next up previous
Next: A.1. FLAT NETS: THE Up: FLAT MINIMA NEURAL COMPUTATION Previous: LIMITATIONS / FINAL REMARKS

APPENDIX - THEORETICAL JUSTIFICATION

Contents:
A.1 Flat nets: the most probable hypotheses
A.2 Why does the Hessian decrease?
A.3 Efficient implementation of the algorithm
A.3.1 Explicit derivative of equation (1)
A.3.2 Fast multiplication by the Hessian
A.4 The algorithm in pseudo code

Note: Appendices A.3.2 and A.4 were omitted from the version for Neural Computation. An alternative version of the entire appendix (but with some minor errors) can be found in Hochreiter & Schmidhuber (1994).



Subsections

Juergen Schmidhuber 2003-02-13


Back to Financial Forecasting page