corrected small error and added explanation

[D-Maar]

mean_per_row = tf$reduce_mean(x, axis = 0L)
does not correspond to
mean_per_row = apply(x, 1, mean)
but
mean_per_row = tf$reduce_mean(x, axis = 1L)
does

Also for tf matrices
result = x - mean_per_row
does not correspond to the same for R matrices, but
tf$transpose(tf$transpose(x) - mean_per_row)
does

Code:

library(tensorflow)
x=matrix(1:100, nrow =10L)
correct_mean_per_row =  apply(x, 1, mean)
wrong_TF = tf$reduce_mean(x, axis = 0L)
correct_TF = tf$reduce_mean(x, axis = 1L)
correct_mean_per_row
wrong_TF
correct_TF

do_something_R = function(x = matrix(0.0, 10L, 10L)){
  mean_per_row = apply(x, 1, mean)
  result = x - mean_per_row
  return(result)
}

do_something_TF_new = function(x = matrix(0.0, 10L, 10L)){
  x = tf$constant(x)  # Remember, this is a local copy!
  mean_per_row = tf$reduce_mean(x, axis = 1L)
  result = tf$transpose(x) - mean_per_row 
  return(tf$transpose(result)) 
}

do_something_TF_old = function(x = matrix(0.0, 10L, 10L)){
  x = tf$constant(x)  # Remember, this is a local copy!
  mean_per_row = tf$reduce_mean(x, axis = 0L)
  result = x - mean_per_row 
  return(result) 
}

x=matrix(1:100, nrow =5L)
do_something_R(x)
do_something_TF_new(x)
do_something_TF_old(x)