use rayon to hopefully speed up linfa-logistic

algorithms/linfa-logistic/Cargo.toml

@@ -23,7 +23,7 @@ optional = true
 version = "1.0"
 
 [dependencies]
-ndarray = { version = "0.15", features = ["approx"] }
+ndarray = { version = "0.15", features = ["rayon", "approx"] }
 ndarray-stats = "0.5.0"
 num-traits = "0.2"
 argmin = { version = "0.9.0", default-features = false }

algorithms/linfa-logistic/src/lib.rs

@@ -454,9 +454,9 @@ fn log_sum_exp<F: linfa::Float, A: Data<Elem = F>>(
 /// Computes `exp(n - max) / sum(exp(n- max))`, which is a numerically stable version of softmax
 fn softmax_inplace<F: linfa::Float, A: DataMut<Elem = F>>(v: &mut ArrayBase<A, Ix1>) {
     let max = v.iter().copied().reduce(F::max).unwrap();
-    v.mapv_inplace(|n| (n - max).exp());
+    v.par_mapv_inplace(|n| (n - max).exp());
     let sum = v.sum();
-    v.mapv_inplace(|n| n / sum);
+    v.par_mapv_inplace(|n| n / sum);
 }
 
 /// Computes the logistic loss assuming the training labels $y \in {-1, 1}$
@@ -479,7 +479,7 @@ fn logistic_loss<F: Float, A: Data<Elem = F>>(
     let yz = x.dot(&params.into_shape((params.len(), 1)).unwrap()) + intercept;
     let len = yz.len();
     let mut yz = yz.into_shape(len).unwrap() * y;
-    yz.mapv_inplace(log_logistic);
+    yz.par_mapv_inplace(log_logistic);
     -yz.sum() + F::cast(0.5) * alpha * params.dot(&params)
 }
 
@@ -495,8 +495,7 @@ fn logistic_grad<F: Float, A: Data<Elem = F>>(
     let yz = x.dot(&params.into_shape((params.len(), 1)).unwrap()) + intercept;
     let len = yz.len();
     let mut yz = yz.into_shape(len).unwrap() * y;
-    yz.mapv_inplace(logistic);
-    yz -= F::one();
+    yz.par_mapv_inplace(|v| logistic(v) - F::one());
     yz *= y;
     if w.len() == n_features + 1 {
         let mut grad = Array::zeros(w.len());
@@ -624,7 +623,7 @@ impl<F: Float, C: PartialOrd + Clone> FittedLogisticRegression<F, C> {
     /// model was fitted.
     pub fn predict_probabilities<A: Data<Elem = F>>(&self, x: &ArrayBase<A, Ix2>) -> Array1<F> {
         let mut probs = x.dot(&self.params) + self.intercept;
-        probs.mapv_inplace(logistic);
+        probs.par_mapv_inplace(logistic);
         probs
     }
 }