## P (C0) and P (C1): use maximum likelihood estimates

## P (x | C0) and P (x | C1): Use Parzen Windows

- ? Gaussian kernels
- Use ?2I as the covariance matrix for each kernel

- Choice of the smoothing parameter ? is often crucial:
- small ? ? lower bias but higher variance estimates of P

Previous slide | Next slide | Back to first slide | View graphic version |