Update count matrices used for Gibbs sampling @param w @param t @param d @param update

Draw a multinomial sample from (un-normalized) vals @param rng @param vals @param normsum @return

Nice summary of rule set and nontrivial groundings

@param c @param T

Ensure that evidence has been applied

Simple conversion of List<Integer> to int[] @param vec @return

Ensure that evidence has been applied

Read a plaintext file containing a sequence of integers @param filename @return

Estimate mean of Phi posterior from current sample @param p @return

Ensure that evidence has been applied

Ensure that evidence has been applied

Ensure that evidence has been applied

Ensure that evidence has been applied

How many (non-trivial) groundings @return

Ensure that evidence has been applied

Ensure that evidence has been applied

Ensure that evidence has been applied

Ensure that evidence has been applied

Return a List of all currently unsatisifed groundings @return

Return a new int[] which is the concatenation of a and b @param a @param b @return

Calc uncollapsed LDA log-likelihood @param nw @param nd @param phi @param theta @param beta @param alpha @return

Static factory for constructing rules from file

Do bookkeeping assoc with reassigning z[idx] to newz @param c @param idx @param newz

Standard (no logic) LDA, using collapsed Gibbs sampling @param c @param p @param numsamp @return

Write logic satisfaction report out to .logic file @param z @param basename

Calc total satisfied weight over all rules @param z @return

Estimate phi/theta and write out to %s.phi/theta @param p Hyperparameters alpha/beta are required to estimate phi/theta @param basefn Base output fil

Groundings are equal if corresponding .values lists are equal (tricky! see http://www.artima.com/lejava/articles/equality.html)

Calc global obj fcn impact of z[idx]'s current value @param z @param idx @return

Evaluate the satisfied weight wrt a single idx in the grounding @param grounding @param z @return

Weight of this rule for random sampling purposes @return [total sampling weight] = [sampling weight] x [num groundings]

Do a single Collapsed Gibbs sample @param c Contains words, documents @param p Contains hyperparameters @param s Sample object to be updated in place

For this rule, is a particular grounding satisfied? @param z @param g @return

For this rule, is a particular grounding satisfied? @param z @param g @return

For this rule, is a particular grounding satisfied? @param z @param g @return

For this rule, is a particular grounding satisfied? @param z @param g @return

Given a sample z, how many groundings are satisfied? @param z @return

Repopulate z-sample, making necessary data struct changes @param z

Dirichlet log-likelihood of multinomial parameter vector x given Dirichlet hyperparameter vector alpha @param alpha @param x @return

Do stochastic gradient descent MAP inference @param c @param p @param relax @param numouter @param numinner @param stepa @param stepb @return

Given Grounding g, return [idx, newz] for best greedy move @param g @return

Actual weight of this logic rule @return [rule weight] = [sampling weight] [step weight]

Do argmax to convert relaxed z to a single "hard" z sample @return

Cast List of LogicRule to GroundableRule @param rules @return

Do a single Ground LogicLDA sample @param logicweights @param c @param p @param s @param onlineInit

Is this index invovled in any logic fcn? @param idx @return

Calc sum of logDir contrib for multinomial parameter matrix x @param alpha @param x @return

Do a single Logic Collapsed Gibbs sample @param logicweights N x T matrix of (independent) rule contributions @param c Contains words, documents @par

Calculate MAP phi @param p @param enw

Calculate MAP theta @param p @param end

How many (non-trivial) groundings does this rule have?

Return a randomly sampled unsatisfied clause, or null if all clauses are satisfied @return

Read an UJMP matrix from text file @param filename @return @throws IOException @throws MatrixException

Just return each line of the file as String @param filename @return

Get the maximum value from an AbstractCollection @param <T> @param seq @return

After an update of z at position idx, update unsat ground clause @param z @param idx

Write these integers out to plaintext file @param filename @param vals @throws IOException

Write matrix out to plaintext file @param filename @param mat @throws IOException

For all zi which are not part of any logic formula, simply assign by argmax (phi theta) @param c @param gr @param phi @param theta @param s @return

Build rule set from .rules file @param basefn @param c @param T @param randseed @param ldaRule @return

External method for doing online-init, then numsamp Gibbs samples @param c Contains words, documents @param p Contains hyperparameters @param numsamp

Externally called method for Ground LogicLDA sampling @param gr @param c @param p @param numsamp @return

External method for doing online-init, then numsamp Logic Gibbs samples @param logicweights N x T matrix of (independent) rule contributions @param c

Do MaxWalkSAT inference @param lstRules @param p @param numiter @param prand @param s @param randseed @return

Simple conversion of List<Double> to double[] @param vec @return

Do an Entropic Mirror Descent step on a single relaxed z-entry @param i Index to update @param gradSignMag Sign and magnitude of gradient @param tGra

Write a single String out to file @param filename @param output @throws IOException

Initialize from a .sample file @param filename @param c

The inverted index maps corpus indices to affected groundings @return

Load any recognized side information files @param basefn @return @throws IOException

Estimate mean of Theta posterior from current sample @param p @return

Once rule is grounded, need to create large data struct of all actual groundings (otherwise sort of 'cheating') @param T

Initialize ground rule data structures (must be called before any other GroundableRule interface calls) @param z

Return 0 if no n-gram seed hit, otherwise return length of longest n-gram seed hit @param w @param start @return

Calculate LDA gradient for a given index i @param i @return

Calculate contrib of phi-counts to loglike @param phi TxW topic P(w|z) matrix @param nw WxT count matrix @return

Calculate contrib of theta-counts to loglike @param theta DxT topic P(z|d) matrix @param nd DxT count matrix @return

How many (non-trivial) groundings does this rule have? (could multiply by 2 to account for each side of bidirectional, but do not, in order to be con

Parse line of doubles @param line @return nothing - but adds elements to vec

Parse line of integers @param line @return nothing - but adds elements to vec

Randomly sample a grounding and take a gradient step wrt to it @param relax Relaxed z-array (will be updated in-place!)

Do LDA-MaxWalkSAT inference @param lstRules @param p @param numiter @param prand @param s @param randseed @return

Run LogicLDA MAP inference, starting from init DiscreteSample @param c @param p @param rs @param s @param numouter @param numinner @return

Get the z-label style weights for all IndependentRule (for use in Logic Collapsed Gibbs, etc) @param N @param T @return

After calling applyEvidence, this returns a 2D array mapping each corpus idx -> logic weights for each potential z-value [w1, ..., wT] This can then

Re-calculate MAP phi @param p @param enw