Deep Learning with Keras :: Cheatsheet

Hex logo for tensorflow - A red hexagon with a stylized 'TFR' (denoting TensorFlow for R) in a lighter shade of red, with the 'F' joined to the 'R'.

Download PDF



Translations (PDF)

Intro

Keras is a high-level neural networks API developed with a focus on enabling fast experimentation. It supports multiple back-ends, including TensorFlow, CNTK and Theano.

TensorFlow is a lower level mathematical library for building deep neural network architectures. The keras R package makes it easy to use Keras and TensorFlow in R.

  1. Define: Model, Sequential model, Multi-GPU model
  2. Compile: Optimizer, Loss, Metrics
  3. Fit: Batch size, Epochs, Validation split
  4. Evaluate: Evaluate, Plot
  5. Predict: Classes, Probability

Read more at:
https://tensorflow.rstudio.com
https://www.manning.com/books/deep-learning-with-r-second-edition

Installation

The keras R package uses the Python keras library. You can install all the prerequisites directly from R: https://tensorflow.rstudio.com/install.

library(keras)
install_keras()

See ?install_keras for GPU instructions.

This installs the required libraries in an Anaconda environment or virtual environment r-tensorflow.

Training an Image Recognizer on MNIST Data

The “Hello, World!” of deep learning

# input layer: use MNIST images
mnist <- dataset_mnist()
x_train <- mnist$train$x
y_train <- mnist$train$y 
x_test <- mnist$test$x
y_test <- mnist$test$y

# reshape and rescale
x_train <- array_reshape(x_train, c(nrow(x_train), 784)) 
x_test <- array_reshape(x_test, c(nrow(x_test), 784)) 
x_train <- x_train / 255
x_test <- x_test / 255

y_train <- to_categorical(y_train, 10) 
y_test <- to_categorical(y_test, 10)

# defining the model and layers
model <- keras_model_sequential() 
model %>%
  layer_dense(units = 256, activation = 'relu', input_shape = c(784)) %>%
  layer_dropout(rate = 0.4) %>% 
  layer_dense(units = 128, activation = 'relu') %>% 
  layer_dense(units = 10, activation = 'softmax')
  
# compile (define loss and optimizer)
model %>%
  compile(
    loss = 'categorical_crossentropy', 
    optimizer = optimizer_rmsprop(), 
    metrics = c('accuracy')
)

# train (fit)
model %>% fit(
  x_train, y_train,
  epochs = 30, batch_size = 128, 
  validation_split = 0.2
)

model %>% evaluate(x_test, y_test) 
model %>% predict_classes(x_test)

Working with keras models

Define a Model

  • keras_model(): Keras Model.

  • keras_model_sequential(): Keras Model composed of a linear stack of layers.

  • multi_gpu_model(): Replicates a model on different GPUs.

Compile a Model

  • compile(object, optimizer, loss, metrics = NULL): Configure a Keras model for training.

Fit a Model

  • fit(object, x = NULL, y = NULL, batch_size = NULL, epochs = 10, verbose = 1, callbacks = NULL, ...): Train a Keras model for a fixed number of epochs (iterations).

  • fit_generator(): Fits the model on data yielded batch-by-batch by a generator.

  • train_on_batch(); test_on_batch(): Single gradient update or model evaluation over one batch of samples.

Evaluate a Model

  • evaluate(object, x = NULL, y = NULL, batch_size = NULL): Evaluate a Keras model.

  • evaluate_generator(): Evaluates the model on a data generator.

Predict

  • predict(): Generate predictions from a Keras model.

  • predict_proba(); predict_classes(): Generates probability or class probability predictions for the input samples.

  • predict_on_batch(): Returns predictions for a single batch of samples.

  • predict_generator(): Generates predictions for the input samples from a data generator.

Other Model Operations

  • summary(): Print a summary of a Keras model.

  • export_savedmodel(): Export a saved model.

  • get_layer(): Retrieves a layer based on either its name (unique) or index.

  • pop_layer(): Remove the last layer in a model.

  • save_model_hdf5(); load_model_hdf5(): Save/Load models using HDF5 files.

  • serialize_model(); unserialize_model(): Serialize a model to an R object.

  • clone_model(): Clone a model instance.

  • freeze_weights(); unfreeze_weights()

Core Layers

  • layer_input(): Input layer.

  • layer_dense(): Add a densely-connected NN layer to an output.

  • layer_activation(): Apply an activation function to an output.

  • layer_dropout(): Applies Dropout to the input.

  • layer_reshape(): Reshapes an output to a certain shape.

  • layer_permute(): Permute the dimensions of an input according to a given pattern.

  • layer_repeat_vector(): Repeats the input n times.

  • layer_lambda(object, f): Wraps arbitrary expression as a layer.

  • layer_activity_regularization(): Layer that applies an update to the cost function based input activity.

  • layer_masking(): Masks a sequence by using a mask value to skip timesteps.

  • layer_flatten(): Flattens an input.

More layers

Convolutional Layers

  • layer_conv_1d(): 1D, e.g. temporal convolution.

  • layer_conv_2d_transpose(): Transposed 2D (deconvolution).

  • layer_conv_2d() : 2D, e.g. spatial convolution over images.

  • layer_conv_3d_transpose(): Transposed 3D (deconvolution).

