Prof. Yi Mei

Prof. Yi Mei

Fellow, Engineering NZ IEEE Senior Member

Professor researching evolutionary computation, machine learning and optimisation

  • Location Wellington, New Zealand
  • Positions Professor
    Programme Director (Computer Science and Computer Graphics)
    School of Engineering and Computer Science (SECS)
    Centre for Data Science and Artificial Intelligence (CDSAI)
    Victoria University of Wellington (VUW)
  • Research groups Evolutionary Computation for Combinatorial Optimisation Group (ECCO)
    Evolutionary Computation and Machine Learning Research Group (ECRG)
    IEEE CIS Taskforce on Evolutionary Scheduling and Combinatorial Optimisation (TESCO)
  • Connect
  • yi.mei@ecs.vuw.ac.nz
  • Google Scholar
  • ORCID
  • ResearchGate
  • LinkedIn
  • University profile

Resources

Open-source code, benchmark datasets and tutorials from my research, followed by external libraries and reading I recommend.

My Code

Implementations of algorithms from my papers. Please cite the corresponding paper if you use them.

MAENS C++

Memetic Algorithm with Extended Neighbourhood Search for the capacitated arc routing problem, a widely used benchmark algorithm with the large-step Merge-Split operator.

GitHub stars GitHub forks Repository · Paper (IEEE TEVC 2009)
GPJSS Java

Genetic programming hyper-heuristics for automatically evolving dispatching rules for (dynamic) job shop scheduling, built on the ECJ library.

GitHub stars GitHub forks Repository · Research theme
gpucarp Java

Meta-heuristics and genetic programming hyper-heuristics that evolve routing policies for the uncertain capacitated arc routing problem; includes the gdb, val and egl datasets.

GitHub stars GitHub forks Repository
JourneyPlanner Java

Earliest-arrival public transit journey planner based on round-based public transit routing (RAPTOR), with GTFS import and an Adelaide example.

GitHub stars GitHub forks Repository

My Datasets

Capacitated arc routing (CARP) benchmark instances used in my research, free to download.

  • gdb23 small instances (~30 nodes, ~60 required edges).
  • val10 groups of medium to large instances (~50 nodes, ~100 required edges); each group has 3–4 instances (A–D) on the same graph with different vehicle capacities.
  • egl8 groups of large instances (~150 nodes, ~200 required edges): e1–e4 and s1–s4 use the same graph with different required edges; 3 capacities (A–C) each.
  • EGL-G2 groups of large instances (~250 nodes, ~400 required edges), 5 capacities (A–E) each.
  • Beijing & HefeiTwo large datasets generated from the road networks of Beijing and Hefei, 10 instances each with thousands of edges.
  • BeullensThe Beullens et al. CARP benchmark instances.
  • Large-scale instancesLarge-scale CARP instances, with the best solutions we found.
  • Multi-depot CARPMulti-depot capacitated arc routing instances.

Tutorials & Talks

tutorials Jupyter Notebook

Jupyter notebook tutorials on artificial intelligence, evolutionary computation and optimisation/operations research, such as Bayesian networks, planning and scheduling, and genetic programming.

GitHub stars GitHub forks Repository

Libraries & Frameworks

Third-party software for evolutionary computation, genetic programming and hyper-heuristics.

ECJJavaHighly recommended for GP DEAPPython gplearnPython MicroGPPython InspyredPython JeneticsJava Open BEAGLEC++ HyFlexJavaHyper-heuristic framework EvoHypJavaHyper-heuristic framework EureqaAppSymbolic regression software

Benchmark Libraries

Further Reading