Wiki
emc-sola / Home
Rezultati
4-fold Cross-validation
| Log loss | Metoda |
|---|---|
| 0.289969772165 | Logistic (lambda=0.05, bias=1), df>10, log10idf, "kalibracija" |
| 0.312606261226 | Softmax (lambda=0.05, bias=1, 60k features), logidf |
| 0.311617325311 | Logistic (lambda=0.05, bias=1, 100k features), logidf |
| 0.473830260423 | MLP (lambda=0.05, hidden=20, maxfun=300, 50k features), logidf |
Teme
- Classifiers
- Logistic regression [Blaz]
- Softmax regression [Jure]
- Neural nets [Marinka]
- Random forest [Marko]
- Trees (adaboost et al., gradient boosting) [Jure]
- Naive Bayes [Janez?]
- SVM [Miha]
- Clustering [Lan Z.]
- Multi-target [Matija]
- Dimensionality reduction / feature selection
- LDA
- PCA [Matija, Lan U. ?????]
- Feature selection
- NNMF [Marinka]
- Misc
- Data overview [Janez?]
- Cross-validation [Jure]
- tf*idf [Lan Z.]
- Calibration [Lan Z., Jure, Marinka]
- Stacking [Blaz]
Prevozi
https://docs.google.com/spreadsheet/ccc?key=0Akez4sX_YKZQdDVoQ3F1REI4WHpZZWwzakpWTlljaEE#gid=0
Urnik
https://docs.google.com/spreadsheet/ccc?key=0Akez4sX_YKZQdHZNQU9wTTNaWV9yRVVRWkF6YWhITlE
Stvari za sabo
- rezervni laptop
Updated