deepbox/metrics

Regression Metrics

Evaluate regression model performance by measuring the distance between predicted and actual continuous values.

mse

mse(yTrue: Tensor, yPred: Tensor): number

Mean Squared Error: (1/n) Σᵢ (yᵢ − ŷᵢ)². Penalizes large errors quadratically. The most common regression metric.

rmse

rmse(yTrue: Tensor, yPred: Tensor): number

Root Mean Squared Error: √MSE. Same units as the target variable, making it more interpretable than MSE.

mae

mae(yTrue: Tensor, yPred: Tensor): number

Mean Absolute Error: (1/n) Σᵢ |yᵢ − ŷᵢ|. More robust to outliers than MSE. Linear penalty for errors.

r2Score

r2Score(yTrue: Tensor, yPred: Tensor): number

Coefficient of determination: R² = 1 − SS_res / SS_tot. 1.0 = perfect prediction, 0.0 = predicting the mean, negative = worse than mean.

adjustedR2Score

adjustedR2Score(yTrue: Tensor, yPred: Tensor, nFeatures: number): number

Adjusted R² that penalizes for number of features. Useful for comparing models with different numbers of predictors.

mape

mape(yTrue: Tensor, yPred: Tensor): number

Mean Absolute Percentage Error: (100/n) Σᵢ |yᵢ − ŷᵢ| / |yᵢ|. Percentage-based, scale-independent. Undefined when true values are zero.

medianAbsoluteError

medianAbsoluteError(yTrue: Tensor, yPred: Tensor): number

Median of absolute errors. Extremely robust to outliers.

explainedVarianceScore

explainedVarianceScore(yTrue: Tensor, yPred: Tensor): number

Fraction of variance in y explained by predictions. Similar to R² but does not penalize for bias.

maxError

maxError(yTrue: Tensor, yPred: Tensor): number

Maximum absolute error: max |yᵢ − ŷᵢ|. Captures the worst-case prediction error.

MSE

MSE = (1/n) Σᵢ (yᵢ − ŷᵢ)²

Where:

ŷᵢ = Predicted value

R²

R² = 1 − Σᵢ(yᵢ − ŷᵢ)² / Σᵢ(yᵢ − ȳ)²

Where:

ȳ = Mean of true values

regression-metrics.ts

import { mse, rmse, mae, r2Score } from "deepbox/metrics";import { tensor } from "deepbox/ndarray";const yTrue = tensor([3.0, -0.5, 2.0, 7.0]);const yPred = tensor([2.5, 0.0, 2.1, 7.8]);mse(yTrue, yPred);    // ~0.375rmse(yTrue, yPred);   // ~0.612mae(yTrue, yPred);    // 0.425r2Score(yTrue, yPred); // ~0.948

Classification Metrics

Clustering Metrics