@@ -1,5 +1,5 @@

 Package: CoGAPS

-Version: 3.1.1

+Version: 3.1.2

 Date: 2018-04-24

 Title: Coordinated Gene Activity in Pattern Sets

 Author: Thomas Sherman, Wai-shing Lee, Conor Kelton, Ondrej Maxian, Jacob Carey,

@@ -11,39 +11,8 @@

 #' data(SimpSim)

 #' plotGAPS(SimpSim.result$Amean, SimpSim.result$Pmean)

 #' @export

-plotGAPS <- function(A, P, outputPDF="")

+plotGAPS <- function(A, P)

 {

-    if (outputPDF != "")

-    {

-        pdf(file=paste(outputPDF, "-Patterns", ".pdf", sep=""))

-    }

-    else

-    {

-        dev.new()

-    }

-

-    arrayIdx <- 1:ncol(P)

-    maxP <- max(P)

-    nPatt <- dim(P)[1]

-    matplot(arrayIdx, t(P), type='l', lwd=3, xlim=c(1,ncol(P)), ylim=c(0,maxP),

-        xlab='Samples',ylab='Relative Strength',col=rainbow(nPatt))

-    title(main='Inferred Patterns')

-    legend("topright", paste("Pattern", 1:nPatt, sep = ""), pch = 1:nPatt,

-        lty=1,cex=0.8,col=rainbow(nPatt),bty="y",ncol=5)

-

-    if (outputPDF == "")

-    {

-        dev.new()

-    }

-    else

-    {

-        dev.off()

-        pdf(file=paste(outputPDF, "-Amplitude", ".pdf", sep=""))

-    }

-

+    plotP(P)

     heatmap(A, Rowv=NA, Colv=NA)

-    if (outputPDF != "")

-    {

-        dev.off()

-    }

 }

@@ -7,25 +7,41 @@

 #' @examples

 #' data(SimpSim)

 #' plotP(SimpSim.result$Pmean, SimpSim.result$Psd)

+#' @importFrom gplots plotCI

 #' @export

-plotP <- function(Pmean, Psd)

+plotP <- function(Pmean, Psd=NULL)

 {

-    Nfactor <- nrow(Pmean)

-    Nobs <- ncol(Pmean)

-    RowP <- 1:Nobs

-    colors <- rainbow(Nfactor)

-    ylimits <- c(0,(max(Pmean + Psd)*1.05))

+    colors <- rainbow(nrow(Pmean))

+    xlimits <- c(0, ncol(Pmean) + 1)

+    ylimits <- c(0, (max(Pmean) * 1.05))

-    plotCI(x=RowP, y=Pmean[1,], col=colors[1],uiw=Psd[1,],

-        ylim=ylimits,type='l',ylab="Relative Amplitude")

+    plot(NULL, xlim=xlimits, ylim=ylimits, ylab="Relative Amplitude")

-    for (i in 2:Nfactor)

+    for (i in 1:nrow(Pmean))

     {

-        points(RowP, Pmean[i,], col=colors[i], pch=i)

-        plotCI(RowP, y=Pmean[i,], col=colors[i], uiw=Psd[i,],

-            add=TRUE)

+        lines(x=1:ncol(Pmean), y=Pmean[i,], col=colors[i])

+        points(x=1:ncol(Pmean), y=Pmean[i,], col=colors[i], pch=i)

     }

+    

+    legend("bottom", paste("Pattern", 1:nrow(Pmean), sep = ""),

+        pch = 1:nrow(Pmean), lty=1, cex=0.8, col=colors, bty="y", ncol=5)

-    legend("bottom", paste("Pattern", 1:Nfactor, sep = ""),

-        pch = 1:Nfactor, lty=1,cex=0.8, col=colors,bty="y",ncol=5)

+    #Nfactor <- nrow(Pmean)

+    #Nobs <- ncol(Pmean)

+    #RowP <- 1:Nobs

+    #colors <- rainbow(Nfactor)

+    #ylimits <- c(0,(max(Pmean + Psd)*1.05))

+    #

+    #plotCI(x=RowP, y=Pmean[1,], col=colors[1], uiw=Psd[1,],

+    #    ylim=ylimits, type='l', ylab="Relative Amplitude")

+    #

+    #for (i in 2:Nfactor)

+    #{

+    #    #points(RowP, Pmean[i,], col=colors[i], pch=i)

+    #    plotCI(RowP, y=Pmean[i,], col=colors[i], uiw=Psd[i,],

+    #        add=TRUE)

+    #}

+    #

+    #legend("bottom", paste("Pattern", 1:Nfactor, sep = ""),

+    #    pch = 1:Nfactor, lty=1,cex=0.8, col=colors,bty="y",ncol=5)

 }

Binary files a/data/GIST_TS_20084.RData and b/data/GIST_TS_20084.RData differ

@@ -4,7 +4,7 @@

 \alias{plotGAPS}

 \title{Plot Decomposed A and P Matrices}

 \usage{

-plotGAPS(A, P, outputPDF = "")

+plotGAPS(A, P)

 }

 \arguments{

 \item{A}{the mean A matrix}

@@ -4,7 +4,7 @@

 \alias{plotP}

 \title{Plot the P Matrix}

 \usage{

-plotP(Pmean, Psd)

+plotP(Pmean, Psd = NULL)

