source: branches/GBT/HeuristicLab.Algorithms.DataAnalysis/3.4/GradientBoostedTrees/LossFunctions/AbsoluteErrorLoss.cs @ 12374

using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using HeuristicLab.Common;
using HeuristicLab.Core;

namespace GradientBoostedTrees {
  public class AbsoluteErrorLoss : ILossFunction {
    public double GetLoss(IEnumerable<double> target, IEnumerable<double> pred, IEnumerable<double> weight) {
      var targetEnum = target.GetEnumerator();
      var predEnum = pred.GetEnumerator();
      var weightEnum = weight.GetEnumerator();

      double s = 0;
      while (targetEnum.MoveNext() & predEnum.MoveNext() & weightEnum.MoveNext()) {
        double res = targetEnum.Current - predEnum.Current;
        s += weightEnum.Current * Math.Abs(res);
      }
      if (targetEnum.MoveNext() | predEnum.MoveNext() | weightEnum.MoveNext())
        throw new ArgumentException("target, pred and weight have differing lengths");

      return s;
    }

    public IEnumerable<double> GetLossGradient(IEnumerable<double> target, IEnumerable<double> pred, IEnumerable<double> weight) {
      var targetEnum = target.GetEnumerator();
      var predEnum = pred.GetEnumerator();
      var weightEnum = weight.GetEnumerator();

      while (targetEnum.MoveNext() & predEnum.MoveNext() & weightEnum.MoveNext()) {
        var res = targetEnum.Current - predEnum.Current;
        if (res > 0) yield return weightEnum.Current;
        else if (res < 0) yield return -weightEnum.Current;
        else yield return 0.0;
      }
      if (targetEnum.MoveNext() | predEnum.MoveNext() | weightEnum.MoveNext())
        throw new ArgumentException("target, pred and weight have differing lengths");
    }

    public LineSearchFunc GetLineSearchFunc(IEnumerable<double> target, IEnumerable<double> pred, IEnumerable<double> weight) {
      var targetArr = target.ToArray();
      var predArr = pred.ToArray();
      var weightArr = weight.ToArray();
      // weights are not supported yet
      // when weights are supported we need to calculate a weighted median
      Debug.Assert(weightArr.All(w => w.IsAlmost(1.0)));

      if (targetArr.Length != predArr.Length || predArr.Length != weightArr.Length)
        throw new ArgumentException("target, pred and weight have differing lengths");

      // line search for abs error
      LineSearchFunc lineSearch = (idx, startIdx, endIdx) => {
        int nRows = endIdx - startIdx + 1;
        var res = from offset in Enumerable.Range(0, nRows)
                  let i = startIdx + offset
                  let row = idx[i]
                  select (targetArr[row] - predArr[row]);
        return res.Median();

        // old code for weighted median calculation
        // int nRows = endIdx - startIdx + 1; // startIdx and endIdx are inclusive
        // if (nRows == 1) return targetArr[idx[startIdx]] - predArr[idx[startIdx]];
        // else if (nRows == 2) {
        //   if (weightArr[idx[startIdx]] > weightArr[idx[endIdx]]) {
        //     return targetArr[idx[startIdx]] - predArr[idx[startIdx]];
        //   } else if (weightArr[idx[startIdx]] < weightArr[idx[endIdx]]) {
        //     return targetArr[idx[endIdx]] - predArr[idx[endIdx]];
        //   } else {
        //     // same weight
        //     return ((targetArr[idx[startIdx]] - predArr[idx[startIdx]]) +
        //        (targetArr[idx[endIdx]] - predArr[idx[endIdx]])) / 2;
        //   }
        // } else {
        //   var ts = from offset in Enumerable.Range(0, nRows)
        //            let i = startIdx + offset
        //            select new { res = targetArr[idx[i]] - predArr[idx[i]], weight = weightArr[idx[i]] };
        //   ts = ts.OrderBy(t => t.res);
        //   var totalWeight = ts.Sum(t => t.weight);
        //   var tsEnumerator = ts.GetEnumerator();
        //   tsEnumerator.MoveNext();
        // 
        //   double aggWeight = tsEnumerator.Current.weight; // weight of first
        // 
        //   while (aggWeight < totalWeight / 2) {
        //     tsEnumerator.MoveNext();
        //     aggWeight += tsEnumerator.Current.weight;
        //   }
        //   return tsEnumerator.Current.res;
        // }
      };
      return lineSearch;

    }

    public override string ToString() {
      return "Absolute error loss";
    }
  }
}