unity-tech-cn
/
ml-agents
46
14
派生 39
代码 合并请求 版本发布
浏览代码

fix bug in RandomNormal (#2294) 

* fix bug in RandomNormal, add test for distribution

* extract epsilon, rename vars
/develop-generalizationTraining-TrainerController
GitHub 6 年前
当前提交
b11efed9
共有 2 个文件被更改，包括 145 次插入 和 102 次删除
  1. 235
      UnitySDK/Assets/ML-Agents/Editor/Tests/RandomNormalTest.cs
  2. 12
      UnitySDK/Assets/ML-Agents/Scripts/InferenceBrain/Utils/RandomNormal.cs

235
UnitySDK/Assets/ML-Agents/Editor/Tests/RandomNormalTest.cs
查看文件


using System;
using System;
using UnityEngine;
using System.Collections;
public class RandomNormalTest
{
public class RandomNormalTest
{
private const float firstValue = -1.19580f;
private const float secondValue = -0.97345f;
private const double epsilon = 0.0001;
[Test]
public void RandomNormalTestTwoDouble()
{
RandomNormal rn = new RandomNormal(2018);
[Test]
public void RandomNormalTestTwoDouble ()
{
RandomNormal rn = new RandomNormal (2018);
Assert.AreEqual(-0.46666, rn.NextDouble(), 0.0001);
Assert.AreEqual(-0.37989, rn.NextDouble(), 0.0001);
}
Assert.AreEqual (firstValue, rn.NextDouble (), epsilon);
Assert.AreEqual (secondValue, rn.NextDouble (), epsilon);
}
[Test]
public void RandomNormalTestWithMean()
{
RandomNormal rn = new RandomNormal(2018, 5.0f);
[Test]
public void RandomNormalTestWithMean ()
{
RandomNormal rn = new RandomNormal (2018, 5.0f);
Assert.AreEqual(4.53333, rn.NextDouble(), 0.0001);
Assert.AreEqual(4.6201, rn.NextDouble(), 0.0001);
}
Assert.AreEqual (firstValue + 5.0, rn.NextDouble (), epsilon);
Assert.AreEqual (secondValue + 5.0, rn.NextDouble (), epsilon);
}
[Test]
public void RandomNormalTestWithStddev()
{
RandomNormal rn = new RandomNormal(2018, 1.0f, 4.2f);
[Test]
public void RandomNormalTestWithStddev ()
{
RandomNormal rn = new RandomNormal (2018, 0.0f, 4.2f);
Assert.AreEqual(-0.9599, rn.NextDouble(), 0.0001);
Assert.AreEqual(-0.5955, rn.NextDouble(), 0.0001);
}
Assert.AreEqual (firstValue * 4.2, rn.NextDouble (), epsilon);
Assert.AreEqual (secondValue * 4.2, rn.NextDouble (), epsilon);
}
[Test]
public void RandomNormalTestWithMeanStddev()
{
RandomNormal rn = new RandomNormal(2018, -3.2f, 2.2f);
[Test]
public void RandomNormalTestWithMeanStddev ()
{
float mean = -3.2f;
float stddev = 2.2f;
RandomNormal rn = new RandomNormal (2018, mean, stddev);
Assert.AreEqual (firstValue * stddev + mean, rn.NextDouble (), epsilon);
Assert.AreEqual (secondValue * stddev + mean, rn.NextDouble (), epsilon);
}
[Test]
public void RandomNormalTestTensorInt ()
{
RandomNormal rn = new RandomNormal (1982);
Tensor t = new Tensor {
ValueType = Tensor.TensorType.Integer
};
Assert.Throws<NotImplementedException> (() => rn.FillTensor (t));
}
[Test]
public void RandomNormalTestDataNull ()
{
RandomNormal rn = new RandomNormal (1982);
Tensor t = new Tensor {
ValueType = Tensor.TensorType.FloatingPoint
};
Assert.Throws<ArgumentNullException> (() => rn.FillTensor (t));
}
[Test]
public void RandomNormalTestDistribution ()
{
float mean = -3.2f;
float stddev = 2.2f;
RandomNormal rn = new RandomNormal (2018, mean, stddev);
int numSamples = 100000;
// Adapted from https://www.johndcook.com/blog/standard_deviation/
// Computes stddev and mean without losing precision
double oldM = 0.0, newM = 0.0, oldS = 0.0, newS = 0.0;
Assert.AreEqual(-4.2266, rn.NextDouble(), 0.0001);
Assert.AreEqual(-4.0357, rn.NextDouble(), 0.0001);
}
