alstat

f, a = subplots(10, 20)
for i in arange(10):
    for j in arange(20):
        a[i, j].imshow(x_train[j + 20 * i])
        a[i, j].axis("off")
        a[i, j].set_adjustable('box-forced')

f.savefig("img1.png", bbox_inches='tight', pad_inches = 0)

alstat

items = [x_train, y_train, x_test, y_test]

# training set
[type(item) for item in items[:2]]    # [<type 'numpy.ndarray'>, <type 'numpy.ndarray'>]
[item.shape for item in items[:2]]    # [(50000, 32, 32, 3), (50000, 1)]

# testing set
[type(item) for item in items[2:]]    # [<type 'numpy.ndarray'>, <type 'numpy.ndarray'>]
[item.shape for item in items[2:]]    # [(10000, 32, 32, 3), (10000, 1)]

alstat

items <- list(x_train, y_train, x_test, y_test)

dim_formatter <- function (x) {
    if (length(dim(x)) > 2)
        paste("(", dim(x)[1], ", ", dim(x)[2], " , ", dim(x)[3], " , ", dim(x)[4], ")", sep = "")
    else
        paste("(", dim(x)[1], ", ", dim(x)[2], ")", sep = "")
}

# training set

alstat

cifar100 <- dataset_cifar100(label_mode = "fine")

x_train <- cifar100$train$x; y_train <- cifar100$train$y
x_test <- cifar100$test$x; y_test <- cifar100$test$y

alstat

(x_train, y_train), (x_test, y_test) = cifar100.load_data(label_mode = "fine")

alstat

library(keras)

# Define the constants
CONST_N <- 2000
CONST_EPOCHS <- 30
CONST_PIXEL_MAX <- 255

alstat

install.packages("keras")

alstat

using Images
using MLDatasets
using Flux: ADAM,
            argmax, 
            Chain, 
            crossentropy, 
            Dense, 
            params,
            relu, 
            softmax, 

alstat

using Flux: ADAM,
            argmax, 
            Chain,
            Dense, 
            params,
            relu, 
            softmax

m = Chain(
    Dense(32^2 * 3, 32 * 10, relu),

alstat

using Flux: onehotbatch

x_vec = [float.(reshape(x_train[:, :, :, i], :)) for i in 1:1000];
X = hcat(x_vec...);            
Y = onehotbatch(y_train_fine[1:1000], 0:99);