version 3.7

Author: Marco Scutari

Changelog

@@ -1,3 +1,28 @@
+bnlearn (3.7)
+
+  * fixed the default setting for the number of particles in cpquery()
+     (thanks Nishanth Upadhyaya).
+  * reimplemented common test patterns in monolithic C functions to speed
+     up constraint-based algorithms.
+  * added support for conditional linear Gaussian (CLG) networks.
+  * fixed several recursion bugs in choose.direction().
+  * make read.{bif,dsc,net}() consistent with the `$<-` method for bn.fit
+     objects (thanks Felix Rios).
+  * support empty networks in read.{bif,dsc,net}().
+  * fixed bug in hc(), triggered when using both random restarts and the
+     maxp argument (thanks Irene Kaplow).
+  * correctly initialize the Castelo & Siebes prior (thanks Irene Kaplow).
+  * change the prior distribution for the training variable in classifiers
+     from the uniform prior to the fitted distribution in the
+     bn.fit.{naive,tan} object, for consistency with gRain and e1071 (thanks
+     Bojan Mihaljevic).
+  * note AIC and BIC scaling in the documentation (thanks Thomas Lefevre).
+  * note limitations of {white,black}lists in tree.bayes() (thanks Bojan
+     Mihaljevic).
+  * better input sanitization in custom.fit() and bn.fit<-().
+  * fixed .Call stack imbalance in random restarts (thanks James Jensen).
+  * note limitations of predict()ing from bn objects (thanks Florian Sieck).
+
 bnlearn (3.6)
 
   * support rectangular nodes in {graphviz,strength}.plot().
@@ -18,12 +43,12 @@ bnlearn (3.6)
   * attach the weights to the return value of cpdist(..., method = "lw").
   * changed the default number of simulations in cp{query, dist}().
   * support interval and multiple-valued evidence for likelihood weighting
-      in cp{query,dist}().
+     in cp{query,dist}().
   * implemented dedup() to pre-process continuous data.
   * fixed a scalability bug in blacklist sanitization (thanks Dong Yeon Cho).
   * fixed permutation test support in relevant().
   * reimplemented the conditional.test() backend completely in C for
-      speed, it is now called indep.test().
+     speed, it is now called indep.test().
 
 bnlearn (3.5)
 

DESCRIPTION

@@ -1,9 +1,9 @@
 Package: bnlearn
 Type: Package
-Title: Bayesian network structure learning, parameter learning and
-        inference
-Version: 3.6
-Date: 2014-06-17
+Title: Bayesian Network Structure Learning, Parameter Learning and
+        Inference
+Version: 3.7
+Date: 2015-01-13
 Depends: R (>= 2.14.0)
 Imports: methods
 Suggests: parallel, graph, Rgraphviz, lattice, gRain
@@ -14,8 +14,8 @@ Description: Bayesian network structure learning, parameter learning and
   This package implements constraint-based (GS, IAMB, Inter-IAMB, Fast-IAMB,
   MMPC, Hiton-PC), pairwise (ARACNE and Chow-Liu), score-based (Hill-Climbing
   and Tabu Search) and hybrid (MMHC and RSMAX2) structure learning algorithms
-  for both discrete and Gaussian networks, along with many score functions
-  and conditional independence tests.
+  for discrete, Gaussian and conditional Gaussian networks, along with many
+  score functions and conditional independence tests.
   The Naive Bayes and the Tree-Augmented Naive Bayes (TAN) classifiers are
   also implemented.
   Some utility functions (model comparison and manipulation, random data
@@ -27,7 +27,7 @@ Description: Bayesian network structure learning, parameter learning and
 URL: http://www.bnlearn.com/
 License: GPL (>= 2)
 LazyData: yes
-Packaged: 2014-06-17 17:15:42 UTC; fizban
+Packaged: 2015-01-13 23:45:31 UTC; fizban
 NeedsCompilation: yes
 Repository: CRAN
-Date/Publication: 2014-06-17 21:16:12
+Date/Publication: 2015-01-14 01:52:50

