Merge pull request #860 from zhengruifeng/mae

Add "mean absolute error" to metrics

doc/parameter.md

@@ -80,6 +80,7 @@ Specify the learning task and the corresponding learning objective. The objectiv
   - User can add multiple evaluation metrics, for python user, remember to pass the metrics in as list of parameters pairs instead of map, so that latter 'eval_metric' won't override previous one
   - The choices are listed below:
   - "rmse": [root mean square error](http://en.wikipedia.org/wiki/Root_mean_square_error)
+  - "mae": [mean absolute error](https://en.wikipedia.org/wiki/Mean_absolute_error)
   - "logloss": negative [log-likelihood](http://en.wikipedia.org/wiki/Log-likelihood)
   - "error": Binary classification error rate. It is calculated as #(wrong cases)/#(all cases). For the predictions, the evaluation will regard the instances with prediction value larger than 0.5 as positive instances, and the others as negative instances.
   - "merror": Multiclass classification error rate. It is calculated as #(wrong cases)/#(all cases).

src/metric/elementwise_metric.cc

@@ -72,6 +72,15 @@ struct EvalRMSE : public EvalEWiseBase<EvalRMSE> {
   }
 };
 
+struct EvalMAE : public EvalEWiseBase<EvalMAE> {
+  const char *Name() const override {
+    return "mae";
+  }
+  inline static float EvalRow(float label, float pred) {
+    return std::abs(label - pred);
+  }
+};
+
 struct EvalLogLoss : public EvalEWiseBase<EvalLogLoss> {
   const char *Name() const override {
     return "logloss";
@@ -114,6 +123,10 @@ XGBOOST_REGISTER_METRIC(RMSE, "rmse")
 .describe("Rooted mean square error.")
 .set_body([](const char* param) { return new EvalRMSE(); });
 
+XGBOOST_REGISTER_METRIC(MAE, "mae")
+.describe("Mean absolute error.")
+.set_body([](const char* param) { return new EvalMAE(); });
+
 XGBOOST_REGISTER_METRIC(LogLoss, "logloss")
 .describe("Negative loglikelihood for logistic regression.")
 .set_body([](const char* param) { return new EvalLogLoss(); });