{"type":"ScholarlyArticle","title":"eNNcode: Optimization-Based Analysis of Neural Networks","authors":[{"name":"Muhammad Sabri Farag Atallah","slug":"muhammad-sabri-farag-atallah","given":"Muhammad Sabri Farag","family":"Atallah","affiliation":""},{"name":"Lukas Dankwart","slug":"lukas-dankwart","given":"Lukas","family":"Dankwart","affiliation":""},{"name":"Daniel Neider","slug":"daniel-neider","given":"Daniel","family":"Neider","affiliation":""},{"name":"Mustafa Yalçıner","slug":"mustafa-yalc-ner","given":"Mustafa","family":"Yalçıner","affiliation":""}],"abstract":"The increasing use of neural networks (NNs) in high-stakes decision-making requires rigorous analysis to ensure safety, fairness, and explainability. \r\nFormal verification tools for neural networks typically focus on determining the satisfiability of properties such as safety or fairness. \r\nHowever, many fairness and explainability tasks go beyond satisfiability and instead rely on arbitrary optimization objectives.\r\nTo address such problems, neural networks are often encoded as mixed-integer linear optimization (MILO) problems with linear objectives. \r\nIn practice, these encodings are usually implemented in an ad-hoc manner, limiting comparability across works, reducing transparency, and increasing implementation effort.\r\nWe address this gap by introducing eNNcode, a user-friendly PyPI library that converts any piecewise-linear neural network in Open Neural Network Exchange (ONNX) format into a MILO instance.\r\nThe library supports arbitrary constraints on input and output nodes, as well as user-defined optimization objectives. \r\nOur experiments show that eNNcode achieves performance comparable to existing libraries, despite its simplicity and ease of use.\r\nOverall, eNNcode facilitates reproducible and standardized optimization-based analysis of neural networks.","doi":"10.24963/ijcai.2026/956","url":"https://ijcai.org/proceedings/2026/956","pdf":"https://ijcai.org/proceedings/2026/0956.pdf","markdown":"https://ijcai.org/proceedings/2026/956.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/956","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"Demo Track","track":"Demo Track","number":956,"pages":{"first":8358,"last":8362},"keywords":["AI: Constraint Satisfaction and Optimization"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}