docs(classical-ml): added random forest example (#242)

Author: sutt0n

.mailmap

@@ -6,4 +6,5 @@
 #
 
 Marcus Cvjeticanin <[email protected]>
-Chase Willden <[email protected]>
+Chase Willden <[email protected]>
+Joseph "Jojo" Sutton <[email protected]>

Cargo.toml

@@ -7,7 +7,7 @@ members = [
     "examples/deep_learning/mnist",
     "examples/deep_learning/imagenet_v2",
     "examples/classical_ml/linear_regression",
-    "examples/classical_ml/logistic_regression",
+    "examples/classical_ml/logistic_regression", "examples/classical_ml/random_forest",
 ]
 resolver = "2"
 
@@ -33,4 +33,4 @@ codegen-units = 1
 
 [profile.bench]
 opt-level = 3
-debug = false
+debug = false

delta/src/classical_ml/algorithms.rs

@@ -313,6 +313,10 @@ where
         }
         counts.into_iter().max_by_key(|&(_, count)| count).map(|(pred, _)| pred).unwrap()
     }
+
+    pub fn calculate_loss(&self, predictions: &Array1<T>, actuals: &Array1<T>) -> T {
+        self.loss_function.calculate(predictions, actuals)
+    }
 }
 
 impl<T, L> Algorithm<T, L> for RandomForest<T, L>
@@ -828,9 +832,7 @@ mod tests {
 
         let mut model = RandomForest::new(CrossEntropy);
 
-        println!("Fitting the model...");
         model.fit(&x_train, &y_train, 0.1, 100);
-        println!("Model fitted.");
 
         let predictions = model.predict(&x_test);
 
@@ -840,7 +842,13 @@ mod tests {
             .zip(y_test.iter())
             .filter(|(&pred, &actual)| (pred - actual).abs() < 1e-6)
             .count();
+
         let accuracy = correct_predictions as f64 / y_test.len() as f64;
+        let loss = model.calculate_loss(&predictions, &y_test);
+
         println!("Test accuracy: {:.2}%", accuracy * 100.0);
+
+        assert!(accuracy > 0.0, "Accuracy should be positive, got: {}", accuracy);
+        assert!(loss < 0.0, "Loss should be less than 0, got: {}", loss);
     }
 }

examples/classical_ml/random_forest/Cargo.toml

@@ -0,0 +1,9 @@
+[package]
+name = "random_forest"
+version = "0.1.0"
+edition = "2024"
+publish = false
+
+[dependencies]
+deltaml = { path = "../../../delta", features = ["classical_ml"] }
+tokio = { workspace = true, features = ["full"] }

examples/classical_ml/random_forest/src/main.rs

@@ -0,0 +1,51 @@
+use deltaml::{
+    classical_ml::{Algorithm, algorithms::RandomForest, losses::MSE},
+    ndarray::{Array1, Array2},
+};
+
+#[tokio::main]
+async fn main() {
+    // create feature data
+    let x_data = Array2::from_shape_vec(
+        (4, 4),
+        vec![
+            25.0,     // 25 years old
+            30.0,     // 30 years old
+            22.0,     // 22 years old
+            40.0,     // 40 years old
+            0.0,      // male
+            1.0,      // female
+            0.0,      // male
+            1.0,      // female
+            1.0,      // engineer
+            2.0,      // doctor
+            3.0,      // student
+            4.0,      // manager
+            60000.0,  // 60000.0 salary
+            80000.0,  // 80000.0 salary
+            20000.0,  // 20000.0 salary
+            100000.0, // 100000.0 salary
+        ],
+    )
+    .unwrap();
+
+    let y_data = Array1::from_vec(vec![1.0, 1.0, 0.0, 1.0]); // high, high, low, high
+
+    // Instantiate the model
+    let mut model = RandomForest::new(MSE);
+
+    // Train the model
+    let learning_rate = 0.01;
+    let epochs = 1000;
+    model.fit(&x_data, &y_data, learning_rate, epochs);
+
+    // Make predictions with the trained model
+    let new_data = Array2::from_shape_vec((4, 1), vec![28.0, 1.0, 1.0, 70000.0]).unwrap();
+    let predictions = model.predict(&new_data);
+
+    println!("Predictions for new data: {:?}", predictions);
+
+    // Calculate accuracy or loss for the test data for demonstration
+    let test_loss = model.calculate_loss(&model.predict(&x_data), &y_data);
+    println!("Test Loss after training: {:.6}", test_loss);
+}