Motivating the "Gaussian" aspect of GP

37 Views Asked by At

In one of his lectures notes on Gaussian Process Regression, Kilian Weinberger notes that

$$ P(Y\mid D,X) = \int_{\mathbf{w}}P(Y,\mathbf{w} \mid D,X) d\mathbf{w} = \int_{\mathbf{w}} P(Y \mid \mathbf{w}, D,X) P(\mathbf{w} \mid D) d\mathbf{w} $$

Using this equation, he motivates Gaussian Process Regression, stating that multiplying two Gaussians will also produce a Gaussian.

My intuition tells me that the equation above has to do with the definition of conditional probability, marginalization, and perhaps the chain rule of probability. However, I am unable to hash out the intermediary steps to go from $P(Y\mid D,X)$ to the final integral.

I'm aware that

$$ p(a) = \int_{b} p(a, b) \, db $$

from which we can easily deduce something like

$$ p(a \vert c) = \int_{b} p(a, b \vert c) \, db $$

I'm also aware of the chain rule of probability, though I'm not entire sure if this is relevant in deriving the desired conclusion.

$$ p(a, b, c) = p(a \vert b, c) \cdot p(b \vert c) \cdot p(c) $$

Thanks in advance for your help!