<div dir="ltr"><div dir="ltr"><div dir="ltr"><div style="color:rgb(31,31,31);font-family:Roboto,Helvetica,Arial,sans-serif;font-size:14px;letter-spacing:0.2px">Dear COIN-OR Community,<br><br>We are pleased to announce that <a href="https://github.com/coin-or/balans">Balans solver</a> has now officially joined the COIN-OR Foundation. <br><br><b>B</b>andits-based <b>A</b>daptive <b>LA</b>rge <b>N</b>eighborhood <b>S</b>earch (Balans) is an online meta-solver for solving Mixed-Integer Programming (MIP) problems. It utilizes adaptive large neighborhood search (ALNS), guided by multi-armed bandits (MAB), operating on top of a MIP solver. More details are available in our <a href="https://www.ijcai.org/proceedings/2025/286">IJCAI'25</a> paper and <a href="https://nbviewer.org/github/skadio/skadio.github.io/blob/master/files/2025_IJCAI_Balans_Kadioglu.pdf">slides</a>. <br><br>The project is open-source under the Apache 2.0 License and now lives at: <a href="https://github.com/coin-or/balans">https://github.com/coin-or/balans</a><br><br>Install Balans directly via pip:
<span style="color:rgb(31,35,40);font-family:"Monaspace Neon",ui-monospace,SFMono-Regular,"SF Mono",Menlo,Consolas,"Liberation Mono",monospace;font-size:13.6px;letter-spacing:normal;background-color:rgba(129,139,152,0.12)">pip install balans</span> and here is a <a href="https://github.com/coin-or/balans?tab=readme-ov-file#quick-start">Quick Start Example</a> with key features summarized below. <br><br>We look forward to your feedback and contributions! <br><br>Best regards, </div><div style="color:rgb(31,31,31);font-family:Roboto,Helvetica,Arial,sans-serif;font-size:14px;letter-spacing:0.2px">Serdar Kadioglu & Bistra Dilkina<br>On behalf all the Balans development team<br><br>=== <br><br>Balans Solver Key features:<br><ul><li><b>Solver-Agnostic:</b> Supports SCIP & Gurobi out-of-the-box, and any MIP solver implementing a fairly standard <a href="https://github.com/coin-or/balans/blob/main/balans/base_mip.py">abstract base MIP class</a>. </li><li><b>Flexible:</b> Running a single destroy/repair neighborhood recovers LNS(MIP). The API allows you to choose which neighborhoods to include, reward mechanism, and learning strategy. </li><li><b>Extensible: </b>Built on top of the open-source <a href="https://github.com/fidelity/mabwiser">MabWiser library</a> (from Fidelity Investments) and the open-source <a href="https://github.com/N-Wouda/ALNS">ALNS library</a> (from RoutingLab), it remains highly modular and configurable.</li><li><b>Parallelizable:</b> ParBalans allows controlling parallel Balans configurations and parallel MIP solving.</li><li><b>Integration Technology: </b>Broadly, Balans combines adaptive search, meta-heuristics, multi-armed bandits, and mixed integer programming in a unified framework. </li><li><b>Contributions:</b> New MIP solver integrations, new destroy/repair neighborhoods, and new learning strategies among others, are most welcome! </li></ul></div></div>
</div>
</div>