test/deepnet.R

In kkdey/classtpx: MAP Estimation using topic models in a semi-supervised set up where there are few samples with known topic proportions or known to belong to some cluster exclusively and there are samples for which no such information are available.

####  Deep Learning with the mxnet package ################################

data(Sonar, package="mlbench")
Sonar[,61] = as.numeric(Sonar[,61])-1
train.ind = c(1:50, 100:150)
train.x = data.matrix(Sonar[train.ind, 1:60])
train.y = Sonar[train.ind, 61]
test.x = data.matrix(Sonar[-train.ind, 1:60])
test.y = Sonar[-train.ind, 61]


mx.set.seed(0)
model <- mx.mlp(train.x, train.y, hidden_node=50, out_node=2, out_activation="softmax",
                num.round=50, array.batch.size=15, learning.rate=0.1, momentum=0.8, 
                eval.metric=mx.metric.accuracy)

graph.viz(model$symbol$as.json())
preds = predict(model, test.x)

pred.label = max.col(t(preds))-1
table(pred.label, test.y)

training.data.frame <- data.frame(cbind(train.y, train.x));
model <- svm(as.factor(train.y) ~ ., data = training.data.frame)

test_class_svm <- predict(model, test.x)
tab_class_svm <- table(test_class_svm, test.y)

data <- mx.symbol.Variable("data")
fc1 <- mx.symbol.FullyConnected(data, name="fc1", num_hidden=128)
act1 <- mx.symbol.Activation(fc1, name="relu1", act_type="relu")
fc2 <- mx.symbol.FullyConnected(act1, name="fc2", num_hidden=64)
act2 <- mx.symbol.Activation(fc2, name="relu2", act_type="relu")
fc3 <- mx.symbol.FullyConnected(act2, name="fc3", num_hidden=10)
softmax <- mx.symbol.SoftmaxOutput(fc3, name="sm")


data <- mx.symbol.Variable("data")
fc1 <- mx.symbol.FullyConnected(data, name="fc1", num_hidden=128)
act1 <- mx.symbol.Activation(fc1, name="tanh1", act_type="tanh")
fc2 <- mx.symbol.FullyConnected(act1, name="fc2", num_hidden=64)
act2 <- mx.symbol.Activation(fc2, name="tanh2", act_type="tanh")
fc3 <- mx.symbol.FullyConnected(act2, name="fc3", num_hidden=32)
act3 <- mx.symbol.Activation(fc3, name="tanh3", act_type="tanh")
fc4 <- mx.symbol.FullyConnected(act3, name="fc4", num_hidden=16)
act4 <- mx.symbol.Activation(fc4, name="tanh4", act_type="tanh")
fc5 <- mx.symbol.FullyConnected(act4, name="fc5", num_hidden=8)
act5 <- mx.symbol.Activation(fc5, name="tanh5", act_type="tanh")
fc6 <- mx.symbol.FullyConnected(act5, name="fc6", num_hidden=2)
softmax <- mx.symbol.SoftmaxOutput(fc6, name="sm")

devices <- mx.cpu()

mx.set.seed(0)
model <- mx.model.FeedForward.create(softmax, X=train.x, y=train.y,
                                     ctx=devices, num.round=10, array.batch.size=10,
                                     learning.rate=0.07, momentum=0.9,  eval.metric=mx.metric.accuracy,
                                     initializer=mx.init.uniform(0.07),
                                     epoch.end.callback=mx.callback.log.train.metric(100))

graph.viz(model$symbol$as.json())




library(mxnet)
data(BostonHousing, package="mlbench")
train.ind = seq(1, 506, 3)
train.x = data.matrix(BostonHousing[train.ind, -14])
train.y = BostonHousing[train.ind, 14]
test.x = data.matrix(BostonHousing[-train.ind, -14])
test.y = BostonHousing[-train.ind, 14]

data <- mx.symbol.Variable("data")

fc1 <- mx.symbol.FullyConnected(data, num_hidden=2)

lro <- mx.symbol.LinearRegressionOutput(fc1)
mx.set.seed(0)
model <- mx.model.FeedForward.create(lro, X=train.x, y=as.array(train.y),
                                     ctx=mx.cpu(), num.round=50, array.batch.size=20,
                                     learning.rate=2e-6, momentum=0.9, eval.metric=mx.metric.rmse)



########  Deep Learning with the h2o package ############################

library(h2o)
localH2O <- h2o.init(ip = "localhost", port = 54321, startH2O = TRUE)
data(Sonar, package="mlbench")
train.ind = c(1:50, 100:150)
train.x = data.matrix(Sonar[train.ind, 1:60])
train.y = Sonar[train.ind, 61]

training_dat <- data.frame(cbind(train.y, train.x));


h2o.init()
prosPath <- system.file("extdata", "prostate.csv", package="h2o")
prostate.hex <- h2o.uploadFile(path = prosPath)
as.data.frame(prostate.hex)

h2o.dat <- as.h2o(training_dat, destination_frame = "dat")

model <- 
  h2o.deeplearning(x = 2:60,  # column numbers for predictors
                   y = 1,   # column number for label
                   training_frame = as.h2o(training_dat, destination_frame = "dat"),
                  # data = dat, # data in H2O format
                   activation = "TanhWithDropout", # or 'Tanh'
                   input_dropout_ratio = 0.2, # % of inputs dropout
                   hidden_dropout_ratios = c(0.5,0.5,0.5), # % for nodes dropout
                   balance_classes = TRUE, 
                   hidden = c(50,50,50), # three layers of 50 nodes
                   epochs = 100) # max. no. of epochs