MD5

@@ -142,28 +142,26 @@ S3method(as.graphNEL, bn.fit)
 S3method(as.graphAM, bn)
 S3method(as.bn, graphAM)
 S3method(as.graphAM, bn.fit)
-S3method(ci.test, character)
-S3method(ci.test, data.frame)
-S3method(ci.test, matrix)
-S3method(ci.test, numeric)
-S3method(ci.test, factor)
-S3method(ci.test, default)
 S3method(print, "bn.fit")
 S3method(print, "bn.fit.dnode")
 S3method(print, "bn.fit.onode")
 S3method(print, "bn.fit.gnode")
+S3method(print, "bn.fit.cgnode")
 S3method(residuals, "bn.fit")
 S3method(residuals, "bn.fit.dnode")
 S3method(residuals, "bn.fit.onode")
 S3method(residuals, "bn.fit.gnode")
+S3method(residuals, "bn.fit.cgnode")
 S3method(fitted, "bn.fit")
 S3method(fitted, "bn.fit.dnode")
 S3method(fitted, "bn.fit.onode")
 S3method(fitted, "bn.fit.gnode")
+S3method(fitted, "bn.fit.cgnode")
 S3method(coef, "bn.fit")
 S3method(coef, "bn.fit.dnode")
 S3method(coef, "bn.fit.onode")
 S3method(coef, "bn.fit.gnode")
+S3method(coef, "bn.fit.cgnode")
 S3method(print, "mvber.moments")
 S3method(print, "bn.kcv")
 S3method(plot, "bn.strength")
@@ -172,6 +170,7 @@ S3method(predict, "bn.fit")
 S3method(predict, "bn.fit.dnode")
 S3method(predict, "bn.fit.onode")
 S3method(predict, "bn.fit.gnode")
+S3method(predict, "bn.fit.cgnode")
 S3method(predict, "bn.naive")
 S3method(predict, "bn.tan")
 S3method("$<-", "bn.fit")

NAMESPACE

@@ -8,7 +8,7 @@ arc.operations = function(x, from, to, op = NULL, check.cycles, update = TRUE,
   # check x's class.
   check.bn(x)
   # check the op code.
-  if (!(op %in% available.ops))
+  if (op %!in% available.ops)
     stop("valid op codes are 'set', 'drop' and 'reverse'.")
   # a valid node is needed.
   check.nodes(nodes = from, graph = x, max.nodes = 1)

R/arc.operations.R

@@ -205,7 +205,7 @@ arc.strength.boot = function(data, cluster = NULL, R, m, algorithm,
     # get the number of slaves.
     s = nSlaves(cluster)
 
-    res = parLapply(cluster, rep(ceiling(R / s), s), bootstrap.batch,
+    res = parallel::parLapply(cluster, rep(ceiling(R / s), s), bootstrap.batch,
             data = data, m = m, arcs = arcs, algorithm = algorithm,
             algorithm.args = algorithm.args, cpdag = cpdag, debug = debug)
 

R/arc.strength.R

@@ -78,7 +78,7 @@ neighbour = function(x, mb, data, alpha, B = NULL, whitelist, blacklist,
   # whitelisted nodes are included (arc orientation is irrelevant),
   # and blacklisted nodes are removed if both directed arcs are banned
   # and both are not in the whitelist.
-  nbrhood = nbrhood[!(nbrhood %in% blacklisted)]
+  nbrhood = nbrhood[nbrhood %!in% blacklisted]
   nbrhood = unique(c(nbrhood, whitelisted))
 
   # use backtracking for a further screening of the nodes to be checked.
@@ -92,7 +92,7 @@ neighbour = function(x, mb, data, alpha, B = NULL, whitelist, blacklist,
 
     # known.bad nodes are not to be checked for inclusion and/or used in
     # the subsets.
-    nbrhood = nbrhood[!(nbrhood %in% known.bad)]
+    nbrhood = nbrhood[nbrhood %!in% known.bad]
 
