Heuristic Algorithms

Heuristics

 A type of Approximation Algorithm

 Key Features of heuristic algorithms:
• No guarantee of near-optimal solution in all
scenarios
• But, in real life, provides excellent results with
relatively short running time
• Many applicable areas, especially useful for
problems with very large scope

2
 Definition
 Heuristic: Derived from Greek word,
heuriskien, means “to search, to discover”
 From Wikipedia:
“A heuristic is a technique designed to solve a problem
that ignores whether the solution can be proven to be
correct, but usually produces a good solution”
“Heuristics are intended to gain computational
performance or conceptual simplicity, potentially at
the cost of accuracy or precision”.

3
 Metaheuristics
 Metaheuristics is a high level framework that
provides guidelines to develop heuristic algorithms
for a large number of problems
 Applicable areas include:
• Engineering Design in VLSI, aerodynamics,
robotics, fluid dynamics and others
• Genetic and Evolutionary Algorithms (key elements
of Machine Learning)
• System Modelling in control, signal and image
processing
• Scheduling, Logistics & Transportation, Route
planning
4
 Graph Search
 Route Planning: Find the shortest route between
two locations in a city (Google Map?)
 System Modelling: Use a graph to model the path
and apply shortest path algorithms
 Shortest Path Algorithms
• Un-weighted graph: BFS
• Weighed graph: Dijkstra
 Disadvantage: In large graphs, evaluation of all
alternate paths (exhaustive search) takes a long
time to complete

5
 Best First Search
 Best First Search avoids exhaustive search by
applying a heuristic:
• A measure of distance of the target from the current
position
 Method of Computing distance:
• Euclidean Method
• Manhattan Method
 Easy to compute measures, but is an
approximation of the actual distance

6
 Best First Search
 Algorithm Approach:
1. At every node, derive the following:
f(n) = h(n), where
h(n) is estimated distance to the target
2. Choose the node with minimum value of f(n)
and proceed till target is reached

 Helps in arriving at the target much faster than

exhaustive search methods

7
 Example

What is the fastest way to reach G from S?

1st Road S

2nd Road
G

3rd Road

9th Lane
3rd Lane

5th Lane

6th Lane

7th Lane
4th Lane

8th Lane
1st Lane

2nd Lane

8
 Dijkstra
Go through all possible combinations …

1st Road S

2nd Road
G

3rd Road

4th Road

5th Road

9th Lane
3rd Lane

4th Lane

5th Lane

6th Lane

7th Lane

8th Lane
1st Lane

2nd Lane

9
 Best First Search
Reach G from S using a short route …

1st Road S

2nd Road
G

3rd Road

9th Lane
3rd Lane

5th Lane

6th Lane

7th Lane
4th Lane

8th Lane
1st Lane

2nd Lane

10
 Best First Search
Blocked by an obstacle … Alternate path?

1st Road S

2nd Road
G

3rd Road

9th Lane
3rd Lane

5th Lane

6th Lane

7th Lane
4th Lane

8th Lane
1st Lane

2nd Lane

11
 Best First Search
 Advantages:
Fast

 Disadvantages:
May not give optimal result when obstacles are in
the way from source to target

12
 A* Search
 More refined heuristic:
Minimises the function:
f(n) = g(n) + h(n), where
g(n) is actual distance from source
h(n) is estimated distance to target
 Finds an optimal path to the target, even in
presence of obstacles
 It is a Fast Algorithm

13
 14

9th Lane
Optimised path around the obstacle

8th Lane
7th Lane
6th Lane
A* Search

5th Lane
4th Lane
3rd Lane
2nd Lane
S
1st Lane

3rd Road
2nd Road
1st Road
 Summary

Search Technique Characteristics

Breadth First Search Optimal but slow

Dijkstra Optimal, slow
Best First Fast, no always optimal
A* Optimal and Fast

15
 Applicable Areas
 Route Planning – computer networks,
airlines
 VLSI Layout – cell layout & routing
 Production Planning
 Protein Sequence Alignment
 Many NP Complete problems – where no
exact algorithms exist, heuristic search is
the only option available

16
The End

17