SHAP values for feature contributions (#2438)

* SHAP values for feature contributions

* Fix commenting error

* New polynomial time SHAP value estimation algorithm

* Update API to support SHAP values

* Fix merge conflicts with updates in master

* Correct submodule hashes

* Fix variable sized stack allocation

* Make lint happy

* Add docs

* Fix typo

* Adjust tolerances

* Remove unneeded def

* Fixed cpp test setup

* Updated R API and cleaned up

* Fixed test typo

src/c_api/c_api.cc

@@ -758,7 +758,8 @@ XGB_DLL int XGBoosterPredict(BoosterHandle handle,
       (option_mask & 1) != 0,
       &preds, ntree_limit,
       (option_mask & 2) != 0,
-      (option_mask & 4) != 0);
+      (option_mask & 4) != 0,
+      (option_mask & 8) != 0);
   *out_result = dmlc::BeginPtr(preds);
   *len = static_cast<xgboost::bst_ulong>(preds.size());
   API_END();

src/gbm/gblinear.cc

@@ -224,7 +224,7 @@ class GBLinear : public GradientBooster {
 
   void PredictContribution(DMatrix* p_fmat,
                            std::vector<bst_float>* out_contribs,
-                           unsigned ntree_limit) override {
+                           unsigned ntree_limit, bool approximate) override {
     if (model.weight.size() == 0) {
       model.InitModel();
     }

src/gbm/gbtree.cc

@@ -233,8 +233,8 @@ class GBTree : public GradientBooster {
 
   void PredictContribution(DMatrix* p_fmat,
                            std::vector<bst_float>* out_contribs,
-                           unsigned ntree_limit) override {
-    predictor->PredictContribution(p_fmat, out_contribs, model_, ntree_limit);
+                           unsigned ntree_limit, bool approximate) override {
+    predictor->PredictContribution(p_fmat, out_contribs, model_, ntree_limit, approximate);
   }
 
   std::vector<std::string> DumpModel(const FeatureMap& fmap,

src/learner.cc

@@ -433,9 +433,9 @@ class LearnerImpl : public Learner {
 
   void Predict(DMatrix* data, bool output_margin,
                std::vector<bst_float>* out_preds, unsigned ntree_limit,
-               bool pred_leaf, bool pred_contribs) const override {
+               bool pred_leaf, bool pred_contribs, bool approx_contribs) const override {
     if (pred_contribs) {
-      gbm_->PredictContribution(data, out_preds, ntree_limit);
+      gbm_->PredictContribution(data, out_preds, ntree_limit, approx_contribs);
     } else if (pred_leaf) {
       gbm_->PredictLeaf(data, out_preds, ntree_limit);
     } else {

src/predictor/cpu_predictor.cc

@@ -206,9 +206,9 @@ class CPUPredictor : public Predictor {
     }
   }
 
-  void PredictContribution(DMatrix* p_fmat,
-                           std::vector<bst_float>* out_contribs,
-                           const gbm::GBTreeModel& model, unsigned ntree_limit) override {
+  void PredictContribution(DMatrix* p_fmat, std::vector<bst_float>* out_contribs,
+                           const gbm::GBTreeModel& model, unsigned ntree_limit,
+                           bool approximate) override {
     const int nthread = omp_get_max_threads();
     InitThreadTemp(nthread,  model.param.num_feature);
     const MetaInfo& info = p_fmat->info();
@@ -225,10 +225,12 @@ class CPUPredictor : public Predictor {
     // make sure contributions is zeroed, we could be reusing a previously
     // allocated one
     std::fill(contribs.begin(), contribs.end(), 0);
-// initialize tree node mean values
-#pragma omp parallel for schedule(static)
-    for (bst_omp_uint i = 0; i < ntree_limit; ++i) {
-      model.trees[i]->FillNodeMeanValues();
+    if (approximate) {
+      // initialize tree node mean values
+      #pragma omp parallel for schedule(static)
+      for (bst_omp_uint i = 0; i < ntree_limit; ++i) {
+        model.trees[i]->FillNodeMeanValues();
+      }
     }
     // start collecting the contributions
     dmlc::DataIter<RowBatch>* iter = p_fmat->RowIterator();
@@ -253,7 +255,11 @@ class CPUPredictor : public Predictor {
             if (model.tree_info[j] != gid) {
               continue;
             }
-            model.trees[j]->CalculateContributions(feats, root_id, p_contribs);
+            if (!approximate) {
+              model.trees[j]->CalculateContributions(feats, root_id, p_contribs);
+            } else {
+              model.trees[j]->CalculateContributionsApprox(feats, root_id, p_contribs);
+            }
           }
           feats.Drop(batch[i]);
           // add base margin to BIAS

src/predictor/gpu_predictor.cu

@@ -384,9 +384,10 @@ class GPUPredictor : public xgboost::Predictor {
   void PredictContribution(DMatrix* p_fmat,
                            std::vector<bst_float>* out_contribs,
                            const gbm::GBTreeModel& model,
-                           unsigned ntree_limit) override {
+                           unsigned ntree_limit,
+                           bool approximate) override {
     cpu_predictor->PredictContribution(p_fmat, out_contribs, model,
-                                       ntree_limit);
+                                       ntree_limit, approximate);
   }
 
   void Init(const std::vector<std::pair<std::string, std::string>>& cfg,