Abstract | ||
---|---|---|
Most of statistical machine translation systems are combi- nations of various models, and tuning of the scaling fac- tors is an important step. However, this optimisation prob- lem is hard because the objective function has many local minima and the available algorithms cannot achieve a global optimum. Consequently, optimisations starting from differ- ent initial settings can converge to fairly different solutions. We present tuning experiments with the Simultaneous Per- turbation Stochastic Approximation (SPSA) algorithm, and compare them to tuning with the widely used downhill sim- plex method. With IWSLT 2006 Chinese-English data, both methods showed similar performance, but SPSA was more robust to the choice of initial settings. |
Year | Venue | Keywords |
---|---|---|
2006 | IWSLT | machine translation,local minima,objective function,stochastic approximation |
Field | DocType | Citations |
Mathematical optimization,Simultaneous perturbation stochastic approximation,Computer science,Machine translation,Global optimum,Maxima and minima,Nelder–Mead method,Scaling | Conference | 5 |
PageRank | References | Authors |
0.46 | 11 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Patrik Lambert | 1 | 277 | 23.36 |
Rafael E. Banchs | 2 | 566 | 63.64 |