R Deep Learning Essentials - Second Edition

Linear regression using TensorFlow

In this first Tensorflow example, we will look at regression. The code for this section is in the Chapter8/regression_tf.R folder:

1. First, we create some fake data for an an input value, x, and an output value, y. We set y to be approximately equal to 0.8 + x * 1.3. We want the application to discover the beta0 and beta1 values, which are 0.8 and 1.3, respectively:

library(tensorflow)set.seed(42)# create 50000 x variable between 0 and 100x_var <- runif(50000,min=0,max=1)#y = approx(1.3x + 0.8)y_var <- rnorm(50000,0.8,0.04) + x_var * rnorm(50000,1.3,0.05)# y_pred = beta0 + beta1 * xbeta0 <- tf$Variable(tf$zeros(shape(1L)))beta1 <- tf$Variable(tf$random_uniform(shape(1L), -1.0, 1.0))y_pred <- beta0 + beta1*x_var ...