Basic bayesian linear regression with training data
enter image description here
Bayesian linear regression with training data
$text{Perform a regression task where } x text{ is a two-dimensional vector and } phi(x) = [1, x_1^2, x_2^3]^T
t_i = w^T phi(x_i) + varepsilon = w_0 + w_1x_{1i}^2 + w_2x_{2i}^3 + varepsilon, text{ where } varepsilon sim mathcal{N}(0, sigma^2)$ Let $w = [0, 2.5, −0.5]$
Perform a batch mode instead of processing samples individually. This means your model will learn from all available training data simultaneously, without the need for iterative updates involving priors and posteriors.