Air Force Research Laboratory (AFRL) is the first member of the research community to enable Kapoho Point and is using it for internal research involving learning based on spiking neural networks and problems that require real-time optimization. "As AFRL missions are conducted in the air and space, the space, weight and power budget of mobile platforms are very limited," said Dr. Qing Wu, senior scientist for data processing and development. "Neuromorphic computing technology provides an excellent computing solution for the needs of running AI algorithms in this environment."

Lowering the threshold for neuromorphic development through the Lava software framework

Updates to the open-source, modular and extensible Lava software framework include a range of improvements to the Loihi 2 feature set, such as programmable neurons, graded events and continuous learning.

Intel Labs has updated the open source Lava framework to support programmable neurons, integer spiking neurons, convolutional networks, and continuous learning. (Image source: Intel Corporation)

The Lava software framework can be downloaded for free on GitHub.

Neuromorphic Ecosystem Project

The Intel Neuromorphic Research Community (INRC) has launched eight Intel-supported university projects, including George Mason University, Queensland University of Technology, Graz University of Technology, University of Zurich, Brown University, Pennsylvania State University, University of Waterloo and University of Göttingen.

Research projects include adaptive robot positioning, wireless biomimetic sensor pulse decoding for brain-computer interfaces, neuromorphic Bayesian optimization, auditory feature detection, and novel brain-like architectures and algorithms.

Since its founding in 2018, the Intel Neuromorphic Research Community has grown to more than 180 members, including university labs, government agencies, and leading global companies such as Accenture, Lenovo, Logitech, and Mercedes-Benz.

Going forward, Intel Labs will continue to provide developers with new tools to make it easier for them to develop applications that solve real-world problems and continue to support community research.

1SDNN results for Loihi 2 are based on the Lava v0.5 benchmark conducted in September 2022, when executing a 9-layer PilotNet DNN inference workload as a Sigma-Delta neural network on Loihi 2, compared to the same network executed on Loihi 1 using SNN frequency encoding. Equivalent DNN operation counts from a traditional DNN execution with the same topology and the same number of 8-bit parameters. See Bojarski, Mariusz, et al., End-to-End Learning for Self-Driving Cars. arXiv preprint arXiv:1604.07316 (2016).