   }#THEN
 
@@ -133,7 +133,7 @@ neighbour = function(x, mb, data, alpha, B = NULL, whitelist, blacklist,
       cat("    > dsep.set = '", dsep.set, "'\n")
 
     a = allsubs.test(x = x, y = y, sx = dsep.set, min = ifelse(empty.dsep, 0, 1),
-          data = data, test = test, alpha = alpha, B = B, debug = debug)
+          data = data, test = test, alpha = alpha, B = B, debug = debug)[1]
 
     if (a > alpha) {
 
@@ -156,7 +156,7 @@ neighbour = function(x, mb, data, alpha, B = NULL, whitelist, blacklist,
 
   # do not even try to remove whitelisted nodes; on the other hand, known.good
   # nodes from backtracking should be checked to remove false positives.
-  sapply(nbrhood[!(nbrhood %in% whitelisted)], nbr, x = x, mb = mb, test = test)
+  sapply(nbrhood[nbrhood %!in% whitelisted], nbr, x = x, mb = mb, test = test)
 
   return(list(mb = mb[[x]], nbr = nbrhood))
 
@@ -177,11 +177,12 @@ vstruct.detect = function(nodes, arcs, mb, data, alpha, B = NULL, test,
 
     tos = parents.backend(arcs, x, TRUE)
 
-    if (length(tos) < 2) return(NULL)
+    if (length(tos) < 2)
+      return(NULL)
 
     # build a list of possibile parents for the node x, i.e. all the subsets
     # of size 2 of the nodes connected to x by incoming arcs.
-    tos.combs = subsets(length(tos), 2, tos)
+    tos.combs = subsets(tos, 2)
     vs = NULL
 
     for (j in 1:nrow(tos.combs)) {
@@ -193,70 +194,28 @@ vstruct.detect = function(nodes, arcs, mb, data, alpha, B = NULL, test,
         cat("  * checking", y, "->", x, "<-", z, "\n")
 
       # check there's no arc from y to z and vice versa.
-      if(!is.listed(arcs, c(y, z)) &&
-         !is.listed(arcs, c(z, y))) {
-
-        mby = mb[[y]][['mb']]
-        mbz = mb[[z]][['mb']]
-
-        # compute mb(y) - {x,z} and mb(z) - {x,y}
-        mby = mby[!(mby %in% c(x, z))]
-        mbz = mbz[!(mbz %in% c(x, y))]
-
-        # choose the smallest one to cut down the number of subsets to test.
-        dsep.set = smaller(mby, mbz)
-
-        if (debug)
-          cat("    > chosen d-separating set: '", dsep.set, "'\n")
-
-        k = 0
-        max_a = a = 0
-
-        repeat {
-
-          dsep.subsets = subsets(length(dsep.set), k, dsep.set)
-
-          for (s in 1:nrow(dsep.subsets)) {
+      if (is.listed(arcs, c(y, z), either = TRUE))
+        next
 
-            a = indep.test(y, z, c(dsep.subsets[s,], x), data = data,
-                  test = test, B = B, alpha = alpha)
-            if (debug)
-              cat("    > testing", y, "vs", z, "given", c(dsep.subsets[s,], x), "(", a, ")\n")
-            max_a = max(a, max_a)
-            if (a > alpha) {
+      # choose the smallest of mb(y) - {x,z} and mb(z) - {x,y} to cut down
+      # the number of subsets to test.
+      dsep.set = smaller(setdiff(mb[[y]][['mb']], c(x, z)),
+                         setdiff(mb[[z]][['mb']], c(x, y)))
 
-              if (debug)
-                cat("    >", y, "and", z, "are independent given '", c(dsep.subsets[s,], x), "' (", a, ")\n")
-              break
-
-            }#THEN
-
-          }#FOR
-
-          if (a <= alpha) {
-
-            if (k < length(dsep.set)) {
-
-              k = k + 1
-
-            }#THEN
-            else {
-
-              if (debug)
-                cat("    @ detected v-structure", y, "->", x, "<-", z, "\n")
-              vs = rbind(vs, data.frame(max_a, y, x, z, stringsAsFactors = FALSE))
-              break
-
-            }#ELSE
+      if (debug)
+        cat("    > chosen d-separating set: '", dsep.set, "'\n")
 
