v0.2.4 release candidate

DESCRIPTION

@@ -1,11 +1,11 @@
 Package: greta
 Type: Package
 Title: Simple and Scalable Statistical Modelling in R
-Version: 0.2.3
-Date: 2018-01-23
+Version: 0.2.4
+Date: 2018-02-20
 Authors@R: person("Nick", "Golding", role = c("aut", "cre"),
     email = "[email protected]")
-Description: Write statistical models in R and fit them by MCMC on CPUs and GPUs, using Google TensorFlow (see <https://goldingn.github.io/greta> for more information).
+Description: Write statistical models in R and fit them by MCMC on CPUs and GPUs, using Google TensorFlow (see <https://greta-dev.github.io/greta> for more information).
 License: Apache License 2.0
 URL: https://github.com/greta-dev/greta
 BugReports: https://github.com/greta-dev/greta/issues

NAMESPACE

@@ -117,6 +117,7 @@ export(hypergeometric)
 export(icauchit)
 export(icloglog)
 export(ilogit)
+export(imultilogit)
 export(install_tensorflow)
 export(inverse_gamma)
 export(iprobit)

NEWS

@@ -1,3 +1,33 @@
+### greta 0.2.4
+================
+
+Fixes:
+
+* improved error checking/messages in `model()`, `%*%`
+* switched docs and examples to always use `<-` for assignment
+* fixed the `n_cores` argument to `model()`
+
+New functionality:
+
+* added a `calculate()` function to compute the values of greta arrays conditional on provided values for others
+* added `imultilogit()` transform
+* added a `chains` argument to `model()`
+* improved HMC self-tuning, including a diagonal euclidean metric
+
+### greta 0.2.3
+================
+
+Fixes:
+
+* fixed breaking change in extraDistr API (caused test errors on CRAN builds)
+* added dontrun statements to pass CRAN checks on winbuilder
+* fixed breaking change in tensorflow API (1-based indexing)
+
+New functionality:
+
+* added `cumsum()` and `cumprod()` functions
+
+
 ### greta 0.2.2
 ================
 

R/calculate.R

@@ -31,18 +31,18 @@
 #'
 #' # define a variable greta array, and another that is calculated from it
 #' # then calculate what value y would take for different values of x
-#' x = normal(0, 1, dim = 3)
-#' a = lognormal(0, 1)
+#' x <- normal(0, 1, dim = 3)
+#' a <- lognormal(0, 1)
 #' y <- sum(x ^ 2) + a
 #' calculate(y, list(x = c(0.1, 0.2, 0.3), a = 2))
 #'
 #'
 #' # define a model
-#' alpha = normal(0, 1)
-#' beta = normal(0, 1)
-#' sigma = lognormal(1, 0.1)
+#' alpha <- normal(0, 1)
+#' beta <- normal(0, 1)
+#' sigma <- lognormal(1, 0.1)
 #' mu <- alpha + iris$Petal.Length * beta
-#' distribution(iris$Petal.Width) = normal(mu, sigma)
+#' distribution(iris$Petal.Width) <- normal(mu, sigma)
 #' m <- model(alpha, beta, sigma)
 #'
 #' # calculate intermediate greta arrays, given some parameter values

R/distribution.R

@@ -23,12 +23,12 @@
 #'
 #' # observed data and mean parameter to be estimated
 #' # (explicitly coerce data to a greta array so we can refer to it later)
-#' y = as_data(rnorm(5, 0, 3))
+#' y <- as_data(rnorm(5, 0, 3))
 #'
-#' mu = uniform(-3, 3)
+#' mu <- uniform(-3, 3)
 #'
 #' # define the distribution over y (the model likelihood)
-#' distribution(y) = normal(mu, 1)
+#' distribution(y) <- normal(mu, 1)
 #'
 #' # get the distribution over y
 #' distribution(y)

R/extract_replace_combine.R

@@ -43,7 +43,7 @@
 #' @examples
 #' \dontrun{
 #'
-#'  x = as_data(matrix(1:12, 3, 4))
+#'  x <- as_data(matrix(1:12, 3, 4))
 #'
 #'  # extract/replace
 #'  x[1:3, ]

