* Load metadata for a model. * @param {string} name Name of model. Should be a key in modelUrls

* Load IMDB targets from a file. * * @param {string} filePath Path to the binary targets file. * @return {tf.Tensor} The targets as `tf.Tensor` of

* Load the universal sentence encoder model

* Generate a number of new fruits at a random locations. * * The number of frtuis created is such that the total number of * fruits will be e

* Get a file by downloading it if necessary. * * @param {string} sourceURL URL to download the file from. * @param {string} destPath Destination fi

* Get extracted files. * * If the files are already extracted, this will be a no-op. * * @param {string} sourcePath Source zip file path. * @para

* Calculate the arithmetic mean of a vector. * * @param {Array} vector The vector represented as an Array of Numbers. * * @returns {number} The ar

* Get model identifier. * * @returns {string} The model identifier.

* A function invoked at the end of a training iteration. * * @param {number} iterationCount A count of how many iterations has completed * so far

* Push a new dictionary of gradients into records. * * @param {{[varName: string]: tf.Tensor[]}} record The record of variable * gradient:

Simulate weight quantization. Args: w: (a numpy.ndarray) The weight to be quantized. bits: (int) number of bits used for the quantizatio

* Render the current state of the system on an HTML canvas. * * @param {CartPole} cartPole The instance of cart-pole system to render. * @param {HT

Reset the game state.

* Run the model in case of image segmentation.

* Randomly sample a batch of items from the replay buffer. * * The sampling is done *without* replacement. * * @param {number} batchSize S

* Shuffles data and target (maintaining alignment) using Fisher-Yates * algorithm.flab

* Collects one random play of the game. Processes the sample to generate * features and labels representation of the play. Calls a UI method to ren

* Tokenize a sentence and tag the tokens. * * @param {string} sentence sentence to tag * @param {string} model name of model to use * * @return {

* The combined encoder-decoder pipeline. * * @param {tf.Model} encoder * @param {tf.Model} decoder * * @returns {tf.Model} the vae.

* The custom loss function for VAE. * * @param {tf.tensor} inputs the encoder inputs a batched image tensor * @param {[tf.tensor]} outputs the vae

* Visualize layers of a model. * * @param {string} tab Name of the tfjs-vis visor tab on which the visualization * will be made. * @param {tf.la

Write a batch of MNIST examples to a JavaScript (.js) file. Args: x: A numpy array representing the image data, with shape (NUM_EXAMPLES,

Convert indices to one-hot encoding. Args: indices: A list of `int` indices with length `n`, each eleemnt of which is assumed to be a zer

* Adds an example to the controller dataset. * @param {Tensor} example A tensor representing the example. It can be an image, * an activatio

* Moves the provided imgNode into a container div, and adds a text div as a * peer. Styles the container div and text div to place the text * on t

* Adds an uploaded image and returns its index.

* Triggers the model to make a prediction on the image referenced by the * image data. After a successful prediction a IMAGE_CLICK_PROCESSED messag

* Convert an image represented as a typed array to a JIMP object. * * @param {Float32Array} imageData * * @returns {Promise[Jimp]} Jimp object rep

* Take an array of images (represented as typedarrays) and return * a tensor representing them. * * @param {Float32Array[]} imagesData * * @retur

* Build a combined ACGAN model. * * @param {number} latentSize Size of the latent vector. * @param {tf.SymbolicTensor} imageClass Symbolic tensor f

* Build the discriminator part of ACGAN. * * The discriminator model of ACGAN takes the input: an image of * MNIST format, of shape [batchSize, 28,

* Build the generator part of ACGAN. * * The generator of ACGAN takes two inputs: * * 1. A random latent-space vector (the latent space is often

* Build a linear-regression model for the temperature-prediction problem. * * @param {tf.Shape} inputShape Input shape (without the batch dimenson).

* Build a new head (i.e., output sub-model) that will be connected to * the top of the truncated base for object detection. * * @param {tf.Shape} i

* Builds object-detection model from MobileNet. * * @returns {[tf.Model, tf.layers.Layer[]]} * 1. The newly-built model for simple object detecti

* Calculate the means and standard deviations of every column. * * TensorFlow.js is used for acceleration.

* Compile model for training. * * @param {number} learningRate The learning rate to use during training.

* Convert all training text to integer indices.

* Convert Iris data arrays to `tf.Tensor`s. * * @param data The Iris input feature data, an `Array` of `Array`s, each element * of which is assum

* Creates a convolutional neural network (Convnet) for the MNIST data. * * @returns {tf.Model} An instance of tf.Model.

* Returns a three layer sequential model suitable for predicting win state from * feature representation. The input shape depends on whether oneHot

* Creates a model consisting of only flatten, dense and dropout layers. * * The model create here has approximately the same number of parameters *

* Create and material-design card for a search match and add * it to the root div for search results.

* Create LSTM model from scratch. * * @param {number | number[]} lstmLayerSizes Sizes of the LSTM layers, as a * number or an non-empty arr

* Create the underlying model of this policy network. * * @param {number | number[]} hiddenLayerSizes Size of the hidden layer, as * a sing

* Create sliders for the latent space. * * @param {tf.Model} generator The generator part of the trained ACGAN.

* Get data identifier. * * @returns {string} The data identifier.

* This is pulled into a separate function to isolate the async code. * @see datasetToArrayHandler

* Convert a 2D tensor into a string with the CharacterTable's vocabulary. * * @param x Input 2D tensor. * @param calcArgmax Whether to perfor

* Convert a 2D tensor into a string with the CharacterTable's vocabulary. * * @param x Input 2D tensor. * @param calcArgmax Whether to perfor

Decode (i.e., translate) an encoded sentence. Args: input_seq: A `numpy.ndarray` of shape `(1, max_encoder_seq_length, num_encoder_tokens)`.