Edition Deep Learning with R, 2nd Edition coincides with new versions of TensorFlow and Keras. These releases bring many improvements that allow for more idiomatic and concise R code.
First, the set of Tensor methods for R base generics has expanded considerably. The set of R generics that work with TensorFlow Tensors is now quite extensive:
methods(class = "tensorflow.tensor") (1) - ! != ( (<-
(6) * / & %/% %%
(11) ^ + < <= ==
(16) > >= | abs acos
(21) all any aperm Arg asin
(26) atan cbind ceiling Conj cos
(31) cospi digamma dim exp expm1
(36) floor Im is.finite is.infinite is.nan
(41) length lgamma log log10 log1p
(46) log2 max mean min Mod
(51) print prod range rbind Re
(56) rep round sign sin sinpi
(61) sort sqrt str sum t
(66) tan tanpi This means that you can often write the same code for TensorFlow Tensors as you would for R arrays. Take, for example, this little function from Chapter 11 of the book:
reweight_distribution <-
function(original_distribution, temperature = 0.5) {
original_distribution %>%
{ exp(log(.) / temperature) } %>%
{ . / sum(.) }
}Note that functions like reweight_distribution() work with both 1D R vectors and 1D TensorFlow tensors, from exp(), log(), /and
sum() are all R generics with methods for TensorFlow Tensors.
In the same vein, this release of Keras brings improvements to the way custom class extensions are defined for Keras. Partly inspired by the novelty R7 syntax, there is a new family of functions: new_layer_class(), new_model_class(),
new_metric_class(), and so on. This new interface greatly simplifies the amount of boilerplate code needed to define custom Keras extensions—a nice R interface that serves as a facade over the mechanics of subclassing Python classes. This new interface is the yang to the yin %py_class%– a way to mimic the Python class definition syntax in R. Of course, the “raw” conversion API
R6Class() to Python via r_to_py() is still available for users who require full control.
This release also brings with it an abundance of small improvements throughout the Keras R interface: updated print() and plot() methods for models, improvements freeze_weights() and load_model_tf()new exported utilities like zip_lists() and %<>%. And let’s not forget to mention a new family of R functions for adjusting the learning rate during training, with a set of built-in plans like
learning_rate_schedule_cosine_decay()supplemented with an interface for creating custom schedules with new_learning_rate_schedule_class().
Full release notes for R packages can be found here:
However, the release notes for the R packages only tell half the story. The R to Keras and TensorFlow interfaces work by injecting the entire Python process into R (via the
reticulate package). One of the main advantages of this design is that R users have full access to everything in both R and Python. In other words, the R interface always has feature parity with the Python interface—anything you can do with TensorFlow in Python, you can do in R just as easily. This means that the release notes for the Python release of TensorFlow are equally relevant for R users:
Thank you for reading!
Photo by Raphael Wilde on Unsplash
Use again
Text and images are licensed under Creative Commons Attribution CC BY 4.0. Figures that have been reused from other sources are not covered by this license and can be identified by the note in their caption: “Image from …”.
Quote
For attribution, please cite this work as
Kalinowski (2022, June 9). Posit AI Blog: TensorFlow and Keras 2.9. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2022-06-09-tf-2-9/
BibTeX citation
@misc{kalinowskitf29,
author = {Kalinowski, Tomasz},
title = {Posit AI Blog: TensorFlow and Keras 2.9},
url = {https://blogs.rstudio.com/tensorflow/posts/2022-06-09-tf-2-9/},
year = {2022}
}