- OctaiPipe’s newly-introduced Federated XGBoost algorithm cuts communication rounds between servers and customers by 66% compared to other Federated Learning implementations, while maintaining model accuracy.
- OctaiPipe launches V2.2 of its federated edge AI platform to enhance model monitoring, continuous distributed learning, platform reliability, security, and cost-effectiveness.
London, UK – May 7th, 2024: OctaiPipe, a Federated Edge AI platform for Industrial IoT, has announced the release of Version 2.2 of its Federated Learning Operations (FL-Ops) platform. This latest version introduces a significant improvement, Federated XGBoost (Extreme Gradient Boosting), which cuts communication rounds between servers and customers by 66% compared to other Federated Learning implementations, while maintaining model accuracy.
Launched in 2022, OctaiPipe’s FL-Ops platform has rapidly become the go-to solution for Critical Infrastructure OEMs, device manufacturers, and Industrial IoT operators, including P&G, ARUP, Domin and SICK, to simplify distributed learning models, ensure device security and improve the performance of critical systems.
Federated learning, a decentralised machine learning approach, allows AI algorithms to be trained on-device at the edge, minimising data movement to the cloud. With this latest version, OctaiPipe enables data scientists to exploit the benefits of Federated Learning with tree based models, reducing CPU, memory, and storage usage while allowing the deployment of AI to the edge at scale. OctaiPipe’s newly-introduced Federated XGBoost algorithm offers unparalleled speed, accuracy, and efficiency, outperforming deep learning approaches while using fewer resources and producing more explainable results.
In addition to Federated XGBoost, Version 2.2 introduces OctaiOxide, an alternative container for machine learning workloads, enabling model inferences at native speeds with minimal RAM usage. This release also includes enhancements to model monitoring, continuous distributed learning, platform reliability, security, and cost-effectiveness.
Eric Topham, CEO of OctaiPipe, stated, “OctaiPipe addresses the critical need to train, deploy, and manage on-device federated learning at scale. Our Federated XGBoost algorithm reduces communication rounds by 66% compared to other Federated Learning implementations currently available in the market, lowering cloud dependency while maintaining model accuracy, resulting in AI that’s efficient and trusted.
For more information on OctaiPipe and Federated Learning for Edge AI for IoT, download the whitepaper here.
About OctaiPipe:
OctaiPipe is a end-to-end Edge AI platform optimised for creating, deploying, and managing machine learning solutions in critical infrastructure environments. Deployments on OctaiPipe are more affordable, secure, scalable, and resilient for on-device intelligence due to its advanced on-device Federated Learning capabilities, as well as innovative edge MLOps technology.
With its cost-efficient, accelerated and federated AI, OctaiPipe is removing the common barriers to market while maximising security and privacy, and minimising Cloud dependency and Cloud costs. OctaiPipe is headquartered in London, UK and is funded by a portfolio of investors including SuperSeed, Forward Partners, D2, Atlas Ventures, Growceanu Angel Investment, Kinled Holding, Martlet Capital, Gelecek Etki VC and Arm-backed Deeptech Labs.
