Index of values

a_mem a n returns the ith element of array a, where i is the id of node n.

Returns the absolute value.

add h x adds the element x to the heap h, growing if necessary.

add hs x adds x to hash set hs.

Adds a child to the node and updates the parameters based on the number of samples in the child node.

add_child parent child appends child to the children of parent and updates the parent of child accordingly.

adotprod xa ya returns the dot product of the two float array xa and ya.

adotprod_autobias xa ya returns the dot product of the two float array xa and ya and assumes that the last entry in ya is the bias term.

adotprod_bias xa ya returns the dot product of the two float array xa and ya plus the bias term bias.

all_var_dists circ returns all plus nodes in the circuit circ that represent distributions.

Returns all indicator nodes (variable nodes).

logsumexp xa computes log(\sum_i e^{x_i}), where x_i is the ith entry of the float array xa.

argmax a returns the index of the largest value in the array a.

argmin a returns the index of the smallest value in the array a.

Creates a new array such that if the ith entry of the given array has value v, then the ith entry of the new array contains the pair (v,i).

build lessthan a builds a heap from the existing array a with respect to the comparison function lessthan.

Create hash maps from variable name -> index, and from each variable's values -> value index.

Construct key statistics for PLL value and gradient computations

Construct key statistics for PLL value and gradient computations, for use with external C implementation.

Construct small cache data structure, not precomputing feature counts

Construct small cache data structure, not precomputing feature counts, for external C implementation.

c_satisfied x cond returns true if x satisfies cond.

check_evidence schema ev runs Data.check_point for evidence ev and returns ev.

check_point schema example is used for simple input validation.

children node returns the children of node.

children bn idx

clear name stops the timer name and sets its elapsed time to zero.

clear h removes all elements from the heap h.

Set of arguments common to most programs

Initializes the common log channels in Libra.

Returns the log partition function.

Returns true when two given conditions are not mutually exclusive.

cond_prob bn x i

Converts a condition to evidence.

condl_to_valsetl schema condl converts list of conditions condl to a list of variable value sets, so that each variable appears only once in the list.

cons x l appends the value x to the list l.

Gets the parameter value stored in the constant node

copy f creates a duplicate of factor f.

count f a applies the boolean function f on each element of the array a and returns the number of times that f returns true.

count f l applies boolean the function f on each element of the list l, and returns the number of times that f returns true.

cpd_to_factors bn cvar cpd converts cpd associated with variable cvar of network bn to Markov network factors

create schema factors returns a Mn.network for the given schema and factors.

create lessthan n creates a heap with the capacity n and the comparison function lessthan.

create n creates a new, empty hash set with initial size n.

Creates a constant node (parameter node).

Initializes the data structure for computing the derivatives.

create_empty_network s creates an empty network with given variable schema s

Returns create_evidence_s of the circuit.schema.

create_evidence schema creates an matrix of ones as the values of indicator variables.

Creates a leaf node for the given data and schema.

Build MN structure for use with external C PLL implementation.

create_node t pt c pm d s creates a new spn node and adds it to Spn.node_array.

Creates a node with the given node specification and children.

Creates a dummy node.

Creates a plus node for the given data and schema.

Creates a plus (sum) node.

Create helper data structure.

Creates a times node for the given data and schema.

Creates a times (product) node.

Creates an indicator node.

create_var idx dim

Prints the given string into the standard output.

delta name returns time passed since the timer name was most recently started or delta was last called for this timer.

Returns the depth of an arithmetic circuit

dfind h key default returns the item with key key if the key is present in h, and returns default otherwise.

dist_var n returns the index of the random variable being modeled given the node n representing a multinomial distribution.

Prints to a log with printf-style formatting in the debug mode.

Prints to a log with printf-style formatting in the debug mode.

Similar to dlogfz, but accepts two lazy arguments.

Similar to dlogfz, but accepts three lazy arguments.

Similar to dlogfz, but accepts four lazy arguments.

dotprod xl yl returns the dot product of the two float lists xl and yl.

dotprod_bias xl yl bias returns the dot product of the two float lists xl and yl plus the bias bias.

dump_data data filename writes data examples to file filename.

dump_schema s filename writes schema s to file filename.

elapsed name returns the amount of time elapsed on the timer name.

Hash of accumulated elapsed times.

