MXNet_Classification_Multi_Layer_Perceptron.R

# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
# Develop a Neural Network with MXNet in Five Minutes
# R Notebook Author: @Korkrid Akepanidtaworn
# Tutorial from MXNet Community
# Source: https://mxnet.incubator.apache.org/tutorials/r/fiveMinutesNeuralNetwork.html
# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
library(mlbench)
library(tidyverse)
library(caret)
library(e1071)

# To install MXNet, you need to run these 4 lines of codes
cran <- getOption("repos")
cran["dmlc"] <- "https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/R/CRAN/"
options(repos = cran)
# install.packages("mxnet")
library(mxnet)

# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
# Load Data
# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------

# This is the data set used by Gorman and Sejnowski in their study of the classification of sonar signals using a neural network [1]. The task is to train a network to discriminate between sonar signals bounced off a metal cylinder and those bounced off a roughly cylindrical rock. Eachpatternisasetof60numbersintherange 0.0to 1.0. Eachnumberrepresentstheenergywithin a particular frequency band, integrated over a certain period of time. The integration aperture for higherfrequenciesoccurlaterintime,sincethesefrequenciesaretransmittedlaterduringthechirp. The label associated with each record contains the letter "R" if the object is a rock and "M" if it is a mine (metal cylinder). The numbers in the labels are in increasing order of aspect angle, but they do not encode the angle directly.

data(Sonar, package="mlbench")
head(Sonar)
# V1     V2     V3     V4     V5     V6     V7     V8     V9    V10    V11    V12    V13    V14    V15    V16    V17
# 1 0.0200 0.0371 0.0428 0.0207 0.0954 0.0986 0.1539 0.1601 0.3109 0.2111 0.1609 0.1582 0.2238 0.0645 0.0660 0.2273 0.3100
# 2 0.0453 0.0523 0.0843 0.0689 0.1183 0.2583 0.2156 0.3481 0.3337 0.2872 0.4918 0.6552 0.6919 0.7797 0.7464 0.9444 1.0000
# 3 0.0262 0.0582 0.1099 0.1083 0.0974 0.2280 0.2431 0.3771 0.5598 0.6194 0.6333 0.7060 0.5544 0.5320 0.6479 0.6931 0.6759
# 4 0.0100 0.0171 0.0623 0.0205 0.0205 0.0368 0.1098 0.1276 0.0598 0.1264 0.0881 0.1992 0.0184 0.2261 0.1729 0.2131 0.0693
# 5 0.0762 0.0666 0.0481 0.0394 0.0590 0.0649 0.1209 0.2467 0.3564 0.4459 0.4152 0.3952 0.4256 0.4135 0.4528 0.5326 0.7306
# 6 0.0286 0.0453 0.0277 0.0174 0.0384 0.0990 0.1201 0.1833 0.2105 0.3039 0.2988 0.4250 0.6343 0.8198 1.0000 0.9988 0.9508
# V18    V19    V20    V21    V22    V23    V24    V25    V26    V27    V28    V29    V30    V31    V32    V33    V34
# 1 0.2999 0.5078 0.4797 0.5783 0.5071 0.4328 0.5550 0.6711 0.6415 0.7104 0.8080 0.6791 0.3857 0.1307 0.2604 0.5121 0.7547
# 2 0.8874 0.8024 0.7818 0.5212 0.4052 0.3957 0.3914 0.3250 0.3200 0.3271 0.2767 0.4423 0.2028 0.3788 0.2947 0.1984 0.2341
# 3 0.7551 0.8929 0.8619 0.7974 0.6737 0.4293 0.3648 0.5331 0.2413 0.5070 0.8533 0.6036 0.8514 0.8512 0.5045 0.1862 0.2709
# 4 0.2281 0.4060 0.3973 0.2741 0.3690 0.5556 0.4846 0.3140 0.5334 0.5256 0.2520 0.2090 0.3559 0.6260 0.7340 0.6120 0.3497
# 5 0.6193 0.2032 0.4636 0.4148 0.4292 0.5730 0.5399 0.3161 0.2285 0.6995 1.0000 0.7262 0.4724 0.5103 0.5459 0.2881 0.0981
# 6 0.9025 0.7234 0.5122 0.2074 0.3985 0.5890 0.2872 0.2043 0.5782 0.5389 0.3750 0.3411 0.5067 0.5580 0.4778 0.3299 0.2198
# V35    V36    V37    V38    V39    V40    V41    V42    V43    V44    V45    V46    V47    V48    V49    V50    V51
# 1 0.8537 0.8507 0.6692 0.6097 0.4943 0.2744 0.0510 0.2834 0.2825 0.4256 0.2641 0.1386 0.1051 0.1343 0.0383 0.0324 0.0232
# 2 0.1306 0.4182 0.3835 0.1057 0.1840 0.1970 0.1674 0.0583 0.1401 0.1628 0.0621 0.0203 0.0530 0.0742 0.0409 0.0061 0.0125
# 3 0.4232 0.3043 0.6116 0.6756 0.5375 0.4719 0.4647 0.2587 0.2129 0.2222 0.2111 0.0176 0.1348 0.0744 0.0130 0.0106 0.0033
# 4 0.3953 0.3012 0.5408 0.8814 0.9857 0.9167 0.6121 0.5006 0.3210 0.3202 0.4295 0.3654 0.2655 0.1576 0.0681 0.0294 0.0241
# 5 0.1951 0.4181 0.4604 0.3217 0.2828 0.2430 0.1979 0.2444 0.1847 0.0841 0.0692 0.0528 0.0357 0.0085 0.0230 0.0046 0.0156
# 6 0.1407 0.2856 0.3807 0.4158 0.4054 0.3296 0.2707 0.2650 0.0723 0.1238 0.1192 0.1089 0.0623 0.0494 0.0264 0.0081 0.0104
# V52    V53    V54    V55    V56    V57    V58    V59    V60 Class
# 1 0.0027 0.0065 0.0159 0.0072 0.0167 0.0180 0.0084 0.0090 0.0032     1
# 2 0.0084 0.0089 0.0048 0.0094 0.0191 0.0140 0.0049 0.0052 0.0044     1
# 3 0.0232 0.0166 0.0095 0.0180 0.0244 0.0316 0.0164 0.0095 0.0078     1
# 4 0.0121 0.0036 0.0150 0.0085 0.0073 0.0050 0.0044 0.0040 0.0117     1
# 5 0.0031 0.0054 0.0105 0.0110 0.0015 0.0072 0.0048 0.0107 0.0094     1
# 6 0.0045 0.0014 0.0038 0.0013 0.0089 0.0057 0.0027 0.0051 0.0062     1

