Artificial Hydrocarbon Networks in R

Quick Start

Here are some quick examples to get you started.

Create a network for a SISO sytem

The following code let’s you create the data to train an artificial hydrocarbon network.

library(ahnr)

# Create data
set.seed(12321)
x <- 2 * runif(1000) - 1;
x <- sort(x)

y <- (x < 0.1) * (0.05 * runif(1000) + atan(pi*x)) + 
    (x >= 0.1 & x < 0.6) * (0.05 * runif(1000) + sin(pi*x)) + 
    (x >= 0.6) * (0.05 * runif(1000) + cos(pi*x))

plot(x, y, type = 'l')

# Create the Sigma dataset
Sigma <- list(X = data.frame(x = x), Y = data.frame(y = y))

# Create network
ahn <- AHNnD(Sigma, 5, 0.01, 100)

The trained network can be used to visualize its performance.

# Create test data
X <- data.frame(x = x)

# Simulate
ysim <- SimAHNnD(ahn, X)

plot(x, y, type = 'l')
lines(x, ysim, type = 'l', col = 'red')
legend(-1, 1, c('Original', 'Simulation'), col = c(1,2), lty = c(1,3), cex = 0.8)

A summary of the network can be obtained with the summary command.

summary(ahn)

## 
## Artificial Hydrocarbon Network trained:
## 
## Number of molecules:
##  5 
## 
## Learning factor:
##  0.01 
## 
## Overall error:
##  0.2875 
## 
## Centers of the molecules:
##                     x
## molecule1  0.36691689
## molecule2  0.06369617
## molecule3 -0.96254244
## molecule4 -0.68667076
## molecule5 -0.94715030
## 
## Molecules:
## Molecule 1:
##           x
## C1   -4.092
## H11  32.303
## H12 -60.912
## H13  31.995
## 
## Molecule 2:
##         x
## C2  0.014
## H21 2.909
## H22 1.057
## 
## Molecule 3:
##          x
## C3  11.446
## H31 25.479
## H32 12.796
## 
## Molecule 4:
##          x
## C4  -0.228
## H41  2.049
## H42  1.101
## 
## Molecule 5:
##            x
## C5   -28.744
## H51  -94.902
## H52 -108.461
## H53  -41.124

Finally, the network itself can be plotted with the plot command. The text of the carbon of the first molecule is in red.

plot(ahn)

Create a network for a MISO sytem

# Create data
set.seed(12321)
t <- seq(0, 15, 0.01)
X <- data.frame(x1 = cos(t), x2 = t)
Y <- data.frame(y = sin(t))

# Create the Sigma dataset
Sigma <- list(X = X, Y = Y)

# Create network
ahn <- AHNnD(Sigma, 5, 0.01, 100)

# Simulate
ysim <- SimAHNnD(ahn, X)

plot(t, Y$y, type = 'l', col = 'black', xlab = 't', ylab = 'output')
lines(t, ysim, col = 'red')
legend(0, -0.5, c('Original', 'Simulation'), col = c(1,2), lty = c(1,3), cex = 0.6)

summary(ahn)

## 
## Artificial Hydrocarbon Network trained:
## 
## Number of molecules:
##  5 
## 
## Learning factor:
##  0.01 
## 
## Overall error:
##  0.2475 
## 
## Centers of the molecules:
##                   x1         x2
## molecule1  0.6871690  6.4083024
## molecule2  0.9325409  0.9380700
## molecule3  0.2068092  6.3405795
## molecule4 -0.8726707  3.4672676
## molecule5  0.8153233 14.7093439
## 
## Molecules:
## Molecule 1:
##         x1     x2
## C1  -1.619 -1.619
## H11 -3.170 -5.443
## H12  0.939  2.007
## H13 -0.226 -0.159
## 
## Molecule 2:
##         x1     x2
## C2  -0.013 -0.013
## H21  0.272  1.039
## H22 -0.248 -0.246
## 
## Molecule 3:
##         x1     x2
## C3  -6.842 -6.842
## H31 -0.364  3.826
## H32 -0.631 -0.249
## 
## Molecule 4:
##         x1      x2
## C4  11.201  11.201
## H41 -6.861 -17.287
## H42 -0.760   2.616
## 
## Molecule 5:
##         x1     x2
## C5  -0.077 -0.077
## H51  0.433 -0.352
## H52 -1.070  0.005
## H53  0.807  0.002

plot(ahn)