R_test_of_nn.R

In mmider/OxWaSPneuralnets: Neural networks

#
#
# gen_data = function(j, n=50){
#   r = seq(0, 1, length.out = n)
#   theta = seq(j*4, (j+1)*4, length.out = n) + rnorm(n, sd=0.2)
#   x = r * cos(theta)
#   y = r * sin(theta)
#   return(data.frame(x1 = x, x2 = y, labels = j))
# }
#
# df = rbind(gen_data(0), gen_data(1), gen_data(2))
# df = df[sample(1:nrow(df)), ]
#
# X = cbind(df$x1, df$x2)
# y = df$labels
#
# plot(X, col=y+1, pch=16)
#
#
# ### fit neural network
#
# # First, do this on command line
# # R CMD SHLIB cs231n.c -lm -lgsl -lgslcblas
#
#
#
# layer_sizes = c(ncol(X),20,3)
# num_layers = 3
# num_iterations = 2000
# batch_size = nrow(X) /10
# step_size = 0.1
# penalty = 0.001
#
#
#
# out = .C("nn",
#          X = as.vector(as.numeric(X)),y = as.double(y),
#          layer_sizes = as.integer(layer_sizes),
#          num_layers = as.integer(num_layers),
#          num_iterations = as.integer(num_iterations),
#          batch_size = as.integer(batch_size),
#          step_size = as.double(step_size),
#          nrow = as.integer(nrow(X)),
#          ncol = as.integer(ncol(X)),
#          penalty = as.double(penalty),
#          output = as.double(-1), output2 = as.double(rep(0,num_iterations)),
#          trans_type = as.integer(2))
#
# out$output
# plot(out$output2, type ="l")

## mnist data
library("OxWaSPneuralnets", lib.loc = "~/R")
load("data/mnist.RData")

X = train$x[1:1000,]
y = train$y[1:1000]


X = train$x[1:1000,]
y = train$y[1:1000]
X_test = test$x[1:500,]
y_test = test$y[1:500]



layer_sizes = c(ncol(X),20,10)
num_layers = 3
num_iterations = 1000
core_num = 3
#batch_size = nrow(X)/8
step_size = 0.0001
penalty = 0.001



out = .C("nn",
         X = as.vector(as.numeric(X)),y = as.double(y),
         X_test = as.vector(as.numeric(X_test)), y_test = as.double(y_test),
         layer_sizes = as.integer(layer_sizes),
         num_layers = as.integer(num_layers),
         num_iterations = as.integer(num_iterations),
         core_num = as.integer(core_num),
         step_size = as.double(step_size),
         nrow = as.integer(nrow(X)),
         ncol = as.integer(ncol(X)),
         nrow_test = as.integer(nrow(X_test)),
         penalty = as.double(penalty),
         output = as.double(-1),
         output2 = as.double(rep(0,num_iterations)),
         output3 = as.double(-1),
         output4 = as.double(rep(0,nrow(X)*10)),
         output5 = as.double(rep(0,nrow(X_test)*10)),
         output6 = as.integer(rep(0, nrow(X))),
         output7 = as.integer(rep(0,nrow(X_test))),
         trans_type = as.integer(1))

out$output
plot(out$output2,type = "l")
out$output3
out$output6