7、深入探索R语言中的深度学习：从DNN算法到参数初始化

深入探索R语言中的深度学习：从DNN算法到参数初始化

1. 编写深度神经网络（DNN）算法

在处理猫和狗的图像分类问题时，我们可以编写如下代码来计算概率和进行预测：

Probs <- list(round(score * 100, 2))
return (Probs)
}
Prob <- compute_Proba(two_layer_model$parameters,
testx,
hidden_layer_act = c('relu', 'relu'),
output_layer_act = 'sigmoid')
labels = ifelse(testy == 1, "dog", "cat")
predicted <- ifelse(
predict_model(two_layer_model$parameters,
testx,
hidden_layer_act = c('relu', 'relu'),
output_layer_act = 'sigmoid') == 0, 'cat', 'dog')
error <- ifelse(predicted == labels, 'No', 'Yes')
index <- c(1:length(labels))
Probs <- as.vector(unlist(Prob[index]))
par(mfrow = c(5, 10), mar = rep(0, 4))
for(i in 1:length(index)){
image(t(apply(matrix(as.matrix(testx[, index[i]]),
c(64, 64, 3),
byrow