University AI - Train an ANN in an Unsupervised Manner

Can we train a feed-forward neural network $\varphi (\underline{x})$ in an unsupervised manner, to estimate a PDF $p(\underline{x})$ ?

⇒ Yes, we can, but it’s tricky

Using an RBF

We define the cost function:

$$C(\tau ,W)=p(\tau |W)$$

Where $p(\tau |W)$ is the Likelihood: $p(\tau |\underline{\Theta })={\prod }_{k=1}^{n}p(\underline{x_k}|\underline{\Theta })$ Where $p(\underline{x_k}|W)$ is defined as:

$$p(\underline{x}|\underline{\Theta })=\sum _{j=1}^{c}P({\omega }_j)p(\underline{x}|{\omega }_j,\underline{{\Theta }_j})$$

This formula means thatTODO

---

Original Files