Skip to content

Document decision-tree baseline models and default sample sizes #3

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
hongqin wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Document decision-tree baseline models and default sample sizes #3

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions decision_tree/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,21 @@
## 1. Preparing data
Because of limitations on memory, for decision trees we cannot use the entire training and test datasets we used for the 2-layer CNN, then we selected 2,000 sequences from each VOC, splitting 20% of them for testing.

### Baseline models and sample sizes
The tree-based baselines used to compare against the CNN are:

* Random Forest (`rf`)
* XGBoost (`xgb`)
* CatBoost (`cat`)

With the default `-num 2000` setting in `data_dt.py` and five VOC classes, this creates:

* 10,000 total sequences for decision-tree experiments
* 8,000 training sequences (80%)
* 2,000 test sequences (20%)

For hyperparameter fine-tuning, `fine_tunning.py` uses 500 sequences per VOC by default (2,500 total sequences).

To create these datasets, first create a folder named `data`, then run:
```
python3 <root_repository>/decision_tree/data_dt.py
Expand Down