cross-chain degenerates into classic when only single chain exists

stan-dev · Feb 18, 2020 · 6052421 · 6052421
1 parent 1174990
commit 6052421
Show file tree

Hide file tree

Showing 5 changed files with 45 additions and 32 deletions.
diff --git a/src/stan/mcmc/hmc/mpi_cross_chain_adapter.hpp b/src/stan/mcmc/hmc/mpi_cross_chain_adapter.hpp
@@ -113,9 +113,13 @@ namespace mcmc {
 
     inline void
     write_num_cross_chain_warmup(callbacks::writer& sample_writer,
-                                 int num_thin) {
-      size_t n = num_cross_chain_draws();
-      sample_writer("num_warmup = " + std::to_string(n / num_thin));
+                                 int num_thin, int num_warmup) {
+      if (use_cross_chain_adapt()) {
+        size_t n = num_cross_chain_draws();
+        sample_writer("num_warmup = " + std::to_string(n / num_thin));
+      } else {
+        sample_writer("num_warmup = " + std::to_string(num_warmup));
+      }
     }
 
     /*
@@ -454,7 +458,9 @@ namespace mcmc {
       sampler.set_nominal_stepsize(new_stepsize);
     }
 
-    inline bool use_cross_chain_adapt() { return true; }
+    inline bool use_cross_chain_adapt() {
+      return num_chains_ > 1;
+    }
   };
 
 #else  // sequential version

diff --git a/src/stan/mcmc/hmc/nuts/adapt_dense_e_nuts.hpp b/src/stan/mcmc/hmc/nuts/adapt_dense_e_nuts.hpp
@@ -31,27 +31,30 @@ class adapt_dense_e_nuts : public dense_e_nuts<Model, BaseRNG>,
       this->stepsize_adaptation_.learn_stepsize(this->nom_epsilon_,
                                                 s.accept_stat());
 
-      bool update;
       if (this -> use_cross_chain_adapt()) {
         this -> add_cross_chain_sample(s.log_prob());
-        update = this -> cross_chain_adaptation(logger);
+        bool update = this -> cross_chain_adaptation(logger);
         if (this -> is_cross_chain_adapted()) {
           update = false;
         }
-      } else {
-        update = this->covar_adaptation_.learn_covariance(this->z_.inv_e_metric_,
-                                                          this->z_.q);
-      }
 
-      if (update) {
-        this->init_stepsize(logger);
+        if (update) {
+          this->init_stepsize(logger);
 
-        this->stepsize_adaptation_.set_mu(log(10 * this->nom_epsilon_));
-        this->stepsize_adaptation_.restart();
+          this->stepsize_adaptation_.set_mu(log(10 * this->nom_epsilon_));
+          this->stepsize_adaptation_.restart();
 
-        if (this -> use_cross_chain_adapt()) {
           this->set_cross_chain_stepsize();          
         }
+      } else {
+        bool update = this->covar_adaptation_.learn_covariance(this->z_.inv_e_metric_,
+                                                          this->z_.q);
+        if (update) {
+          this->init_stepsize(logger);
+
+          this->stepsize_adaptation_.set_mu(log(10 * this->nom_epsilon_));
+          this->stepsize_adaptation_.restart();
+        }
       }
     }
     return s;

diff --git a/src/stan/mcmc/hmc/nuts/adapt_diag_e_nuts.hpp b/src/stan/mcmc/hmc/nuts/adapt_diag_e_nuts.hpp
@@ -31,27 +31,31 @@ class adapt_diag_e_nuts : public diag_e_nuts<Model, BaseRNG>,
       this->stepsize_adaptation_.learn_stepsize(this->nom_epsilon_,
                                                 s.accept_stat());
 
-      bool update;
+
       if (this -> use_cross_chain_adapt()) {
         this -> add_cross_chain_sample(s.log_prob());
-        update = this -> cross_chain_adaptation(logger);
+        bool update = this -> cross_chain_adaptation(logger);
         if (this -> is_cross_chain_adapted()) {
           update = false;
         }
-      } else {
-        update = this->var_adaptation_.learn_variance(this->z_.inv_e_metric_,
-                                                      this->z_.q);
-      }
 
-      if (update) {
-        this->init_stepsize(logger);
+        if (update) {
+          this->init_stepsize(logger);
 
-        this->stepsize_adaptation_.set_mu(log(10 * this->nom_epsilon_));
-        this->stepsize_adaptation_.restart();
+          this->stepsize_adaptation_.set_mu(log(10 * this->nom_epsilon_));
+          this->stepsize_adaptation_.restart();
 
-        if (this -> use_cross_chain_adapt()) {
           this->set_cross_chain_stepsize();          
         }
+      } else {
+        bool update = this->var_adaptation_.learn_variance(this->z_.inv_e_metric_,
+                                                           this->z_.q);
+        if (update) {
+          this->init_stepsize(logger);
+
+          this->stepsize_adaptation_.set_mu(log(10 * this->nom_epsilon_));
+          this->stepsize_adaptation_.restart();
+        }
       }
     }
     return s;

diff --git a/src/stan/services/util/mpi_cross_chain.hpp b/src/stan/services/util/mpi_cross_chain.hpp
@@ -54,7 +54,7 @@ namespace util {
 
     static void write_num_warmup(Sampler& sampler,
                                  callbacks::writer& sample_writer,
-                                 int num_thin) {}
+                                 int num_thin, int num_warmup) {}
   };
 
   /*
@@ -81,8 +81,8 @@ namespace util {
 
     static void write_num_warmup(Sampler& sampler,
                                  callbacks::writer& sample_writer,
-                                 int num_thin) {
-      sampler.write_num_cross_chain_warmup(sample_writer, num_thin);      
+                                 int num_thin, int num_warmup) {
+      sampler.write_num_cross_chain_warmup(sample_writer, num_thin, num_warmup);
     }
   };
 #endif
@@ -111,9 +111,9 @@ namespace util {
 
     static void write_num_warmup(Sampler& sampler,
                                  callbacks::writer& sample_writer,
-                                 int num_thin) {
+                                 int num_thin, int num_warmup) {
       mpi_cross_chain_impl<Sampler, has_cross_chain_warmup<Sampler>::value>::
-        write_num_warmup(sampler, sample_writer, num_thin);
+        write_num_warmup(sampler, sample_writer, num_thin, num_warmup);
     }
   };
 

diff --git a/src/stan/services/util/run_adaptive_sampler.hpp b/src/stan/services/util/run_adaptive_sampler.hpp
@@ -73,7 +73,7 @@ void run_adaptive_sampler(Sampler& sampler, Model& model,
                              mpi_cross_chain<Sampler>::num_draws(sampler),
                              num_warmup + num_samples, num_thin, refresh, save_warmup,
                              true, writer, s, model, rng, interrupt, logger);
-  mpi_cross_chain<Sampler>::write_num_warmup(sampler, sample_writer, num_thin);
+  mpi_cross_chain<Sampler>::write_num_warmup(sampler, sample_writer, num_thin, num_warmup);
 
   clock_t end = clock();
   double warm_delta_t = static_cast<double>(end - start) / CLOCKS_PER_SEC;