Hyper-Heuristic
ONE-D BIN-PACKING Channel Assignment Problem Travelling Salesman Problems
郭益銘
Hyper-Heuristic
HH vs. MH
Hyper-Heuristic Meta-Heuristic
Objective style General Particular
Problem domain range Wide Narrow (class of problems)
Use Easier Hard
Implement Cheaper Expensive
The goal is to produce good
quality solutions in this more
Goal Particular Problem
general framework
Hyper-Heuristic
Concept
• A hyper-heuristic can be (often is) a
(meta-)heuristic and it can operate on
(meta-)heuristics.
Hyper-Heuristic
1. Accept or reject
the new solution.
2. Decide which LLH
to call at each
iteration.
• Fitness
• Computation Time
Only non-domain
data can pass
through the harrier.
Learning a procedure Channel assignment Self-adaptive
Title that can solve hard bin- optimization using a hyper- Hyperheuristic and
packing problems heuristic Greedy Search
Bin-Packing Problems Channel Assignment Travelling Salesman
Problem
(BPP) Problems (CAP) Problems (TSP)
1. Co-channel constraint 1. No repeatable
2. Adjacent channel 2. All of the cities
Constraint constraint should be visited
3. Co-site constraint once and only once
Hyper- XCS
Heuristic A modern learning classi£er TABU GP
system
H1 - First-Fit-Decreasing (FFD) H1 - Sort the channel from lowest
H2 - Next-Fit-Decreasing (NFD) to highest, delete the call with the
H3 - Djang and Finch’s highest channel assignment,
algorithm (DJD) randomly insert at any point and
Low-Level H1 - NATURAL
H4 - Djang and Finch, more reassign the channel.
H2 - 2-CHANGE
Heuristic tuples(DJT) H2 - Same as H1, but randomly
H3 - IF_2-CHANGE
(LLH) H5 - FFD+Filler select the call to delete.
H4 - IF_3-CHANGE
H6 – NFD + Filler H3 - Same as H1, but randomly
H7 – DJD + Filler select the call to delete.
H8 – DJT + Filler H4 - Same as H1, but randomly
select the call to delete.
ONE-D BIN-PACKING
ONE-D BIN-PACKING
• In its simplest incarnation there is an
unlimited supply of identical bins and there is
a set of objects to be packed into as few bins
as possible.
• Each object has an associated scalar (think of
it as the object’s weight) and a bin cannot
carry more than a certain total weight.
ONE-D BIN-PACKING
Bins Capacity : 524
largest first, first fit’ heuristic
1. 442 46 12 12 12 => 524 1. 442 37 37 =>516
2. 252 252 10 10 => 524 2. 252 252 12 => 516
3. 252 252 10 10 => 524 3. 252 252 12 => 516
4. 252 252 10 10 => 524 4. 252 252 12 => 516
5. 252 127 127 9 9 => 524 5. 252 127 127 10 => 516
6. 127 127 127 106 37 => 524 6. 127 127 127 106 10 10 10 => 517
7. 106 106 106 85 84 37 => 524 7. 106 106 106 85 84 10 10 9 =>516
sum: 3668 8. 9 => 9
sum: 3622
Channel Assignment Problem
Channel assignment in cellular communication
using a great deluge hyper-heuristic
The basic concept of the Great Deluge Hyper-heuristic Algorithm :
Suggestion:
Channel Assignment Problem
Channel assigmnent optimisation using a hyper-heuristic
Travelling Salesman Problems
Self-adaptive Hyperheuristic and
Greedy Search
Travelling Salesman Problems
2-CHANGE primitive
Travelling Salesman Problems