LEARN: Learning Efficient Automated Reasoning on the Net

Introduction

Certified computer systems are becoming the key in the increasingly complex decision-making activities of our modern society. Among others, provably correct/secure solutions within Artificial Intelligence (AI), Autonomous Systems, Big Data, Blockchain, Decentralized Finance (DeFi), Cloud Computing, or Machine Learning (ML) are indispensable in the ongoing digitalization of our society.

The Paradox of Computer Systems

While the explosion in applications of computer systems leads to great increases in productivity, wealth, and convenience, it creates a paradoxical situation: we rely on computer systems despite the fact that countless application scenarios showcase that computer systems are not (properly) certified and hence are error-prone.

The Impact of Software Errors

Unfortunately, a single software error can quickly escalate into a big issue with huge economic damage. Software errors could be prevented by rigorous code reviews in the software logic. However, who can tell software developers which logical formalism should be used?

Certification Mechanisms

Moreover, which certification mechanism is best to be used during code review for ensuring system safety and security?

LEARN's Solution

LEARN answers these questions by automating reasoning about software and discovering proof strategies for software certification.

Features of LEARN

• Introduces a web-based framework for learning and predicting logic models of software properties.
• Discovers collections of proof strategies without relying on a server back-end, by automating inductive reasoning over commonly used software data structures.

Cost Reduction

LEARN eliminates massive costs arising from correcting defective software releases by proving an easy-to-deploy certification platform for developers and organizations.