pub fn normalize_dataset<T: TaskLabelType + Copy>(
dataset: &mut Dataset<T>,
scaler: ScalerType,
)Examples found in repository?
examples/02_logistic_regression.rs (line 14)
10fn main() -> std::io::Result<()> {
11 let path = ".data/MobilePhonePricePredict/train.csv";
12 let mut dataset: Dataset<usize> = Dataset::<usize>::from_name(path, DatasetName::MobilePhonePricePredictDataset, None);
13
14 normalize_dataset(&mut dataset, ScalerType::Standard);
15
16 let mut res = dataset.split_dataset(vec![0.8, 0.2], 0);
17 let (train_dataset, test_dataset) = (res.remove(0), res.remove(0));
18
19 let mut rng = RandGenerator::new(0);
20
21 // initing weights is also important, or the gradients may not be good
22 let mut model = LogisticRegression::new(train_dataset.feature_len(), train_dataset.class_num(), Some(Penalty::RidgeL2(1e-1)),|item| item.iter_mut().for_each(move |i| {*i = rng.gen_f32()}));
23
24 let mut train_dataloader = Dataloader::new(train_dataset, 8, true, None);
25
26 const EPOCH: usize = 3000;
27 let mut best_acc = vec![];
28 for ep in 0..EPOCH {
29 let mut losses = vec![];
30 let lr = match ep {
31 i if i < 200 => 1e-3,
32 _ => 2e-3,
33 };
34 for (feature, label) in train_dataloader.iter_mut() {
35 let loss = model.one_step(&feature, &label, lr, Some(NormType::L2(1.0)));
36 losses.push(loss);
37 }
38 let (_, acc) = evaluate(&test_dataset, &model);
39 best_acc.push(acc);
40 let width = ">".repeat(ep * 100 / EPOCH);
41 print!("\r{width:-<100}\t{:.3}\t{acc:.3}", losses.iter().sum::<f32>() / losses.len() as f32);
42 stdout().flush()?;
43 }
44 let acc = best_acc.iter().fold(0.0, |s, i| f32::max(s, *i));
45 let best_ep = argmax(&NdArray::new(best_acc), 0);
46 println!("\nbest acc = {acc} ep {}", best_ep[0][0]);
47 assert!(acc > 0.9); // gradient clip greatly helps it
48
49 Ok(())
50}