Merge Release/v4.7.0 into develop

.github/ISSUE_TEMPLATE.md

@@ -25,4 +25,4 @@ If this is a **feature request**, show what you expect to happen if the feature
 
 
 #### Current Version:
-v4.6.2
+v4.7.0

RELEASE-NOTES.txt

@@ -1,5 +1,24 @@
 Stan Math Library Release Notes
 
+======================================================================
+v4.7.0 (5 September 2023)
+======================================================================
+
+- Improved accuracy of `matrix_exp_multiply`. (#2619)
+- Added framework for vectorising ternary functions (#2642)
+- Added `select()` function for vectorised ternary operations, as well as the `any()` and `all()` boolean reduction functions (#2853)
+- Added a minimum number of iterations (5) to the `grad_2f1` function to avoid early convergence (#2858)
+- Fixed some spelling errors in documentation (#2908)
+- Added new function in `stan::math::internal`, `finite_diff_hessian_vector_product_auto` (#2914)
+- Fixed an issue with the return type of `eigenvalues` being incorrect (#2915)
+- Delete jquery files from vendored Boost documentation (#2916)
+- Fixed fvar/higher-order autodiff compatibility with GLM distributions (#2917)
+- Fixed a type issue in the tests for inv_wishart_cholesky. (#2920)
+- The constraint tolerance can now be adjusted by defining the `STAN_MATH_CONSTRAINT_TOLERANCE` macro before including Stan headers. The default value remains `1E-8`. (#2921)
+- Use -dumpfullversion if available in makefiles (#2922)
+- Vectorised binary signature for `log_sum_exp` (#2930)
+- Added new functions `qr_thin`, `eigendecompose_sym`, `eigendecompose`, `complex_schur_decompose`, `svd`, and `csr_extract`. Each of these is equivalent to calling several existing functions, but should be more efficient due to sharing work. For example, `svd(m)` is equivalent to `(svd_U(m), singular_values(m), svd_V(m))`. (#2931)
+
 ======================================================================
 v4.6.2 (15 May 2023)
 ======================================================================

doxygen/doxygen.cfg

@@ -48,7 +48,7 @@ PROJECT_NAME           = "Stan Math Library"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = 4.6.2
+PROJECT_NUMBER         = 4.7.0
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

stan/math/prim/functor/operands_and_partials.hpp

@@ -54,8 +54,7 @@ class ops_partials_edge;
  *  for this specialization must be a `Arithmetic`
  */
 template <typename ViewElt, typename Op>
-class ops_partials_edge<ViewElt, Op, require_st_arithmetic<Op>> {
- public:
+struct ops_partials_edge<ViewElt, Op, require_st_arithmetic<Op>> {
   using inner_op = std::conditional_t<is_eigen<value_type_t<Op>>::value,
                                       value_type_t<Op>, Op>;
   using partials_t = empty_broadcast_array<ViewElt, inner_op>;

stan/math/prim/prob/categorical_logit_glm_lpmf.hpp

@@ -29,8 +29,8 @@ namespace math {
  * @tparam T_alpha type of the intercept vector
  * @tparam T_beta type of the matrix of weights
  * @param y a scalar or vector of classes. If it is a scalar it will be
- * broadcast - used for all instances. Values should be between 1 and number of
- * classes, including endpoints.
+ * broadcast - used for all instances. Values should be between 1 and number
+ * of classes, including endpoints.
  * @param x design matrix or row vector. If it is a row vector it will be
  * broadcast - used for all instances.
  * @param alpha intercept vector (in log odds)
@@ -148,8 +148,8 @@ return_type_t<T_x, T_alpha, T_beta> categorical_logit_glm_lpmf(
           = beta_y
             - (exp_lin.matrix() * beta_val.transpose()).array().colwise()
                   * inv_sum_exp_lin;
-      // TODO(Tadej) maybe we can replace previous block with the following line
-      // when we have newer Eigen  partials<0>(ops_partials) = beta_val(y
+      // TODO(Tadej) maybe we can replace previous block with the following
+      // line when we have newer Eigen  partials<0>(ops_partials) = beta_val(y
       // - 1, all) - (exp_lin.matrix() * beta.transpose()).colwise() *
       // inv_sum_exp_lin;
     }
@@ -183,8 +183,9 @@ return_type_t<T_x, T_alpha, T_beta> categorical_logit_glm_lpmf(
           beta_derivative.col(y_seq[i] - 1) += x_val.row(i);
         }
       }
-      // TODO(Tadej) maybe we can replace previous loop with the following line
-      // when we have newer Eigen  partials<2>(ops_partials)(Eigen::all, y
+      // TODO(Tadej) maybe we can replace previous loop with the following
+      // line when we have newer Eigen  partials<2>(ops_partials)(Eigen::all,
+      // y
       // - 1) += x_val.colwise.sum().transpose();
 
       partials<2>(ops_partials) = std::move(beta_derivative);

stan/math/prim/prob/neg_binomial_2_log_glm_lpmf.hpp

@@ -47,8 +47,8 @@ namespace math {
  * this can be a vector (of the same length as y, for heteroskedasticity)
  * or a scalar.
  *
- * @param y failures count scalar or vector parameter. If it is a scalar it will
- * be broadcast - used for all instances.
+ * @param y failures count scalar or vector parameter. If it is a scalar it
+ * will be broadcast - used for all instances.
  * @param x design matrix or row vector. If it is a row vector it will be
  * broadcast - used for all instances.
  * @param alpha intercept (in log odds)

stan/math/prim/prob/ordered_logistic_glm_lpmf.hpp

@@ -20,9 +20,9 @@ namespace stan {
 namespace math {
 
 /** \ingroup multivar_dists
- * Returns the log PMF of the ordinal regression Generalized Linear Model (GLM).
- * This is equivalent to and faster than ordered_logistic_lpmf(y, x * beta,
- * cuts).
+ * Returns the log PMF of the ordinal regression Generalized Linear Model
+ * (GLM). This is equivalent to and faster than ordered_logistic_lpmf(y, x *
+ * beta, cuts).
  *
  * @tparam T_y type of integer vector of classes. It can be either
  * `std::vector<int>` or `int`.
@@ -31,8 +31,8 @@ namespace math {
  * @tparam T_cuts type of the vector of cutpoints
  *
  * @param y a scalar or vector of classes. If it is a scalar it will be
- * broadcast - used for all instances. Values should be between 1 and number of
- * classes, including endpoints.
+ * broadcast - used for all instances. Values should be between 1 and number
+ * of classes, including endpoints.
  * @param x design matrix or row vector. If it is a row vector it will be
  * broadcast - used for all instances.
  * @param beta weight vector

stan/math/version.hpp

@@ -12,8 +12,8 @@
 #endif
 
 #define STAN_MATH_MAJOR 4
-#define STAN_MATH_MINOR 6
-#define STAN_MATH_PATCH 2
+#define STAN_MATH_MINOR 7
+#define STAN_MATH_PATCH 0
 
 namespace stan {
 namespace math {