vectorized densities evaluate OK -- need to check results

Author: jonathan-taylor

selectiveInference/R/funs.randomized.R

@@ -124,7 +124,7 @@ randomizedLASSO = function(X,
                 inactive_set = inactive_set,
                 unpenalized_set = unpenalized_set,
                 sign_soln = sign_soln,
-                opt_transform = opt_transform,
+                optimization_transform = opt_transform,
                 internal_transform = internal_transform
                 ))
 

selectiveInference/src/Rcpp-randomized.cpp

@@ -7,7 +7,7 @@ Rcpp::NumericVector log_density_gaussian_(double noise_scale,
 					  Rcpp::NumericMatrix internal_state,         // D -- data state -- matrix of shape (nopt, npts)
 					  Rcpp::NumericMatrix optimization_linear,    // A_O -- linear part for optimization variables
 					  Rcpp::NumericMatrix optimization_state,     // O -- optimization state -- matrix of shape (ninternal, npts)
-					  Rcpp::NumericMatrix offset) {               // h -- offset in affine transform -- "p" dimensional 
+					  Rcpp::NumericVector offset) {               // h -- offset in affine transform -- "p" dimensional 
 
   int npt = internal_state.ncol();         // Function is vectorized
   if (optimization_state.ncol() != npt) {  // Assuming each column is an internal or opt state because arrays are column major
@@ -43,7 +43,7 @@ Rcpp::NumericVector log_density_gaussian_(double noise_scale,
 Rcpp::NumericVector log_density_gaussian_conditional_(double noise_scale,                         // Scale of randomization
 						      Rcpp::NumericMatrix optimization_linear,    // A_O -- linear part for optimization variables
 						      Rcpp::NumericMatrix optimization_state,     // O -- optimization state -- matrix of shape (ninternal, npts)
-						      Rcpp::NumericMatrix offset) {               // h -- offset in affine transform -- "p" dimensional 
+						      Rcpp::NumericVector offset) {               // h -- offset in affine transform -- "p" dimensional 
 
   int npt = optimization_state.ncol();         // Function is vectorized
   int ndim = optimization_linear.nrow();  

tests/randomized/test_randomized.R

@@ -1,6 +1,6 @@
 library(selectiveInference)
 
-test = function() {
+smoke_test = function() {
 
     n = 100; p = 50
     X = matrix(rnorm(n * p), n, p)
@@ -10,6 +10,27 @@ test = function() {
     ridge_term = .1 / sqrt(n)
     selectiveInference:::randomizedLASSO(X, y, lam, noise_scale, ridge_term)
 }
+A = smoke_test()
 
-A=test()
-#print(test())
+density_test = function() {
+
+    random_lasso = smoke_test()
+    p = nrow(random_lasso$internal_transform$linear_term)
+    internal_state = matrix(rnorm(p * 20), p, 20)
+    optimization_state = matrix(rnorm(p * 20), p, 20)
+    offset = rnorm(p)
+
+    selectiveInference:::log_density_gaussian_(10.,
+                                               random_lasso$internal_transform$linear_term,
+                                               internal_state,
+                                               random_lasso$optimization_transform$linear_term,
+					       optimization_state,
+                                               offset)
+
+    selectiveInference:::log_density_gaussian_conditional_(10.,
+                                                           random_lasso$optimization_transform$linear_term,
+                    				           optimization_state,
+                                                           offset)
+}
+
+density_test()