  • layer_conv_3d(): 3D, e.g. spatial convolution over volumes.

  • layer_conv_lstm_2d(): Convolutional LSTM.

  • layer_separable_conv_2d(): Depthwise separable 2D.

  • layer_upsampling_1d(); layer_upsampling_2d(); layer_upsampling_3d(): Upsampling layer.

  • layer_zero_padding_1d(); layer_zero_padding_2d(); layer_zero_padding_3d(): Zero-padding layer.

  • layer_cropping_1d(); layer_cropping_2d(); layer_cropping_3d(): Cropping layer.

Pooling Layers

  • layer_max_pooling_1d(); layer_max_pooling_2d(); layer_max_pooling_3d(): Maximum pooling for 1D to 3D.

  • layer_average_pooling_1d(); layer_average_pooling_2d(); layer_average_pooling_3d(): Average pooling for 1D to 3D.

  • layer_global_max_pooling_1d(); layer_global_max_pooling_2d(); layer_global_max_pooling_3d(): Global maximum pooling.

  • layer_global_average_pooling_1d(); layer_global_average_pooling_2d(); layer_global_average_pooling_3d(): Global average pooling.

Activation Layers

  • layer_activation(object, activation): Apply an activation function to an output.

  • layer_activation_leaky_relu(): Leaky version of a rectified linear unit.

  • layer_activation_parametric_relu(): Parametric rectified linear unit.

  • layer_activation_thresholded_relu(): Thresholded rectified linear unit.

  • layer_activation_elu(): Exponential linear unit.

Dropout Layers

  • layer_dropout(): Applies dropout to the input.

  • layer_spatial_dropout_1d(); layer_spatial_dropout_2d(); layer_spatial_dropout_3d(): Spatial 1D to 3D version of dropout

Recurrent Layers

  • layer_simple_rnn(): Fully-connected RNN where the output is to be fed back to input.

  • layer_gru(): Gated recurrent unit - Cho et al.

  • layer_cudnn_gru(): Fast GRU implementation backed by CuDNN.

  • layer_lstm(): Long-Short Term Memory unit - Hochreiter 1997.

  • layer_cudnn_lstm(): Fast LSTM implementation backed by CuDNN.

Locally Connected Layers

  • layer_locally_connected_1d(); layer_locally_connected_2d(): Similar to convolution, but weights are not shared, i.e. different filters for each patch.

Preprocessing

Sequence Preprocessing

  • pad_sequences(): Pads each sequence to the same length (length of the longest sequence).

  • skipgrams(): Generates skipgram word pairs.

  • make_sampling_table(): Generates word rank-based probabilistic sampling table.

Text Preprocessing

  • text_tokenizer(): Text tokenization utility.

  • fit_text_tokenizer(): Update tokenizer internal vocabulary.

  • save_text_tokenizer(); load_text_tokenizer(): Save a text tokenizer to an external file.

  • texts_to_sequences(); texts_to_sequences_generator(): Transforms each text in texts to sequence of integers.

  • texts_to_matrix(); sequences_to_matrix(): Convert a list of sequences into a matrix.

  • text_one_hot(): One-hot encode text to word indices.

  • text_hashing_trick(): Converts a text to a sequence of indexes in a fixed-size hashing space.

  • text_to_word_sequence(): Convert text to a sequence of words (or tokens).

Image Proprocessing

  • image_load(): Loads an image into PIL format.

  • flow_images_from_data(); flow_images_from_directory(): Generates batches of augmented/normalized data from images and labels, or a directory.

  • image_data_generator(): Generate minibatches of image data with real-time data augmentation.

  • fit_image_data_generator(): Fit image data generator internal statistics to some sample data.

  • generator_next(): Retrieve the next item.

  • image_to_array(); image_array_resize(); image_array_save(): 3D array representation.

Pre-trained models

Keras applications are deep learning models that are made available alongside pre-trained weights. These models can be used for prediction, feature extraction, and fine-tuning.

  • application_xception(); xception_preprocess_input(): Xception v1 model.

  • application_inception_v3(); inception_v3_preprocess_input(): Inception v3 model, with weights pre-trained on ImageNet.

  • application_inception_resnet_v2(); inception_resnet_v2_preprocess_input(): Inception-ResNet v2 model, with weights trained on ImageNet.

  • application_vgg16(); application_vgg19(): VGG16 and VGG19 models.

  • application_resnet50(): ResNet50 model.

  • application_mobilenet(); mobilenet_preprocess_input(); mobilenet_decode_predictions(); mobilenet_load_model_hdf5(): MobileNet model architecture.

ImageNet is a large database of images with labels, extensively used for deep learning.

  • imagenet_preprocess_input(); imagenet_decode_predictions(): Preprocesses a tensor encoding a batch of images for ImageNet, and decodes predictions.

Callbacks

A callback is a set of functions to be applied at given stages of the training procedure. You can use callbacks to get a view on internal states and statistics of the model during training.

  • allback_early_stopping(): Stop training when a monitored quantity has stopped improving.

  • callback_learning_rate_scheduler(): Learning rate scheduler.

  • callback_tensorboard(): TensorBoard basic visualizations.

CC BY SA Posit Software, PBC • info@posit.co • posit.co

Learn more at tensorflow.rstudio.com.

Updated: 2023-06.

packageVersion("keras")
[1] '2.11.1'