tks0123456789 / XGB_experiments

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

XGBoost experiments

name #instances #features Task #classes
Covertype Data Set 581012 54 Classification 7
Class #Instances
1| 283301
2| 211840
3|  35754
4|   2747
5|   9493
6|  17367
7|  20510
file name description data
XGBoost_refresh.ipynb updater=refresh Covtype
  • exp001
    • model : CPU, GPU, hist_256, hist_1024
    • objective : Binary classification
    • metric : Logloss
    • dataset : make_classification
    • n_train : 10K, 100K, 1M, 2M, 4M
    • n_valid : n_train/4
    • n_features : 32
    • n_rounds : 50
    • n_clusters_per_class: 8
    • max_depth : 5, 10, 15
  • exp002
    • model : hist_depthwise, hist_lossguide
    • objective : Binary classification
    • metric : Logloss
    • dataset : make_classification
    • n_train : 10K, 100K, 1M, 2M, 4M
    • n_valid : n_train/4
    • n_features : 32
    • n_rounds : 50
    • n_clusters_per_class: 8
    • max_depth : 10, 15, 20
    • num_leaves : 256, 1024, 4096

About

License:MIT License


Languages

Language:Jupyter Notebook 89.4%Language:Python 10.6%