stan-dev/math

Reported by @sims1253 for bernoulli_logit_lpmf here and bernoulli_logit_glm_lpmf here. Here are some tests that would fail if inserted into math/test/unit/math/mix/prob/bernoulli_logit_glm_lpmf_test.cpp Line 4 in 344d716 TEST_F(AgradRev, mathMixScalFun_bernoulli_logit_glm_lpmf) { // internal cutoff for extreme values const double cutoff = 20; Eigen::MatrixXd x_one(1, 1); x_one << 1.0; Eigen::Matrix<double, Eigen::Dynamic, 1> beta_one(1, 1); beta_one << (cutoff + 5); stan::test::ad_tolerances tols; tols.gradient_grad_ = 1e-6; stan::test::expect_ad(tols, f(0), x_one, 0.0, beta_one); stan::test::expect_ad(tols, f(1), x_one, 0.0, beta_one); stan::test::expect_ad(tols, f(2), x_one, 0.0, beta_one); Eigen::Matrix<double, Eigen::Dynamic, 1> beta_flipped = beta * -1; stan::test::expect_ad(tols, f(0), x_one, 0.0, beta_flipped); stan::test::expect_ad(to

For example, in prim/fun/square.hpp. Note that some care must be taken to avoid changing the rounding behavior by casting the input before multiplying With default toolchain settings (gcc/clang default to -fmath-errno), std::pow must set errno on domain/range/overflow errors, so the optimizer cannot rewrite it to a multiply in general. glibc's pow is a branchy multi-path implementation where x * x is one instruction. Originally posted by @sims1253 in #3368

The code was generated by chatgpt sol. The math patch code is attached. The algorithm is from https://dl.acm.org/doi/10.1145/3423597. @WardBrian this is the solver we discussed about. functions { // Replace this body to prototype a different scalar equation f(x, ...)=0. // This example solves x^3 - theta[1] = 0. real itp_residual(real x, vector theta, array[] real x_r, array[] int x_i) { return x * x * x - theta[1]; } int itp_n_half(real width, real x_tolerance, int max_num_steps) { int n_half = 0; real bisection_width = 2.0 * x_tolerance; while (bisection_width < width) { if (n_half >= max_num_steps) { reject("solve_itp_prototype: max_num_steps is too small for the ", "requested bracket width and x_tolerance"); } bisection_width *= 2.0; n_half += 1; } return n_half; } // Overflow-resistant regula-falsi point for opposite-signed fa and fb. real itp_false_position(real a, real b

To facilitate working with the sum-to-zero constraint in multivariate contexts, it would be valuable to be able to specify not only individual sum_to_zero_vectors but also matrices for which all rows or all columns are constrained to sum to zero. To my knowledge, there is no universally standard name for such a mathematical object (nor to Claude's knowledge, which only produced a list of suggestions), but for the sake of starting a conversation, I would propose something like row_centered_matrix and col_centered_matrix (and perhaps row_centred_matrix and col_centred_matrix as aliases for spelling-dialect neutrality). The primary use case for me would be hierarchical modelling with correlated multivariate random effects, e.g., random-slopes models or multivariate item-response models. To reduce problems with funnel geometries during sampling, it can be wise to write such models using ‘raw’ parameters with standard normal priors and then construct the actual parameters in the transfo

I would rather write is_odd(n) than n % 2 == 1 or more obscure n % 2.. Same for is_even(n) vs. n % 2 == 0 or more obscurely !(n % 2). This is about the simplest pair of functions to add because they don't need overloads for different scalar types and don't have derivatives. It would be great if they could be vectorized over arrays of any shape to return an array of the same shape with 0/1 values based on whether each element is odd or not. This should then be added to the Stan language and doc.