Download presentation
Presentation is loading. Please wait.
Published byBennett Todd Modified over 9 years ago
1
PSO implementation on Hadoop and standalone
2
Flowchart1 Inner loop outer loop Output Total gbest … … Initial particles (random) Split1 of particles SplitN of particles gbest1 gbestN Total gbest split
3
Flowchart1 Map phase Reduce phase Inner loop outer loop Output Total gbest … … Initial particles (random) Split1 of particles SplitN of particles gbest1 gbestN Total gbest split Inner loop
4
Flowchart2 Inner loop outer loop Output Total gbest … … Pseudo randomly initialize particles Split1 of particles SplitN of particles gbest1 gbestN Total gbest
5
Flowchart2 Inner loop outer loop (feedback total gbest) Output Total gbest … … Pseudo randomly initialize particles Split1 of particles SplitN of particles gbest1 gbestN HDFS Map phase
6
Experiment results # of lamps : 12 targets : 5 * 1 light ranks : 16 Differ_rate : 0.0001 # of splits : 4
7
Experiment results # of lamps : 12 targets : 5 * 1 light ranks : 16 Differ_rate : 0.0001 # of splits : 8
8
Experiment results # of lamps : 396 targets : 90 * 33 light ranks : 16 Differ_rate : 0.0001 # of splits : 4
9
Experiment results # of lamps : 396 targets : 90 * 33 light ranks : 16 Differ_rate : 0.0001 # of splits : 8
10
Exchange of particles Assume that we have an example: # of splits is 4, # of particles in each split is 2500, and we assign 2500/2(ideal) particles to exchange between splits. Under this condition, each split is responsible for emitting(pseudo randomly) 1250/4 particles to exchange. Constraints: # of particles in each split ≥ ( # of splits -1)* # of splits *2
11
Exchange of particles ID00 103104 207208 311 ID10 103104 207208 311 ID20 103104 207208 311 ID30 103104 207208 311 ID0[0…103] ID0[104…207] ID0 ID0[208…311] ID1[0…103] ID1[104…207] ID1 ID1[208…311] ID2[0…103] ID2[104…207] ID2 ID2[208…311] ID3[0…103] ID3[104…207] ID3 ID3[208…311] …… ID3[0…103],ID2[104…207],ID1[208…311] ……
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.