| 26 |
| etc. |
|
|
|
|
| • | The learning rule
for a logistic DAG is: |
||
| • | With replicated
weights this becomes: |
||
| h2 |
| v2 |
| h1 |
| v1 |
| h0 |
|
|
|
| The derivatives |
||
| for the
recognition |
||
| weights
are zero. |
||
| v0 |