-          }#THEN
-          else {
+      assoc = allsubs.test(x = y, y = z, fixed = x, sx = dsep.set, data = data,
+                test = test, B = B, alpha = alpha, debug = debug)
+      a = assoc[1]
+      max_a = assoc[3]
 
-            break
+      if (a <= alpha) {
 
-          }#ELSE
+        if (debug)
+          cat("    @ detected v-structure", y, "->", x, "<-", z, "\n")
 
-        }#REPEAT
+        vs = rbind(vs, data.frame(max_a, y, x, z, stringsAsFactors = FALSE))
 
       }#THEN
 
@@ -288,13 +247,13 @@ vstruct.apply = function(arcs, vs, nodes, strict, debug = FALSE) {
       }#THEN
 
       if (strict)
-        stop(paste("vstructure", v["y"], "->", v["x"], "<-", v["z"],
-          "is not applicable, because one or both arcs are oriented",
-          "in the opposite direction."))
+        stop("vstructure ", v["y"], " -> ", v["x"], " <- ", v["z"],
+          " is not applicable, because one or both arcs are oriented",
+          " in the opposite direction.")
       else
-        warning(paste("vstructure", v["y"], "->", v["x"], "<-", v["z"],
-          "is not applicable, because one or both arcs are oriented",
-          "in the opposite direction."))
+        warning("vstructure ", v["y"], " -> ", v["x"], " <- ", v["z"],
+          " is not applicable, because one or both arcs are oriented",
+          " in the opposite direction.")
 
       return(NULL)
 
@@ -314,13 +273,13 @@ vstruct.apply = function(arcs, vs, nodes, strict, debug = FALSE) {
       }#THEN
 
       if (strict)
-        stop(paste("vstructure", v["y"], "->", v["x"], "<-", v["z"],
-          "is not applicable, because one or both arcs introduce cycles",
-          "in the graph."))
+        stop("vstructure ", v["y"], " -> ", v["x"], " <- ", v["z"],
+          " is not applicable, because one or both arcs introduce cycles",
+          " in the graph.")
       else
-        warning(paste("vstructure", v["y"], "->", v["x"], "<-", v["z"],
-          "is not applicable, because one or both arcs introduce cycles",
-          "in the graph."))
+        warning("vstructure ", v["y"], " -> ", v["x"], " <- ", v["z"],
+          " is not applicable, because one or both arcs introduce cycles",
+          " in the graph.")
 
       return(NULL)
 

R/backend-indep.R

@@ -43,10 +43,10 @@
   if (!is.null(x$learning$blacklist))
     x$learning$blacklist[] = value[match(x$learning$blacklist, nodes)]
   # ... Castelo & Siebes prior specification...
-  if (("prior" %in% names(x$learning$args)) && 
+  if (("prior" %in% names(x$learning$args)) &&
       (x$learning$args$prior == "cs")) {
 
-    x$learning$args$beta[, c("from", "to")] = 
+    x$learning$args$beta[, c("from", "to")] =
       value[match(x$learning$args$beta[, c("from", "to")], nodes)]
     attr(x$learning$args$beta, "nodes") = value
 
@@ -99,7 +99,7 @@
   }#FOR
 
   # relabel the nodes.
-  names(x) = value 
+  names(x) = value
 
   # update the training node of classifiers.
   if (is(x, c("bn.naive", "bn.tan")))

R/backend-s4.R

@@ -50,7 +50,7 @@ arcs.to.be.reversed = function(arcs, blacklist, nparents, maxp = Inf) {
     arcs = arcs[!which.listed(arcs[, c(2, 1), drop = FALSE], blacklist), , drop = FALSE]
 
   if (!missing(nparents))
-    arcs = arcs[nparents[arcs[, 1]] < maxp, ]
+    arcs = arcs[nparents[arcs[, 1]] < maxp, , drop = FALSE]
 
   return(arcs)