for (int i = 0; i < numSamples; i++) {
double x = rn.NextDouble ();
if (i == 0) {
oldM = newM = x;
oldS = 0.0;
} else {
newM = oldM + (x - oldM) / i;
newS = oldS + (x - oldM) * (x - newM);
[Test]
public void RandomNormalTestTensorInt()
{
RandomNormal rn = new RandomNormal(1982);
Tensor t = new Tensor
{
ValueType = Tensor.TensorType.Integer
};
// set up for next iteration
oldM = newM;
oldS = newS;
}
}
Assert.Throws<NotImplementedException>(() => rn.FillTensor(t));
}
double sampleMean = newM;
double sampleVariance = newS / (numSamples - 1);
double sampleStddev = Math.Sqrt (sampleVariance);
[Test]
public void RandomNormalTestDataNull()
{
RandomNormal rn = new RandomNormal(1982);
Tensor t = new Tensor
{
ValueType = Tensor.TensorType.FloatingPoint
};
// Note a larger epsilon here. We could get closer to the true values with more samples.
Assert.AreEqual (mean, sampleMean, 0.01);
Assert.AreEqual (stddev, sampleStddev, 0.01);
Assert.Throws<ArgumentNullException>(() => rn.FillTensor(t));
}
}
[Test]
public void RandomNormalTestTensor()
{
RandomNormal rn = new RandomNormal(1982);
Tensor t = new Tensor
{
ValueType = Tensor.TensorType.FloatingPoint,
Data = Array.CreateInstance(typeof(float), new long[3] {3, 4, 2})
};
[Test]
public void RandomNormalTestTensor ()
{
RandomNormal rn = new RandomNormal (1982);
Tensor t = new Tensor {
ValueType = Tensor.TensorType.FloatingPoint,
Data = Array.CreateInstance (typeof (float), new long [3] { 3, 4, 2 })
};
rn.FillTensor(t);
rn.FillTensor (t);
float[] reference = new float[]
{
-0.2139822f,
0.5051259f,
-0.5640336f,
-0.3357787f,
-0.2055894f,
-0.09432302f,
-0.01419199f,
0.53621f,
-0.5507085f,
-0.2651141f,
0.09315512f,
-0.04918706f,
-0.179625f,
0.2280539f,
0.1883962f,
0.4047216f,
0.1704049f,
0.5050544f,
-0.3365685f,
0.3542781f,
0.5951571f,
0.03460682f,
-0.5537263f,
-0.4378373f,
};
float [] reference = new float []
{
-0.4315872f,
0.9561074f,
-1.130287f,
-0.7763879f,
-0.3027347f,
-0.1377991f,
-0.02921959f,
0.9520947f,
-1.11074f,
-0.5018106f,
0.1413168f,
-0.07491868f,
-0.2645015f,
0.3331701f,
0.3716498f,
1.088157f,
0.3414804f,
1.167787f,
-0.5105762f,
0.5396146f,
1.225356f,
0.06144788f,
-1.092338f,
-1.177194f,
};
int i = 0;
foreach (float f in t.Data)
{
Assert.AreEqual(f, reference[i], 0.0001);
++i;
}
int i = 0;
foreach (float f in t.Data) {
Assert.AreEqual (f, reference [i], epsilon);
++i;
}
}
}
}
}
}

12
UnitySDK/Assets/ML-Agents/Scripts/InferenceBrain/Utils/RandomNormal.cs
查看文件


using System;
using System;
namespace MLAgents.InferenceBrain.Utils
{

private readonly double m_mean;
private readonly double m_stddev;
private readonly System.Random m_random;
public RandomNormal(int seed, float mean = 0.0f, float stddev = 1.0f)
{
m_mean = mean;

// Each iteration produces two numbers. Hold one here for next call
private bool m_hasSpare = false;
private double m_spare = 0.0f;
/// <summary>
/// Return the next random double number
/// </summary>

s = u * u + v * v;
} while (s >= 1.0 || s == 0.0);
s = Math.Sqrt(-2.0 * Math.Log(s) / 2);
s = Math.Sqrt(-2.0 * Math.Log(s) / s);
m_spare = u * s;
m_hasSpare = true;

}
}
}
/// Fill a pre-allocated Tensor with random numbers
/// Fill a pre-allocated Tensor with random numbers
/// </summary>
/// <param name="t">The pre-allocated Tensor to fill</param>
/// <exception cref="NotImplementedException">Throws when trying to fill a Tensor of type other than float</exception>

撰写 预览
正在加载...
取消
保存