Skip to content

INSTALL_NEURAL_NETWORK.md

Latest commit

 

History

History
128 lines (96 loc) · 3.26 KB

INSTALL_NEURAL_NETWORK.md

File metadata and controls

128 lines (96 loc) · 3.26 KB

Installing Keras3 for Neural Network Models

The neural network models (model_07 and model_08) require the keras3 package. Follow these steps to install it:

Step 1: Fix CRAN Mirror Issue

If you're getting "unable to access index for repository" errors, first fix your CRAN mirror:

# Set a working CRAN mirror
options(repos = c(CRAN = "https://cloud.r-project.org"))

# Or choose a mirror interactively
chooseCRANmirror()

Step 2: Install Required Dependencies

# Install devtools or remotes (try both if one fails)
install.packages("devtools")
# OR
install.packages("remotes")

Step 3: Install keras3 Package

Option A: Install from GitHub (Recommended for R 4.4)

# Using devtools
devtools::install_github("rstudio/keras3")

# OR using remotes
remotes::install_github("rstudio/keras3")

Option B: Install from CRAN (if available for your R version)

install.packages("keras3")

Note: If you're using R 4.4, keras3 might not be available on CRAN yet. Use Option A (GitHub) instead.

Quick Installation Script

For easiest installation, run the provided script:

Rscript install_keras3.R

Or in R:

source("install_keras3.R")

Step 2: Install TensorFlow Backend

After installing keras3, TensorFlow will be automatically installed when you first run:

library(keras3)
use_backend("tensorflow")

This happens automatically when you run the neural network model scripts.

Step 3: Verify Installation

You can verify the installation by running:

library(keras3)
use_backend("tensorflow")
cat("Keras3 and TensorFlow are ready!\n")

Troubleshooting

CRAN Mirror Issues:

If you see "unable to access index for repository" errors:

# Set a working CRAN mirror
options(repos = c(CRAN = "https://cloud.r-project.org"))
# Then try installing again

Package Not Available for R Version:

If you see "package 'keras3' is not available for this version of R":

  • keras3 might require R 4.5+ or need to be installed from GitHub
  • Use the GitHub installation method: devtools::install_github("rstudio/keras3")

If devtools/remotes installation fails:

  1. Try installing from a different CRAN mirror:

    options(repos = c(CRAN = "https://cloud.r-project.org"))
install.packages("devtools")

  2. Or download and install manually from:

If TensorFlow installation fails:

  1. Make sure you have Python installed on your system
  2. Install reticulate package: install.packages("reticulate")
  3. Try installing TensorFlow manually: 
    library(reticulate)
py_install("tensorflow")

If you get "no package called 'keras3'" error:

  • Make sure you've installed keras3 from GitHub successfully
  • Check if the package is loaded: library(keras3)
  • Try restarting R session and installing again
  • Verify installation: packageVersion("keras3")

Alternative: Skip Neural Network Models

If you don't need to run the neural network models, you can skip them and run only the other models:

  • model_01_glm_lasso_min.R
  • model_02_glm_lasso_1se.R
  • model_03_random_forest.R
  • model_04_ranger.R
  • model_05_xgboost.R
  • model_06_simple_glm.R

These models don't require keras3.