Skip to content
Toynn Docs ToyNN Docs

Intro to Errors

Errors contains statistical error functions.

Currently there are no functions or classes to get error regarding classifiers

Usage

import toynn from 'toynn';


let yTrue = [1, 2, 3, 4, 5];
let yPred = [1.1, 1.99, 3, 4.05, 5];


console.log(toynn.errors.RSS(yTrue, yPred).result); // 0.0126
console.log(toynn.errors.RSS(yTrue, yPred).formula); // sum(square(yTrue - yPred))

Premade Errors

Errors comes with a lot of error functions already implemented. Following are the functions that comes pre loaded with Errors:

  1. error :- Returns a list of y^ - y,
  2. MAE,
  3. MSE,
  4. RSS,
  5. RSME

Custom Error

You can use StatError class to create a custom error.

Note: While creating custom error do not use anonymous functions

You can use StatError class’s object to create a custom error.

  • use(func: Function): StatErrorReturn : Takes in a function and use it to transform loss value for each row. Can’t be chained,

  • apply(func: Function): StatErrorReturn : Takes in a function and use it to compute result value. Can be chained.

Note: You should always pass a named function inside use and apply or else they’ll throw Error

Example

The RSS error function is implemented as follows:

function RSS(yTrue: StatErrorInput, yPred: StatErrorInput): StatErrorReturn {
    return new StatError(yTrue, yPred).use(square).apply(sum);
}

Extra Funtions

Errors comes with 3 more extra functions

  1. mean,
  2. sum,
  3. square

You can use these instead of creating one from scratch.