Sonar$Class = ifelse(Sonar$Class == 'R', 0, 1)

# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
# Split Training/Testing 80:20
# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------

# 80% of the sample size
smp_size <- floor(0.60 * nrow(Sonar))

# Set the seed to make your partition reproducible
set.seed(1234)
train_ind <- sample(seq_len(nrow(Sonar)), size = smp_size)

train <- Sonar[train_ind, ]
test <- Sonar[-train_ind, ]

train_x = as.matrix(train[, 1:60])
train_y = as.numeric(as.matrix(train[, 61]))
test_x = as.matrix(test[, 1:60])
test_y = as.numeric(as.matrix(test[, 61]))

# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
# Train the Neural Network Model
# Description: use a multi-layer perceptron as our classifier. 
# In mxnet, we have a function called mx.mlp for building a general multi-layer neural network 
# to do classification or regression.
# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
mx.set.seed(1234)
model <- mx.mlp(train_x, 
                train_y, 
                hidden_node=10, 
                out_node=2, 
                out_activation="softmax",
                num.round = 20, 
                array.batch.size = 15, 
                learning.rate = 0.07, 
                momentum = 0.9,
                eval.metric = mx.metric.accuracy)
# Start training with 1 devices
# [1] Train-accuracy=0.400000009271834
# [2] Train-accuracy=0.459259268310335
# [3] Train-accuracy=0.43703705072403
# [4] Train-accuracy=0.451851861344443
# [5] Train-accuracy=0.429629640446769
# [6] Train-accuracy=0.474074088864856
# [7] Train-accuracy=0.614814837773641
# [8] Train-accuracy=0.622222244739532
# [9] Train-accuracy=0.60000001721912
# [10] Train-accuracy=0.629629648394055
# [11] Train-accuracy=0.651851872603099
# [12] Train-accuracy=0.703703721364339
# [13] Train-accuracy=0.711111128330231
# [14] Train-accuracy=0.688888907432556
# [15] Train-accuracy=0.674074093500773
# [16] Train-accuracy=0.762962977091471
# [17] Train-accuracy=0.785185197989146
# [18] Train-accuracy=0.800000011920929
# [19] Train-accuracy=0.770370384057363
# [20] Train-accuracy=0.75555557012558

summary(model)
# Length Class         Mode
# symbol     1      Rcpp_MXSymbol S4  
# arg.params 4      -none-        list
# aux.params 0      -none-        list

graph.viz(model$symbol)

# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
# Predict on the Test Data
# -------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------
test_pred = predict(model, test_x) # Array of Predicted Probabilities
test_pred = t(test_pred) # Predicted probabilities (for Prediction Wrangling)
test_pred = as.data.frame(test_pred)
test_pred
# [,1]         [,2]
# [1,] 0.426815450 0.5731846094
# [2,] 0.959362805 0.0406371839
# [3,] 0.773524523 0.2264754623
# [4,] 0.133803442 0.8661965728
# [5,] 0.784839451 0.2151605636
# [6,] 0.988470435 0.0115295900
# [7,] 0.992558479 0.0074415053
# [8,] 0.998107433 0.0018925916
# [9,] 0.618567407 0.3814325333
# [10,] 0.979754388 0.0202456508
# [11,] 0.997008383 0.0029916782
# [12,] 0.992276728 0.0077233198
# [13,] 0.999820054 0.0001799719

# Assume that [,1] is probability for 0
# Assume that [,2] is probability for 1
test_pred = test_pred %>%
  dplyr::rename(prob_0 = V1,
                prob_1 = V2) %>%
  dplyr::mutate(Class_Pred = ifelse(prob_1 > prob_0, 1, 0))

Class = test$Class

test_eval = cbind(Class, test_pred)

confusionMatrix(as.factor(test_eval$Class_Pred),
                as.factor(test_eval$Class),
                positive = '1')
# Confusion Matrix and Statistics
# 
# Reference
# Prediction  0  1
# 0 32 20
# 1  2 30
# 
# Accuracy : 0.7381         
# 95% CI : (0.6307, 0.828)
# No Information Rate : 0.5952         
# P-Value [Acc > NIR] : 0.0044768      
# 
# Kappa : 0.4989         
# Mcnemar's Test P-Value : 0.0002896      
#                                          
#             Sensitivity : 0.6000         
#             Specificity : 0.9412         
#          Pos Pred Value : 0.9375         
#          Neg Pred Value : 0.6154         
#              Prevalence : 0.5952         
#          Detection Rate : 0.3571         
#    Detection Prevalence : 0.3810         
#       Balanced Accuracy : 0.7706         
#                                          
#        'Positive' Class : 1