update needs to know the context for some learners

src/bandit/bandit.cpp

@@ -92,8 +92,8 @@ T Bandit<T>::choose(tree<Node>& tree, Fitness& f) {
 }
 
 template <typename T>
-void Bandit<T>::update(T arm, float reward) {
-    this->pbandit->update(arm, reward);
+void Bandit<T>::update(T arm, float reward, tree<Node>* tree, Fitness* f) {
+    this->pbandit->update(arm, reward, tree, f);
 }
 
 } // MAB

src/bandit/bandit.h

@@ -113,7 +113,7 @@ struct Bandit
      * @param arm The chosen arm.
      * @param reward The received reward.
      */
-    void update(T arm, float reward);
+    void update(T arm, float reward, tree<Node>* tree=nullptr, Fitness* f=nullptr);
 };
 
 //TODO: serialization should save the type of bandit and its parameters

src/bandit/bandit_operator.cpp

@@ -51,7 +51,7 @@ T BanditOperator<T>::choose(tree<Node>& tree, Fitness& f)
 
 
 template<typename T>
-void BanditOperator<T>::update(T arm, float reward)
+void BanditOperator<T>::update(T arm, float reward, tree<Node>* tree, Fitness* f)
 {
     // TODO: Implement the logic for updating the bandit operator's internal state
     // based on the received rewards

src/bandit/bandit_operator.h

@@ -62,7 +62,7 @@ class BanditOperator
      * @param arm The arm for which to update the reward.
      * @param reward The reward value.
      */
-    virtual void update(T arm, float reward);
+    virtual void update(T arm, float reward, tree<Node>* tree=nullptr, Fitness* f=nullptr); // TODO: this should not have a default value in the future
 protected:    
     std::map<T, float> probabilities;
 };

src/bandit/dummy.cpp

@@ -16,7 +16,8 @@ T DummyBandit<T>::choose(tree<Node>& tree, Fitness& f) {
 }
 
 template <typename T>
-void DummyBandit<T>::update(T arm, float reward) {
+void DummyBandit<T>::update(T arm, float reward, tree<Node>* tree, Fitness* f) {
+    // Do nothing
 }
 
 } // MAB

src/bandit/dummy.h

@@ -20,7 +20,7 @@ class DummyBandit : public BanditOperator<T>
 
     std::map<T, float> sample_probs(bool update);
     T choose(tree<Node>& tree, Fitness& f);
-    void update(T arm, float reward);
+    void update(T arm, float reward, tree<Node>* tree=nullptr, Fitness* f=nullptr);
 
 private:
     // additional stuff should come here

src/bandit/linear_thompson.cpp

@@ -31,7 +31,7 @@ T LinearThompsonSamplingBandit<T>::choose(tree<Node>& tree, Fitness& f) {
 }
 
 template <typename T>
-void LinearThompsonSamplingBandit<T>::update(T arm, float reward) {
+void LinearThompsonSamplingBandit<T>::update(T arm, float reward, tree<Node>* tree, Fitness* f) {
 
 }
 

src/bandit/linear_thompson.h

@@ -26,7 +26,7 @@ class LinearThompsonSamplingBandit : public BanditOperator<T>
 
     std::map<T, float> sample_probs(bool update);
     T choose(tree<Node>& tree, Fitness& f);
-    void update(T arm, float reward);
+    void update(T arm, float reward, tree<Node>* tree=nullptr, Fitness* f=nullptr);
 private:
 };
 

src/bandit/thompson.cpp

@@ -85,7 +85,7 @@ T ThompsonSamplingBandit<T>::choose(tree<Node>& tree, Fitness& f) {
 }
 
 template <typename T>
-void ThompsonSamplingBandit<T>::update(T arm, float reward) {
+void ThompsonSamplingBandit<T>::update(T arm, float reward, tree<Node>* tree, Fitness* f) {
     // reward must be either 0 or 1
 
     alphas[arm] += reward;

src/bandit/thompson.h

@@ -26,7 +26,7 @@ class ThompsonSamplingBandit : public BanditOperator<T>
 
     std::map<T, float> sample_probs(bool update);
     T choose(tree<Node>& tree, Fitness& f);
-    void update(T arm, float reward);
+    void update(T arm, float reward, tree<Node>* tree=nullptr, Fitness* f=nullptr);
 private:
     bool dynamic_update;
     float C = 1000;