Graph Algorithm Efficient Shortest Path Estimation Mentee: Yonk - - PowerPoint PPT Presentation

Graph Algorithm – Efficient Shortest Path Estimation

Mentee: Yonk Shi (CSE, Moorpark College) Mentor: Arijit Khan (CS, UCSB) Faculty Advisor: Dr. Xifeng Yan Computer Science Department of UCSB INSET Program

1

Shortest Path Algorithm

Simple case:

• We are trying to develop a general

algorithm for graph navigation

• It will work with any dataset, i.e. Google,

Facebook, Last.fm

• It is optimized for massive databases
• It is extremely efficient regardless of the

size of the graph.

Reality:

Source: my facebook, www.facebook.com/yonkshi A B

Algorithm

MDS: Multidimensional Scaling

Dimensionless Data 2 Dimensional Data

• Preserved Distances
• Preserved Paths

A B

S1 S2

A B

S1 S2

Algorithm

In reality, MDS generates an approximation of coordinates, thus the distance is approximated

Experimental Results

1 2 3 4 5 6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Distance

Distances Calculated by Different Algorithms

MDS-distance MDS-Dij-Distance Real DIjsktra Distance Our Algorithm

Experimental Results

1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Steps

Steps Taken by Different Algorithms

Our Algorithm Dijsktra

10% Failure Rate

Experimental Results

As much as 3000x Faster than Dijkstra’s Algorithm

5 10 15 20 25 30 35

Average Time Taken by Algorithms

Our Algorithm Dijkstra's

Conclusion

• We have designed a shortest path algorithm
• It is very efficient and accurate for large databases
• It is much faster than Dijkstra’s Algorithm

Our future goals:

• Reduce failure rate to 0% (while maintaining accuracy)
• Increase high efficiency and accuracy
• Add “Label” information for even more accurate search

Thank You!

Special Thanks To: My mentor Arijit Khan Faculty Advisor Dr Xifeng Yan Professor Christine Aguilera from Moorpark College Professor Martin Chetlen from Moorpark College Awesome Roommates Adam, Michael and Jose

• 0.6
• 0.4
• 0.2

0.2 0.4 0.6 0.8 1

• 1
• 0.5

0.5 1 Path 1340