Convert an error type into a human-readable error string, mainly for debugging purposes.

ev_intersect e1 e2 returns true when evidence e1 and e2 are not mutually exclusive.

ev_subset e1 e2 returns true if evidence e1 is subset of evidence e2 (i.e.

ev_union circuit ev1 ev2 creates another evidence e such that e is the union of ev1 and ev2.

exists p [|a1; ...; an|] checks if at least one element of the array satisfies the predicate p.

Return the Expat library version as a string (e.g.

expected_log_value f logmarg negmarg returns the expected value of factor f given a fully factorized distribution.

Create a new XML_Parser object for parsing an external general entity.

f_satisfied x f returns true if an instance x satisfies all conditions of feature f.

Tautology function, i.e.

Compute PLL of a single example from a feature array.

fast_exp x returns the exponential of the given argument, but approximates very small return values with zero.

feature_contains_var f v returns true if feature f has a condition of variable v.

feature_node f returns the acnode related to the feature f.

See Mn.Factor.vars.

feature_value f returns the value of feature f when its condition is satisfied.

Checks whether the filename ends with .bif

Checks whether the filename ends with .cn, .dn, or .bn

Checks whether the filename ends with .dn

Returns true if filename ends with .mn.

Returns true if filename ends with .uai.

Checks whether the filename ends with .xmod

filter f hs applies the boolean function f on every item in hs and remove the item if f returns false.

filter p a returns all the elements of the array a that satisfy the predicate p.

Inform the parser that the entire document has been parsed.

find p a returns the first element of the array a that satisfies the predicate p.

Returns a flat array of all indicator nodes for all random variables.

Flatten a matrix (array of array) into an array.

The elements of the argument are all concatenated together (in the same order) to give the result.

fmatch x cond_ar checks whether example x satisfies set of conditions cond_ar.

fold f hs init computes (f iN ... (f i1 init)...), where i1 ... iN are the items in hs.

for_all p [|a1; ...; an|] checks if all elements of the array satisfy the predicate p.

fweight x f returns the weight of feature f if example x satisfies the feature.

get h i gets the ith element in the heap h.

Get the base for resolving relative URIs.

Computes derivatives of an arithmetic circuit with respect to all features conditioned on evidence.

get_logmarginals circ ds ev returns the log-marginals of all variables in circuit circ given evidence ev.

get_marginals circ ds ev return the marginals of all variables in circuit circ given evidence ev.

get_mpe circ ds e computes the MPE state of variables given the circuit circ, its empty derivative data structure ds, and the evidence ev.

Returns an spnode with the given id from node_array.

get_range mn v returns the cardinality of variable v in network mn.

get_range bn i

Returns the size of an SPN network (the lengths of node_array)

Returns a string value corresponding to the node type.

grow h automatically resizes the size of the heap h.

h_split node num_part lambda concurrent_thr ratio learns a sum node using sample clustering for conditional distributions.

Similar to h_split, but never raises HSplitNotPossible, so it always find a split.

id node returns the id of node.

Helper function for iterating over all states of the parent variables.

Reads real-valued evidence vector from the input.

Reads a list of real-valued evidence vectors from the input.

Reads one data example from the input.

Reads a list of data examples from the input.

Reads a list of data examples and validates them against the given schema.

Load features from an input channel.

Load features from an existing lexing buffer.

Reads vector of marginal distributions from the input.

Reads one weighted data example from the input, e.g., 10 | 0 1 1 1 is equivalent to 10 copies of the example 0 1 1 1.

Reads a list of weighted data examples from the input.

Reads a list of weighted data examples and validates them against the given schema.

Checks whether the node is a constant (parameter) node.

is_empty h checks if the heap h is empty.

Checks whether the node is a null node.

Checks whether the node is a plus (sum) node.

is_sot_dist node returns true if node is a sum of variable products (constants times indicator variables).

Checks whether the node is a times (product) node.

Checks whether the node is an indicator node.

is_var_product node returns true if node is a product of a constant and an indicator variable.

iter f h applies the function f over the elements of the heap h.

iter f hs applies f to all items in the hash set hs.

iter2 f [|a1; ...; an|] [|b1; ...; bn|] calls in turn f a1 b1; ...; f an bn.

iter2 f [a1; ...; an] [b1; ...; bn] [c1; ...; c_n] calls in turn f a1 b1 c1; ...; f an bn cn] .

keys_to_list h returns a list of all keys in hashtable h.

last_elapsed name returns the amount of time elapsed between last start and stop on timer name.

Hash of last elapsed times.

Loads a Markov network in .mn format from a channel.

