** Next:** Adaptive sequence chunkers
** Up:** Remedies
** Previous:** Searching without gradients

Bengio and Frasconi [4]
propose a probabilistic approach for propagating targets.
With so-called ``state networks'', at a given
time, their system
can be in one of only different discrete states.
Parameters are
adjusted using the expectation-maximization algorithm.
But to solve problems that require a significant amount of memory to
store contextual information,
such systems would require an unacceptable number
of states (i.e., state networks).

Juergen Schmidhuber
2003-02-19