R/functions.R

@@ -86,18 +86,18 @@
 #' @examples
 #' \dontrun{
 #'
-#' x = as_data(matrix(1:9, nrow = 3, ncol = 3))
-#' a = log(exp(x))
-#' b = log1p(expm1(x))
-#' c = sign(x - 5)
-#' d = abs(x - 5)
+#' x <- as_data(matrix(1:9, nrow = 3, ncol = 3))
+#' a <- log(exp(x))
+#' b <- log1p(expm1(x))
+#' c <- sign(x - 5)
+#' d <- abs(x - 5)
 #'
-#' e = diag(x)
+#' e <- diag(x)
 #' diag(x) <- e + 1
 #'
-#' z = t(a)
+#' z <- t(a)
 #'
-#' y = sweep(x, 1, e, '-')
+#' y <- sweep(x, 1, e, '-')
 #' }
 NULL
 

R/greta_model_class.R

@@ -44,10 +44,10 @@ NULL
 #' \dontrun{
 #'
 #' # define a simple model
-#' mu = variable()
-#' sigma = lognormal(1, 0.1)
-#' x = rnorm(10)
-#' distribution(x) = normal(mu, sigma)
+#' mu <- variable()
+#' sigma <- lognormal(1, 0.1)
+#' x <- rnorm(10)
+#' distribution(x) <- normal(mu, sigma)
 #'
 #' m <- model(mu, sigma)
 #'
@@ -81,9 +81,6 @@ model <- function (...,
     }
   }
 
-  # flush all tensors from the default graph
-  tf$reset_default_graph()
-
   # nodes required
   target_greta_arrays <- list(...)
 
@@ -102,6 +99,27 @@ model <- function (...,
 
   }
 
+  # check they are greta arrays
+  are_greta_arrays <- vapply(target_greta_arrays,
+                             inherits, "greta_array",
+                             FUN.VALUE = FALSE)
+
+  if (!all(are_greta_arrays)) {
+
+    unexpected_items <- names(target_greta_arrays)[!are_greta_arrays]
+
+    msg <- ifelse(length(unexpected_items) > 1,
+                  "The following objects passed to model() are not greta arrays: ",
+                  "The following object passed to model() is not a greta array: ")
+
+    stop (msg,
+          paste(unexpected_items, sep = ", "),
+          call. = FALSE)
+
+  }
+
+
+
   if (length(target_greta_arrays) == 0) {
     stop ('could not find any non-data greta arrays',
           call. = FALSE)

R/inference.R

@@ -120,10 +120,10 @@ stash_trace <- function (trace, raw) {
 #' @examples
 #' \dontrun{
 #' # define a simple model
-#' mu = variable()
-#' sigma = lognormal(1, 0.1)
-#' x = rnorm(10)
-#' distribution(x) = normal(mu, sigma)
+#' mu <- variable()
+#' sigma <- lognormal(1, 0.1)
+#' x <- rnorm(10)
+#' distribution(x) <- normal(mu, sigma)
 #' m <- model(mu, sigma)
 #'
 #' # carry out mcmc on the model

R/operators.R

@@ -40,23 +40,23 @@
 #' @examples
 #' \dontrun{
 #'
-#' x = as_data(-1:12)
+#' x <- as_data(-1:12)
 #'
 #' # arithmetic
-#' a = x + 1
-#' b = 2 * x + 3
-#' c = x %% 2
-#' d = x %/% 5
+#' a <- x + 1
+#' b <- 2 * x + 3
+#' c <- x %% 2
+#' d <- x %/% 5
 #'
 #' # logical
-#' e = (x > 1) | (x < 1)
-#' f = e & (x < 2)
-#' g = !f
+#' e <- (x > 1) | (x < 1)
+#' f <- e & (x < 2)
+#' g <- !f
 #'
 #' # relational
-#' h = x < 1
-#' i = (-x) >= x
-#' j = h == x
+#' h <- x < 1
+#' i <- (-x) >= x
+#' j <- h == x
 #' }
 NULL
 
@@ -145,15 +145,16 @@ NULL
     y <- elem_list[[2]]
 
     # check they're matrices
-    if (length(dim(x)) != 2 | length(dim(x)) != 2)
-      stop ('only two-dimensional greta arrays can be matrix-multiplied',
+    if (length(dim(x)) != 2 | length(dim(y)) != 2) {
+      stop ("only two-dimensional greta arrays can be matrix-multiplied",
             call. = FALSE)
+    }
 
     # check the dimensions match
     if (dim(x)[2] != dim(y)[1]) {
-      msg <- sprintf('incompatible dimensions: %s vs %s',
-                     paste0(dim(x), collapse = 'x'),
-                     paste0(dim(y), collapse = 'x'))
+      msg <- sprintf("incompatible dimensions: %s vs %s",
+                     paste0(dim(x), collapse = "x"),
+                     paste0(dim(y), collapse = "x"))
       stop (msg)
     }
 

