123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177
import { Tensor } from "./tensor/tensor.ts";import { Rank, Shape } from "./api/shape.ts";import { Layer } from "./api/layer.ts";
/** * The Backend is responsible for eveything related to the neural network. */export interface Backend {  /**   * The train method is a function that trains a neural network using a set of training data.   * It takes in an array of DataSet objects, the number of epochs to train for, and the learning rate.   * The method modifies the weights and biases of the network to minimize the cost function and improve its accuracy on the training data.   */  train(    datasets: DataSet[],    epochs: number,    batches: number,    rate: number,  ): void;
  /**   * The predict method is a function that takes in a Tensor object   * representing the input to the neural network and returns a Promise that resolves to a Tensor object representing the output of the network.   * This method is used to make predictions on new data after the network has been trained.   */  predict(input: Tensor<Rank>): Promise<Tensor<Rank>>;
  /**   * The save method is a function that saves the network to a Uint8Array.   * This method is used to save the network after it has been trained.   */  save(): Uint8Array;
  /**   * The saveFile method is a function that takes in a string representing the path to a file and saves the network to that file.   * This method is used to save the network after it has been trained.   */  saveFile(path: string): void;}
/** * NetworkConfig represents the configuration of a neural network. */export type NetworkConfig = {  size: Shape[Rank];  layers: Layer[];  cost: Cost;  silent?: boolean;};
/** * Activation functions are used to transform the output of a layer into a new output. */export enum Activation {  /**   * Sigmoid activation function f(x) = 1 / (1 + e^(-x))   */  Sigmoid = "sigmoid",
  /**   * Tanh activation function f(x) = (e^x - e^-x) / (e^x + e^-x)   * This is the same as the sigmoid function, but is more robust to outliers   */  Tanh = "tanh",
  /**   * ReLU activation function f(x) = max(0, x)   * This is a rectified linear unit, which is a smooth approximation to the sigmoid function.   */  Relu = "relu",
  /**   * Relu6 activation function f(x) = min(max(0, x), 6)   * This is a rectified linear unit with a 6-value output range.   */  Relu6 = "relu6",
  /**   * Leaky ReLU activation function f(x) = x if x > 0, 0.01 * x otherwise   */  LeakyRelu = "leakyrelu",
  /**   * Elu activation function f(x) = x if x >= 0, 1.01 * (e^x - 1) otherwise   * This is a rectified linear unit with an exponential output range.   */  Elu = "elu",
  /**   * Selu activation function f(x) = x if x >= 0, 1.67 * (e^x - 1) otherwise   * This is a scaled version of the Elu function, which is a smoother approximation to the ReLU function.   */  Selu = "selu",}
export enum Cost {  /**   * Cross entropy cost function is the standard cost function for binary classification.   */  CrossEntropy = "crossentropy",
  /**   * Hinge cost function is the standard cost function for multi-class classification.   */  Hinge = "hinge",
  /**   * Mean squared error cost function is the standard cost function for regression.   */  MSE = "mse",}
/** * DataSet is a container for training data. */export type DataSet = {  inputs: Tensor<Rank>;  outputs: Tensor<Rank>;};
export enum LayerType {  Activation = "activation",  BatchNorm2D = "batchnorm2d",  Conv2D = "conv2d",  Dense = "dense",  Dropout1D = "dropout1d",  Dropout2D = "dropout2d",  Pool2D = "pool2d",  Flatten = "flatten",  Softmax = "softmax",}
/** * BackendType represents the type of backend to use. */export enum BackendType {  /**   * CPU backend   */  CPU = "cpu",
  /**   * GPU backend   */  GPU = "gpu",
  /**   * Web Assembly backend   */  WASM = "wasm",}
/** * Init represents the type of initialization to use. */export enum Init {  /**   * Uniform initialization   */  Uniform = "uniform",
  /**   * Xavier initialization   */  Xavier = "xavier",
  /**   * XavierN initialization   */  XavierN = "xaviern",
  /**   * Kaiming initialization   */  Kaiming = "kaiming",}