examples: add basic calibration of AR model written in Java

Author: AldoGl

examples/AR1_process_in_Java.ipynb

@@ -0,0 +1,27 @@
+import java.util.Random;
+
+public class ARModel {
+
+    public static void main(String[] args) {
+
+        double constant = Double.parseDouble(args[0]);
+        double autoregressiveParameter = Double.parseDouble(args[1]);
+
+        int nPeriods = Integer.parseInt(args[2]);
+        long seed = Long.parseLong(args[3]);
+
+        double[] timeSeries = new double[nPeriods];
+        timeSeries[0] = 0;
+
+        Random rnd = new Random();
+        rnd.setSeed(seed);
+
+        for (int i = 1; i < nPeriods; i++) {
+            timeSeries[i] = constant + autoregressiveParameter * timeSeries[i-1] + rnd.nextGaussian();
+       }
+
+	   for (int i = 0; i < nPeriods; i++) {
+	        System.out.println(timeSeries[i]);
+	   }
+    }
+}

examples/models/ar_model_java/ARModel.java

@@ -0,0 +1,119 @@
+import os
+import subprocess
+
+import numpy as np
+
+
+def ar1_model(theta, N, rndSeed=0):
+    """Autoregressive model.
+
+    A simple AR(1) model. This is a Python wrapper for an underlying
+        Java implementation. In order to run this code you need to first
+        compile the Java code via "javac ARModel.java".
+
+    Args:
+        theta: the two parameters of the process
+        N: the length of the generated time series
+        rndSeed: the random seed of the simulation
+
+    Returns:
+        the generated time series
+    """
+
+    # AR(1) constant term
+    const = theta[0]
+    # AR(1) multiplicative term
+    mul_par = theta[1]
+
+    # the path of the Java executable
+    file_path = os.path.realpath(os.path.dirname(__file__))
+
+    command = "java -classpath " + file_path + " ARModel {} {} {} {}".format(
+        const, mul_par, N, rndSeed
+    )
+
+    res = subprocess.run(
+        command.split(),
+        shell=False,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+    )
+
+    stdout = res.stdout
+
+    # remove first lines and last line
+    lines = stdout.split("\n")
+
+    # parse the result of the simulation
+    time_series = []
+    for line in lines[:-1]:
+        
+        splitted_line = line.split()
+        time_series.append(float(splitted_line[-1]))
+
+    time_series = np.array([time_series]).T
+
+    return time_series
+
+
+def ar1_model_not_random(theta, N, rndSeed=0):
+    """Autoregressive model.
+
+    A simple AR(1) model. This is a Python wrapper for an underlying
+        Java implementation. In order to run this code you need to first
+        compile the Java code via "javac ARModel.java".
+
+    Args:
+        theta: the two parameters of the process
+        N: the length of the generated time series
+        rndSeed: the random seed of the simulation
+
+    Returns:
+        the generated time series
+    """
+
+    # AR(1) constant term
+    const = theta[0]
+    # AR(1) multiplicative term
+    mul_par = theta[1]
+
+    # the path of the Java executable
+    file_path = os.path.realpath(os.path.dirname(__file__))
+
+    # fixed seed to zero in this
+    seed = 0
+    command = "java -classpath " + file_path + " ARModel {} {} {} {}".format(
+        const, mul_par, N, rndSeed
+    )
+
+    res = subprocess.run(
+        command.split(),
+        shell=False,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+    )
+
+    stdout = res.stdout
+
+    # remove first lines and last line
+    lines = stdout.split("\n")
+
+    # parse the result of the simulation
+    time_series = []
+    for line in lines[:-1]:
+        splitted_line = line.split()
+        time_series.append(float(splitted_line[-1]))
+
+    time_series = np.array([time_series]).T
+
+    return time_series
+
+
+
+if __name__ == "__main__":
+    results = ar1_model([0., 1.], 10, 3)
+
+    print(results)
+    print(results.shape)