ARC@ORU Meeting AI Challenges in Automation with Recent Advances in Hyperdimensional Computing

Venue: ARC, Örebro University

Remote participation: Join by this Zoom link

No registration is needed.

About the Seminar

This talk examines key challenges associated with the use of artificial intelligence in automation and discusses how selected recent results in hyperdimensional computing may help address them. As an illustrative use-case we will use Cognitive Management in next-generation mobile networks. The underlying challenges and insights are also relevant to broader automation settings.

As mobile networks become increasingly dynamic, heterogeneous, distributed, and service-driven, network automation must evolve beyond predefined, rule-based procedures. Future management systems need to observe complex operational conditions, learn from continuously changing data, reason about system states, and support timely and reliable decisions. The use of conventional AI techniques for these tasks, however, raises several practical challenges. These include limited transparency and explainability, difficulties in validating and trusting model outputs, adaptation to changing operating conditions, robustness to noisy or incomplete data, and high computational and energy requirements.

Such challenges are particularly important in near-real-time management, where decisions must be produced under strict latency, reliability, memory, and resource constraints. In this context, recent developments in hyperdimensional computing offer potentially useful properties, including distributed and compositional representations, computationally lightweight operations, robustness to noise and uncertainty, and efficient incremental or online learning.

The talk will present selected recent results in hyperdimensional computing and will relate them to the requirements of Cognitive Network Management. The objective is not to position HDC as a universal replacement for established AI methods, but to examine where it can complement them and help enable trustworthy, resource-efficient, and time-critical automation in next-generation mobile networks.

Speaker

Evgeny Osipov is a full professor in Dependable Communication and Computation Systems at LTU. 

His main research interests are within Artificial Intelligence. His area of expertise is Vector Symbolic Architectures, also known as  hyperdimensional computing. VSA is a computing framework for explaining  and expressing AI functionality from the standpoints of mathematical properties of random hyperdimensional spaces. It serves as a bridge between the symbolic and connectionist AI approaches and is prospected as one of the main candidates for creating artificial general intelligence.