Reads an arithmetic circuit from a file description.

load_auto filename loads an MN, inferring its filetype by its filename.

Read a BN from a file, inferring the format from filename.

Input a BN in BIF format.

Input a BN in CN format.

load_data file reads examples from file filename and returns them as a list.

load_data_ar file reads examples from file filename and returns them as an array.

load_evidence filename reads real-valued evidence vectors from file filename and returns them as a list.

load_evidence_ar filename reads real-valued evidence vectors from file filename and returns them as an array.

Constructs a SPN structure given the model file

load_schema filename loads a schema from file filename.

Loads a Markov network written in the UAI file format.

Reads an arithmetic circuits and log-linear features from .ac file.

Input a BN in .xmod format.

Prints the arguments to a log in the verbose mode.

Constant string debug for logging in the debug mode.

Checks if the given name maps to any output channel.

Used for mapping output channel to a specific name.

Constant string normal for logging in the normal mode.

Zero constant.

Returns the output channel corresponding to the given name.

Outputs a string to the specified log, if defined.

log_value x f returns the value of the factor for the configuration given by example x.

Constant string verbose for logging in the verbose mode.

Negative infinity constant.

logf outname fmt arg1 ... argN prints to a log named outname with printf-style formatting.

loglikelihood bn x computes the log-likelihood of the network on a single example x.

logprob_ev scratch circ ev returns the log probability of evidence structure ev in circuit circ.

logprob_x scratch circ x returns the log probability of example x in circuit circ.

logsumexp xl computes log(\sum_i e^{x_i}), where x_i is the ith entry of the float list xl.

logsumexp2 x y computes log(exp(x)+exp(y), but avoids numerical underflow.

Maps Circuit.circuit.nodes to their logvalues after the evaluation of the circuit.

Creates an array of child lists, given an array of parent lists

Creates list of variables in topographic order

Creates an indicator variable for each state of every variable based on the given schema

map2 f [|a1; ...; an|] [|b1; ...; bn|] is [|f a1 b1; ...; f an bn|].

map f [a1; ...; an] [b1; ...; bn] [c1; ...; c_n] returns the list f a1 b1 c1; ...; f an bn cn.

match_beg s1 s2 i checks if the first i characters of s1 and s2 are equal.

match_end s1 s2 i checks if the last i characters of s1 and s2 are equal.

max a returns the maximum value of the array a.

maxf x y returns the maximum of the float values x and y.

mb_logdist mn x i returns the log probability distribution over variable i in network mn, given the values of its Markov blanket variables specified in instance x.

mb_logprob bn x i

mb_prob bn x i

mem hs x checks if x exists in hs.

min h accesses the smallest element in the heap h.

min a returns the mininum value of the array a.

minf x y returns the minmum of the float values x and y.

minimize f x decrease maxiter finds the minimum of function f, starting its search at point x.

minimize_l1 c f x decrease maxiter finds the minimum of function f with L1 penalty weight c, starting its search at point x.

minimize_l1_simple c f x finds the minimum of function f with L1 penalty weight c, starting its search at point x.

minimize_simple f x finds the minimum of function f, starting its search at point x.

Copy constructor.

Prints to a log with printf-style formatting in the normal mode.

Returns the node at a specific index.

Stores the nodes of an SPN network.

node_iter f root applies function f to every node in the circuit rooted at root.

node_logscore bn x v

node_map f root applies function f to every node in the circuit rooted at root and creates a list of the results.

node_name returns the name of node node, e.g., "n5".

normalize_inplace_log xa Normalizes the elements of xa so that they represent the log probabilities of a valid probability distribution.

normalize_inplace_raw xa Normalizes the elements of xa so that they represent a valid probability distribution.

Similar to normalize_inplace_log, but returns a new array as the result (does not modify the argument.

Similar to normalize_inplace_raw, but returns a new array as the result (does not modify the argument.

Creates a node with type NullNode, and its parent and children lists are empty.

num_children node returns the number of children of node.

Return the number of edges in an arithmetic circuit.

Return the number of parameters in an arithmetic circuit.

numchildren bn idx

numparams bn idx

numparents bn idx

Returns the number of variables in an arithmetic circuit.

of_graph schema vn vnl root constructs a new circuit from schema and root node and places it in topological order.

Puts all nodes in the circuit into a topological partial order.

Writes a Markov network in .mn format to a channel.

Stores an arithmetic circuit in the regular plain text format.

Output a BN in BIF format.