R/package.R

@@ -26,13 +26,13 @@
 #' # a simple Bayesian regression model for the iris data
 #'
 #' # priors
-#' int = normal(0, 5)
-#' coef = normal(0, 3)
-#' sd = lognormal(0, 3)
+#' int <- normal(0, 5)
+#' coef <- normal(0, 3)
+#' sd <- lognormal(0, 3)
 #'
 #' # likelihood
 #' mean <- int + coef * iris$Petal.Length
-#' distribution(iris$Sepal.Length) = normal(mean, sd)
+#' distribution(iris$Sepal.Length) <- normal(mean, sd)
 #'
 #' # build and sample
 #' m <- model(int, coef, sd)

R/probability_distributions.R

@@ -537,7 +537,7 @@ cauchy_distribution <- R6Class (
       s <- parameters$scale
 
       log_prob = function (x)
-        -tf$log(fl(pi) * s * (fl(1) + tf$square((x - loc) / s)))
+        tf$negative(tf$log(fl(pi) * s * (fl(1) + tf$square((x - loc) / s))))
 
       cdf = function (x)
         fl(1 / pi)  * tf$atan((x - loc) / s) + fl(0.5)
@@ -1359,39 +1359,39 @@ distribution_classes_module <- module(uniform_distribution,
 #' \dontrun{
 #'
 #' # a uniform parameter constrained to be between 0 and 1
-#' phi = uniform(min = 0, max = 1)
+#' phi <- uniform(min = 0, max = 1)
 #'
 #' # a length-three variable, with each element following a standard normal
 #' # distribution
-#' alpha = normal(0, 1, dim = 3)
+#' alpha <- normal(0, 1, dim = 3)
 #'
 #' # a length-three variable of lognormals
-#' sigma = lognormal(0, 3, dim = 3)
+#' sigma <- lognormal(0, 3, dim = 3)
 #'
 #' # a hierarchical uniform, constrained between alpha and alpha + sigma,
-#' eta = alpha + uniform(0, 1, dim = 3) * sigma
+#' eta <- alpha + uniform(0, 1, dim = 3) * sigma
 #'
 #' # a hierarchical distribution
-#' mu = normal(0, 1)
-#' sigma = lognormal(0, 1)
-#' theta = normal(mu, sigma)
+#' mu <- normal(0, 1)
+#' sigma <- lognormal(0, 1)
+#' theta <- normal(mu, sigma)
 #'
 #' # a vector of 3 variables drawn from the same hierarchical distribution
-#' thetas = normal(mu, sigma, dim = 3)
+#' thetas <- normal(mu, sigma, dim = 3)
 #'
 #' # a matrix of 12 variables drawn from the same hierarchical distribution
-#' thetas = normal(mu, sigma, dim = c(3, 4))
+#' thetas <- normal(mu, sigma, dim = c(3, 4))
 #'
 #' # a multivariate normal variable, with correlation between two elements
 #' Sig <- diag(4)
 #' Sig[3, 4] <- Sig[4, 3] <- 0.6
-#' theta = multivariate_normal(rep(mu, 4), Sig)
+#' theta <- multivariate_normal(rep(mu, 4), Sig)
 #'
 #' # 10 independent replicates of that
-#' theta = multivariate_normal(rep(mu, 4), Sig, dim = 10)
+#' theta <- multivariate_normal(rep(mu, 4), Sig, dim = 10)
 #'
 #' # a Wishart variable with the same covariance parameter
-#' theta = wishart(df = 5, Sigma = Sig)
+#' theta <- wishart(df = 5, Sigma = Sig)
 #'
 #' }
 NULL

R/tf_functions.R

@@ -212,6 +212,11 @@ tf_icloglog <- function (x)
 tf_icauchit <- function (x)
   fl(1 / pi) * tf$atan(x) + fl(0.5)
 
+tf_imultilogit <- function (x) {
+  zeros <- tf$zeros(shape(nrow(x), 1L), tf_float())
+  latent <- tf$concat(list(x, zeros), 1L)
+  tf$nn$softmax(latent)
+}
 
 # map R's extract and replace syntax to tensorflow, for use in operation nodes
 # the following arguments are required:
@@ -328,6 +333,7 @@ tf_functions_module <- module(tf_as_logical,
                               tf_iprobit,
                               tf_icloglog,
                               tf_icauchit,
+                              tf_imultilogit,
                               tf_extract,
                               tf_recombine,
                               tf_replace,