What is it about?

- The aim of the article is to conceptualise a more compact and efficient version of algorithms for artificial intelligence (AI). - The core objective is to construct the design for a self-optimising and self-adapting autonomous artificial intelligence (AutoAI) that can be applied for edge analytics using real-time data. - This article undertakes experimental developments in research on how AI algorithms can operate on low memory / low computation IoT devices and how AI can be designed and constructed to procreate and write its own algorithms.

Featured Image

Why is it important?

- The new concept of autonomous AI depends on data training preparation for multiple AI challenges (self-evolving, self-procreating, self-optimising and self-adaptive) - If AI algorithms are not trained to take risks and learning from its own experience, then the algorithms are missing the training of experimenting in uncertain environment. - To address this challenge, we need to enable AI to learn by itself by exploration and exploitation.


- The article presents a new ordered pipeline approach, based on integrating a variety of existing methods in an ordered approach, to increase the efficiency of algorithms in low memory / low computation IoT devices. The new ordered pipeline approach builds upon the state-of-the-art literature on AI and IoT devices (i.e., MCUNet, TinyNAS, TinyEngine), existing datasets (e.g., Caltech 101, Caltech 256, ImageNet) and integrates some of the concepts from early literature on AI algorithms (i.e., The Boltzmann Machine, The Helmholtz Machine and The Elman Network). - The design is multidisciplinary as it integrates knowledge and methods from statistics and mathematical sciences, engineering sciences, computer sciences and healthcare disciplines.

Dr Petar Radanliev
University of Oxford

Read the Original

This page is a summary of: Review of Algorithms for Artificial Intelligence on Low Memory Devices, IEEE Access, January 2021, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/access.2021.3101579.
You can read the full text:



The following have contributed to this page