WEBVTT
Kind: captions
Language: en

00:00:14.420 --> 00:00:18.060
Hello everybody. Today, I want to introduce

00:00:18.220 --> 00:00:20.220
Particle Swarm Optimization

00:00:20.660 --> 00:00:21.320
(PSO)

00:00:21.540 --> 00:00:23.680
and Its Applications.

00:00:24.760 --> 00:00:28.700
This is the outline of today’s course including Introduction,

00:00:29.460 --> 00:00:31.420
Particle Swarm Optimization,

00:00:31.740 --> 00:00:32.880
Applications.

00:00:33.280 --> 00:00:36.500
Now I am going to start with the Introduction.

00:00:37.120 --> 00:00:45.680
PSO was proposed by J. Kennedy an R. Eberhart in 1995.

00:00:46.300 --> 00:00:50.040
It simulates birds searching for food

00:00:50.580 --> 00:00:55.060
or the movement of fishes’ shoal.

00:00:56.160 --> 00:01:04.500
Particles in the swarm move around the search space looking for the optimum solution

00:01:05.260 --> 00:01:10.460
and adjust their position according to inertia,

00:01:10.880 --> 00:01:14.920
individual experience and  social experience.

00:01:15.420 --> 00:01:18.780
Now I am going to introduce the PSO algorithm.

00:01:19.360 --> 00:01:25.120
At first we have to initialize the particles from the solution space.

00:01:25.620 --> 00:01:30.260
A particle should be with position and velocity.

00:01:30.820 --> 00:01:31.880
In step 2,

00:01:32.440 --> 00:01:36.480
we have to Evaluate the fitness of each particle

00:01:36.780 --> 00:01:38.980
according to the fitness function.

00:01:39.600 --> 00:01:43.780
Then we can update the individual best solution

00:01:44.120 --> 00:01:49.360
PBest and the global best solution GBest.

00:01:50.020 --> 00:01:59.600
After that we update the velocity and position of each particle using these two equations.

00:02:00.420 --> 00:02:05.380
We can see here the velocity is updated according to the

00:02:05.640 --> 00:02:06.400
inertial,

00:02:06.740 --> 00:02:07.520
cognition

00:02:07.860 --> 00:02:10.200
and social experience

00:02:10.980 --> 00:02:15.100
Where omega, c1 and c2 are constants

00:02:15.960 --> 00:02:19.800
Random1 and random2 are random variables.

00:02:20.860 --> 00:02:24.800
After updating the velocity and position,

00:02:25.480 --> 00:02:32.600
go to step 2, and repeat until termination condition is reached.

00:02:33.280 --> 00:02:37.480
We can see an example for PSO solution update.

00:02:38.220 --> 00:02:42.260
Assume that the Current solution is (2, 2),

00:02:42.720 --> 00:02:48.300
the Particle’s best solution PBest is (2, 8),

00:02:49.100 --> 00:02:53.680
the Global best solution GBest is (7, 2),

00:02:54.240 --> 00:02:57.480
the Inertia: v(k) is (1, 2),

00:02:58.280 --> 00:03:03.040
omega equals to c1 equals to c2 equals to 1,

00:03:03.920 --> 00:03:09.560
random1 is 0.5, and random 2 is 0.4.

00:03:10.200 --> 00:03:16.720
Then we can obtain cognitive experience being (0,6)

00:03:17.100 --> 00:03:21.560
and social experience being (5, 0).

00:03:21.920 --> 00:03:22.420
Therefore,

00:03:22.940 --> 00:03:28.240
we have new velocity being (2, 5)

00:03:28.480 --> 00:03:31.900
and new position being (4, 7).

00:03:32.680 --> 00:03:37.300
Now we make a comparison between PSO and GA.

00:03:38.200 --> 00:03:43.840
We can see in this table that GA is Easier than PSO

00:03:44.280 --> 00:03:48.400
to find the global optimum due to “mutation”.

00:03:49.080 --> 00:03:57.040
However, the computation of GA is relatively more complicated than PSO.

00:03:57.920 --> 00:04:01.580
Finally we can see some applications of PSO.

00:04:02.140 --> 00:04:07.680
PSO can be applied for various optimization problems

00:04:07.800 --> 00:04:08.640
for example,

00:04:09.180 --> 00:04:11.940
Energy-Storage Optimization.

00:04:12.640 --> 00:04:19.460
Moreover, since PSO can simulate the movement of a particle swarm,

00:04:19.960 --> 00:04:26.840
this can also be applied to movie film as shown in this figure.