Output a BN in CN format.

Stores an arithmetic circuit in the dot format (for visualizing).

Writes an example to the output channel.

output_factor out f writes factor f to channel out.

See Mn.Factor.output_factor.

output_feature out f writes feature f to the output channel out.

See Mn.Factor.output_factor.

Stores the feature list given the circuit root.

output_float out f writes the float value f on the output channel out.

output_int out i writes the integer value i on the output channel out.

Stores an arithmetic circuit in the javascript format for visualizing using Digraph.

Writes vector of marginal distributions to the output.

output_node out node writes node node to the output channel out.

output_root out schema root recursively writes circuit to output channel out, starting at the root.

Stores the given arithmetic circuit and feature list in .ac format, given the circuit schema, root node, and feature list.

dump_schema o s writes schema s to output channel o.

output_string ouput sep sa concatinates the string elements of the string array sa using the separator sep, and then writes the result on the output channel out.

See Mn.Factor.output_factor.

Writes a Markov network in the UAI file format.

Stores the given arithmetic circuit and feature list in .ac format.

Output a BN in .xmod format.

parents node returns the parents of node.

parents bn idx

Let the parser parse a chunk of an XML document.

Let the parser parse a chunk of an XML document in a substring

Create a new XML parser.

Create a new XML parser that has namespace processing in effect

partition p a returns a pair of arrays (a1, a2), where a1 is the array of all the elements of a that satisfy the predicate p, and a2 is the array of all the elements of a that do not satisfy p.

Build a factor from parser output.

pll mn x computes the pseudo-likelihood of sample x given network mn.

pll bn x computes the pseudo-log-likelihood of the network on a single example x.

Construct an MN from the PLL cache data and the weight vector.

Compute PLL of MN.

Compute value and gradient of PLL without using full cache

Compute value and gradient PLL without using full cache.

Compute value and gradient of PLL using cache

Compute PLL and its gradient using cache.

prefix s1 s2 checks if s2 is the prefix of s1.

Prints an arithmetic circuit to stdout.

Print evidence data structure to stdout.

Prints marginals to stdout.

Writes the SPN structure into the output file

Writes the information of a single SPN node into an output file.

Prints a single node to stdout.

Prints a single node to stdout, including a newline at the end.

Prints SPN information.

Prunes features by removing empty, redundant, and contradictory circuit features.

Prunes a circuit conditioned on evidence.

range n creates an array with values from 0 to n-1.

range n creates a list with values from 0 to n-1.

raw_logprob mn state returns the unnormalized log probability of the variable configuration state according to network mn.

realloc h n grows the capacity of the heap h to n.

Rebuilds the circuit data structure from the root node.

rebuild h fixes all out-of-order elements in the heap h in linear time.

register_log name out maps the output channel out to the string name.

relatedl_a size rel nl returns an array in which the ith element is true if the node with id i is reachable from some node in nl through relation rel.

rem_first l returns the tail of the list l.

rem_item l i returns a list that has the same order as the list l, which does not include the value i.

remove h i removes the ith element in the heap h.

remove_all h f removes all elements from the heap h that match the criterion f.

remove_all_children parent removes all children of parent.

remove_all_parents child removes all parents of child.

remove_child parent child removes child from the children of parent and updates child accordingly.

remove_fast x l removes the value x from the list l.

remove_min h removes the smallest element in heap h (root).

Removes redundant conditions from a condition list.

rescale alpha f rescales factor f by a factor of alpha.

rev_flatten a dimx dimy creates a matrix given an array and the dimension of the matrix.

rev_inplace a reverses the array a in place.

rev_iter f a is the same as iter f a, but iterates from right to left.

rev_iter f l is the same as iter f l, but iterates from right to left.

root_to_list root generates a list of all nodes reachable from root.

Generates a single iid sample from the BN probability distribution.

Utility function for sampling a multinomial given an array containing the probability of each state.

Infers the data schema from the given examples.

Returns the schema of an arithmetic circuit.

Set the base to be used for resolving relative URIs in system identifiers.

set_const node wt sets the value of const node node to wt.

set_cpt bn var_idx parents cpt_values updates a CPT.

set_cptree bn var_idx root updates the CPD of variable var_idx in network bn with a CPD with root node root.

set_factorset bn var_idx fl updates the CPD of variable var_idx in network bn with factor list fl.

set_feature_value f wt sets the value of feature f to wt, modifying the associated constant node.

