Hello World: Linear Regression (y = 2x + 1)

Hello World: Linear Regression (y = 2x + 1)

Objective

Train a simple regression model, deploy it, run predictions, then verify results match the known function.

Step 1: Create table and load data

CREATE TABLE hw_train (x DOUBLE, y DOUBLE);

INSERT INTO hw_train (x, y) VALUES
(0, 1),(1, 3),(2, 5),(3, 7),(4, 9),
(5, 11),(6, 13),(7, 15);

Step 2: Create AutoML experiment

CREATE EXPERIMENT hw_linreg AS
SELECT
  x,
  y AS target
FROM hw_train
WITH (
  task_type = 'regression',
  target_column = 'target',
  algorithms = ['linear_regression','gradient_boosting'],
  optimization_metric = 'rmse',
  max_trials = 10,
  validation_strategy = 'kfold',
  n_folds = 3
);

Step 3: Deploy best model

DEPLOY MODEL hw_model FROM EXPERIMENT hw_linreg;

Step 4: Predict on new values

-- Values to predict
  PREDICT y_hat USING hw_model AS
  SELECT 8 as x

Step 5: Validate predictions against expected values

Predict against the test set inline via AUTOML.PREDICT(...) and put the predicted y_hat next to y_expected so they can be compared side-by-side.

-- Expected y for y = 2x + 1
CREATE TABLE hw_expected (x DOUBLE, y_expected DOUBLE);
INSERT INTO hw_expected VALUES (8.0, 17.0), (12.5, 26.0);

-- Predict inline and project alongside the expected values
SELECT
  x,
  AUTOML.PREDICT('hw_model', x) AS y_hat,
  y_expected
FROM hw_expected;

The result returns one row per test point with y_hat and y_expected side-by-side. For y = 2x + 1 the trained model should produce y_hat ≈ 17.0 for x = 8.0 and y_hat ≈ 26.0 for x = 12.5.

Optional: Inspect experiment and trials

-- Experiment overview
SELECT * FROM automl_experiments WHERE name = 'hw_linreg';

-- Top trials by RMSE
SELECT trial_id, model_type, hyperparameters, cv_score, training_time
FROM automl_trials
WHERE experiment_id = (SELECT id FROM automl_experiments WHERE name = 'hw_linreg')
ORDER BY cv_score ASC
LIMIT 5;

Expected Outcomes

• Best algorithm: linear_regression.

• Coefficients: slope ≈ 2.0, intercept ≈ 1.0.

• Quality: RMSE ≈ 0 on hw_train.

• Prediction examples:

  • Input x = 8.0 → y_hat = 17.0.
  • Input x = 12.5 → y_hat = 26.0.

• Validation query: mismatches = 0.