Suppose that we have the data (tau_{mathrm{n}}=left{x_{1}, ldots, x_{n} ight}) in (mathbb{R}) and decide to train the
Question:
Suppose that we have the data \(\tau_{\mathrm{n}}=\left\{x_{1}, \ldots, x_{n}\right\}\) in \(\mathbb{R}\) and decide to train the two-component Gaussian mixture model
\[ g(x \mid \boldsymbol{\theta})=w_{1} \frac{1}{\sqrt{2 \pi \sigma_{1}^{2}}} \exp \left(-\frac{\left(x-\mu_{1}\right)^{2}}{2 \sigma_{1}^{2}}\right)+w_{2} \frac{1}{\sqrt{2 \pi \sigma_{2}^{2}}} \exp \left(-\frac{\left(x-\mu_{2}\right)}{2 \sigma_{2}^{2}}\right) \]
where the parameter vector \(\boldsymbol{\theta}=\left[\mu_{1}, \mu_{2}, \sigma_{1}, \sigma_{2}, w_{1}, w_{2}\right]^{\top}\) belongs to the set
\[ \Theta=\left\{\boldsymbol{\theta}: w_{1}+w_{2}=1, w_{1} \in[0,1], \mu_{i} \in \mathbb{R}, \sigma_{i}>0, \forall i\right\} \]
Suppose that the training is via the maximum likelihood in (2.28). Show that
\[ \sup _{\boldsymbol{\theta} \in \Theta} \frac{1}{n} \sum_{i=1}^{n} \ln g\left(x_{i} \mid \boldsymbol{\theta}\right)=\infty \]
In other words, find a sequence of values for \(\theta \in \Theta\) such that the likelihood grows without bound. How can we restrict the set \(\Theta\) to ensure that the likelihood remains bounded?
Step by Step Answer:
Data Science And Machine Learning Mathematical And Statistical Methods
ISBN: 9781118710852
1st Edition
Authors: Dirk P. Kroese, Thomas Taimre, Radislav Vaisman, Zdravko Botev