Formal Learning Systems

Three key scientific problems:

how to design learning models that satisfy rigorous formal specifications and guarantees

how to maintain robustness, generalization, and adaptability under uncertainty and environment changes without violating these guarantees

how to unify formal methods with data-driven learning to achieve scalable, interpretable, and deployable intelligent systems.

Module resources

Three module pillars

The original three-pillar module area is now upgraded to a video-first layout. Each row presents one externally linked video together with its PDF resource, followed by the related publication list below.

Edit the


       embedUrl

and


       videoUrl

values in the script block at the bottom of this page to load the real external videos.

Selected output

Related publications

Representative papers are pulled from the current publication archive and embedded here so each module page has both visual entry points and a paper list beneath them.

Browse all publications

Active Learning Based Requirement Mining for Cyber-Physical Systems

Gang Chen, Zachary Sabato, Zhaodan Kong
2016 IEEE 55th Conference on Decision and Control (CDC), 4586-4593

Conference

We study requirement mining for cyber-physical systems by combining signal temporal logic with active learning. The proposed GP-ACB algorithm accelerates the search for parametric temporal logic requirements by selecting informative samples, leading to faster convergence than existing Gaussian-process-based alternatives.

Data-Driven Approximate Abstraction for Black-Box Piecewise Affine Systems

Gang Chen, Zhaodan Kong
Preprint

Preprint

We develop a data-driven algorithm to construct approximate abstractions for black-box piecewise affine systems by integrating system identification, abstraction, and active sampling under temporal-logic specifications with bounded error and probability guarantees.

Requirement Mining from Closed-Loop Control Models via Human-Computer Collaboration

Penghong Lu, Gang Chen
2024 4th International Conference on Computer, Control and Robotics (ICCCR), Shanghai, China, 2024

Conference

Position in the lab

How this module fits the SRIS portfolio

Module identity

Formal Learning Systems is presented as a reusable research block that can connect to papers, projects, talks, and demonstrations across the website.

Typical outputs

This module can aggregate theory papers, application case studies, tutorial materials, project narratives, and visual evidence under one stable page structure.

Cross-links

Use the quick-access tiles below to move from this module page to research overviews, team pages, publication lists, and lab visuals.

Related themes