![[2602.21546] Mamba Meets Scheduling: Learning to Solve Flexible Job Shop Scheduling with Efficient Sequence Modeling](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.21546] Mamba Meets Scheduling: Learning to Solve Flexible Job Shop Scheduling with Efficient Sequence Modeling
Summary
This article presents a novel approach to the Flexible Job Shop Problem (FJSP) using a state-space model called Mamba, which improves efficiency in scheduling tasks across machines.
Why It Matters
The Flexible Job Shop Problem is critical in manufacturing and production scheduling. By introducing a more efficient model, this research could enhance operational efficiency in various industries, leading to cost savings and improved productivity.
Key Takeaways
- The proposed Mamba model offers linear computational complexity for FJSP.
- It surpasses existing graph-attention-based methods in efficiency and performance.
- The architecture includes a dual Mamba block for feature extraction and an efficient cross-attention decoder.
- Experimental results indicate faster solving speeds and improved outcomes compared to state-of-the-art methods.
- This research has significant implications for optimizing scheduling in manufacturing.
Computer Science > Machine Learning arXiv:2602.21546 (cs) [Submitted on 25 Feb 2026] Title:Mamba Meets Scheduling: Learning to Solve Flexible Job Shop Scheduling with Efficient Sequence Modeling Authors:Zhi Cao, Cong Zhang, Yaoxin Wu, Yaqing Hou, Hongwei Ge View a PDF of the paper titled Mamba Meets Scheduling: Learning to Solve Flexible Job Shop Scheduling with Efficient Sequence Modeling, by Zhi Cao and 4 other authors View PDF HTML (experimental) Abstract:The Flexible Job Shop Problem (FJSP) is a well-studied combinatorial optimization problem with extensive applications for manufacturing and production scheduling. It involves assigning jobs to various machines to optimize criteria, such as minimizing total completion time. Current learning-based methods in this domain often rely on localized feature extraction models, limiting their capacity to capture overarching dependencies spanning operations and machines. This paper introduces an innovative architecture that harnesses Mamba, a state-space model with linear computational complexity, to facilitate comprehensive sequence modeling tailored for FJSP. In contrast to prevalent graph-attention-based frameworks that are computationally intensive for FJSP, we show our model is more efficient. Specifically, the proposed model possesses an encoder and a decoder. The encoder incorporates a dual Mamba block to extract operation and machine features separately. Additionally, we introduce an efficient cross-attention decoder to l...