Enable the parsing of parameter entities

Sets the parent of a node, and also updates the children of the parent node.

set_weights w wi f updates the weight in factor f using a global weight vector w.

set_weights mn w updates weights using weight vector w.

sibling_vars circ n returns a list of all indicator nodes for all values of the same variable given the indicator node n for a particular variable/value combination.

sift_down h root moves root down to satisfy the heap property for the heap h.

sift_up h child moves child up to satisfy the heap property for the heap h.

Comparision function used for sorting conditions.

simplify ev f simplifies factor f by conditioning it on evidence ev.

simplify mn ev builds a new MN with simpler factors, conditioned on evidence ev.

simplify bn ev simplifies CPDs of netwprk bn given evidence ev if the CPDs are represented using trees or factor sets.

simplify_feature ev f removes all conditions of feature f that are satisfied by evidence ev and returns a new, simpler feature.

size h returns the number of items in the heap h.

Transform a list of triples into a triple of lists: split3 [(a1,b1,c1); ...; (an,bn,cn)] is ([a1; ...; an], [b1; ...; bn]).

starts name starts a timer with the name name.

Hash of start times.

stats_add s w x adds one observation of x to the counters

Create an object to help compute of mean and variance statistics.

stats_mean s returns the arithmetic mean of the observations

stats_n s returns the number of (weighted) observations

stats_stddev s returns the standard deviation of the observations

stats_stderr s returns the standard deviation of the mean

stats_sum s returns the sum of the observations

stats_var s returns the estimated variance of the observations

stats_wadd s w x adds w observations of x to the counters

stop name stops the timer name.

Infers the data schema from the given input stream.

Converts a float array to a string.

Converts the given feature to a string

Converts a float list to a string.

Converts an int array to a string.

Converts an int list to a string.

string_of_node node returns a string representation of node node.

suffix s1 s2 checks if s2 is the suffix of s1.

sum a returns the sum of the elements of the int array a.

sum l return the sum of the elements of the int list l.

sum_map f hs applies f to all items in hs, returns the sum of all the results.

sum_map f a applies the function f on each element of the int array a and then returns the sum of the results.

sum_map f l applies the function f on each element of the int list l and then returns the sum of the results.

sum a returns the sum of the elements of the float array a.

sum l returns the sum of the elements of the float list l.

sum_map f hs applies f to all items in hs, returns the sum of all the results.

sum_map f a applies the function f on each element of the float array a and then returns the sum of the results.

sumf_map f l applies the function f on each element of the float list l and then returns the sum of the results.

swap h x y swaps two items x and y in the heap h.

Converts table CPD to features

take a ai removes the indices ai from the array a.

takeCol m cols removes the columns cols from the matrix m.

takeRow m rows removes the rows rows from the matrix m.

to_array h constructs an array from the elements of the heap h.

Converts factor to a set of conjunctive features.

to_features mn converts the factors in network mn to a list of features.

to_list h constructs a list from the elements of the heap h.

Converts the given hash set into a list.

Converts the Bayesian network to a Markov network by introducing one factor for each CPD.

Converts the given example to a string.

Converts the given schema to a string.

to_table schema f converts factor f to another factor with a table representation.

transpose m transposes the matrix m represented as an array of arrays.

transpose m transposes the matrix m represented as a list of lists.

tree_logprob x cpdnode

tree_parents numvars root

Generates an equivalent set of features for a given tree factor.

See Mn.Factor.vars.

unlink node removes all parents and children of node.

unregister_log name removes the channel-name mapping for the given name.

Updates the children lists and topological variable order

Update parent references for all nodes to be consistent.

v_split node cut_thr learns a product node using variable clustering.

v_split_gtest node cut_thr learns a product node using variable clustering.

var circ i returns the array of indicator nodes for the given random variable index i and the circuit circ.

var_details node returns the variable-value pair if node is an indicator variable node

var_value node returns the value if node is a variable node

var_var node returns the variable if node is a variable node

varname bn idx

Prints to a log with printf-style formatting in the verbose mode.

Prints to a log with printf-style formatting in the verbose mode.

Similar to vlogfz, but accepts two lazy arguments.

Similar to vlogfz, but accepts three lazy arguments.

Similar to vlogfz, but accepts four arguments.

wrap x returns a list that contains x.

write_auto filename mn writes Markov network mn to the output file filename, inferring its filetype by its filename.

Write the BN to a file, inferring the format from filename.

Converts a xml_error to a string