Update of metrics.

src/Metrics/metrics.jl

@@ -1,4 +1,81 @@
-export calc_metrics, get_metrics
+export calc_metrics, get_metrics, mae, rmse, rsq, mean_cos
+
+"""
+    mae(x_pred, x)
+    
+`x_pred`: vector of predicted values. E.g. predicted energies.
+`x`: vector of true values. E.g. DFT energies.
+
+Returns mean absolute error.
+"""
+function mae(x_pred, x)
+    return sum(abs.(x_pred .- x)) / length(x)
+end
+
+"""
+    rmse(x_pred, x)
+    
+`x_pred`: vector of predicted values. E.g. predicted energies.
+`x`: vector of true values. E.g. DFT energies.
+
+Returns mean root mean square error.
+"""
+function rmse(x_pred, x)
+    return sqrt(sum((x_pred .- x) .^ 2) / length(x))
+end
+
+"""
+    rsq(x_pred, x)
+    
+`x_pred`: vector of predicted values. E.g. predicted energies.
+`x`: vector of true values. E.g. DFT energies.
+
+Returns R-squared.
+"""
+function rsq(x_pred, x)
+    return 1 - sum((x_pred .- x) .^ 2) / sum((x .- mean(x)) .^ 2)
+end
+
+"""
+    mean_cos(x_pred, x)
+    
+`x_pred`: vector of predicted forces,
+`x`: vector of true forces.
+
+Returns mean cosine.
+"""
+function mean_cos(x_pred, x)
+    x_pred_v = collect(eachcol(reshape(x_pred, 3, :)))
+    x_v = collect(eachcol(reshape(x, 3, :)))
+    x_cos = dot.(x_v, x_pred_v) ./ (norm.(x_v) .* norm.(x_pred_v))
+    x_mean_cos = mean(filter(!isnan, x_cos))
+    return x_mean_cos
+end
+
+"""
+    get_metrics(
+        x_pred,
+        x;
+        metrics = [mae, rmse, rsq],
+        label = "x"
+    )
+    
+`x_pred`: vector of predicted forces,
+`x`: vector of true forces.
+`metrics`: vector of metrics.
+`label`: label used as prefix in dictionary keys.
+
+Returns and OrderedDict with different metrics.
+"""
+function get_metrics(
+    x_pred,
+    x;
+    metrics = [mae, rmse, rsq],
+    label = "x"
+)
+    return OrderedDict( "$(label)_$(Symbol(m))" => m(x_pred, x)
+                         for m in metrics)
+end
 
 
 """
@@ -17,6 +94,7 @@ function calc_metrics(x_pred, x)
     return x_mae, x_rmse, x_rsq
 end
 
+
 """
     get_metrics( e_train_pred, e_train, e_test_pred, e_test)