this is for holding javascript data
Xavier Holt edited subsection_Prior_for_Weights_In__.tex
over 8 years ago
Commit id: f6a40099050cdadccb07d0675dc6162dd956c5e6
deletions | additions
diff --git a/subsection_Prior_for_Weights_In__.tex b/subsection_Prior_for_Weights_In__.tex
index 3813dae..c7f7b04 100644
--- a/subsection_Prior_for_Weights_In__.tex
+++ b/subsection_Prior_for_Weights_In__.tex
...
In pursuit of generalisability, we adopt the framework that values should be at or near zero with high probability. The most widely used weight-distribution in logistic regression is Guassian, so we take this approach as a useful starting point.
\subsubsection{Guassian Prior: $\mathbf{w} \sim
\mathcal{N}(\mathbf{0}, \mathcal{N}_{d\times d}(\mathbf{0}, \text{diag}(\boldsymbol{\sigma}))$}
Clearly having a zero mean vector is desirable. Additionally, we have opted to assume that the weights are heteroscedastic but independent of one another. That is, the covariance matrix is a diagonal $d\times d$ matrix $\Sigma = \text{diag}(\boldsymbol{\sigma}) = \text{diag}(\sigma_1, \sigma_2, \dots \sigma_d)$. Consequently,