 }

 \arguments{

 \item{Pmean}{matrix of mean values of P}

@@ -3,7 +3,6 @@

 #include "file_parser/CsvParser.h"

 #include <Rcpp.h>

-#include <omp.h>

 // [[Rcpp::export]]

 Rcpp::List cogapsFromFile_cpp(const std::string D)

@@ -12,9 +11,12 @@ Rcpp::List cogapsFromFile_cpp(const std::string D)

     //Rcpp::Rcout << csv.nRow() << "," << csv.nCol() << '\n';

     //RowMatrix mat(D);

-    RowMatrix mat(D);

-    Rcpp::Rcout << mat.nRow() << "," << mat.nCol() << '\n';

-    //Rcpp::Rcout << gaps::algo::sum(mat) << '\n';

+    RowMatrix rmat(D);

+    ColMatrix cmat(D);

+    Rcpp::Rcout << rmat.nRow() << "," << rmat.nCol() << '\n';

+    Rcpp::Rcout << cmat.nRow() << "," << cmat.nCol() << '\n';

+    Rcpp::Rcout << gaps::algo::sum(rmat) << '\n';

+    Rcpp::Rcout << gaps::algo::sum(cmat) << '\n';

 }

 // [[Rcpp::export]]

@@ -27,11 +29,6 @@ const Rcpp::NumericMatrix &FP, unsigned checkpointInterval,

 const std::string &cptFile, unsigned pumpThreshold, unsigned nPumpSamples,

 unsigned nCores)

 {

-    #ifdef __GAPS_OPENMP__

-        unsigned availableCores = omp_get_max_threads();

-        Rprintf("Running on %d out of %d available cores\n", nCores, availableCores);

-    #endif

-

     // create internal state from parameters and run from there

     GapsRunner runner(D, S, nFactor, nEquil, nEquilCool, nSample,

         nOutputs, nSnapshots, alphaA, alphaP, maxGibbmassA, maxGibbmassP, seed,

@@ -49,6 +49,14 @@ mStatistics(D.nrow(), D.ncol(), nFactor)

 // execute the steps of the algorithm, return list to R

 Rcpp::List GapsRunner::run()

 {

+    #ifdef __GAPS_OPENMP__

+        if (mPrintMessages)

+        {

+            unsigned availableCores = omp_get_max_threads();

+            Rprintf("Running on %d out of %d available cores\n", mNumCores, availableCores);

+        }

+    #endif

+

     // reset the checkpoint timer

     mStartTime = bpt_now();

     mLastCheckpoint = mStartTime;

@@ -251,6 +251,7 @@ unsigned col)

         mDomain.updateMass(pos, mass);

         mMatrix(row, col) += mass;

         impl()->updateAPMatrix(row, col, mass);

+        mQueue.acceptBirth();

     }

     else

     {

@@ -1,4 +1,4 @@

-PKG_CPPFLAGS =  -DSIMD -DGAPS_NDEBUG -msse4.2 -DBOOST_MATH_PROMOTE_DOUBLE_POLICY=0

+PKG_CPPFLAGS =  -DSIMD -DGAPS_NDEBUG -msse4.2 -DBOOST_MATH_PROMOTE_DOUBLE_POLICY=0 -fopenmp

 OBJECTS =   AtomicDomain.o \

             Cogaps.o \

             GapsRunner.o \

@@ -26,7 +26,15 @@ void ProposalQueue::setAlpha(float alpha)

 void ProposalQueue::populate(AtomicDomain &domain, unsigned limit)

 {

     unsigned nIter = 0;

-    while (nIter++ < limit && makeProposal(domain));

+    bool success = true;

+    while (nIter++ < limit && success)

+    {

+        success = makeProposal(domain);

+        if (!success)

+        {

+            mUseCachedRng = true;

+        }

+    }

 }

 // TODO efficiently allow clearing a range of proposals

@@ -48,8 +56,7 @@ unsigned ProposalQueue::size() const

 const AtomicProposal& ProposalQueue::operator[](int n) const

 {

-    return mQueue[mQueue.size() - 1 - n];

-    //return mQueue[n];

+    return mQueue[n];

 }

 void ProposalQueue::acceptDeath()

@@ -64,11 +71,14 @@ void ProposalQueue::rejectDeath()

     mMinAtoms++;

 }

-void ProposalQueue::rejectBirth()

+void ProposalQueue::acceptBirth()

 {

     #pragma omp atomic

-    mMinAtoms--;

+    mMinAtoms++;

+}

+void ProposalQueue::rejectBirth()

+{

     #pragma omp atomic

     mMaxAtoms--;

 }

@@ -96,26 +106,29 @@ bool ProposalQueue::makeProposal(AtomicDomain &domain)

     }

     float bdProb = mMaxAtoms < 2 ? 0.6667f : 0.5f;

-    float u1 = gaps::random::uniform(); // cache these values if a failure

+

+    mU1 = mUseCachedRng ? mU1 : gaps::random::uniform();

+    mU2 = mUseCachedRng ? mU2: gaps::random::uniform();

+    mUseCachedRng = false;

+

     float lowerBound = deathProb(mMinAtoms);

     float upperBound = deathProb(mMaxAtoms);

-    if (u1 <= bdProb)

+    if (mU1 <= bdProb)

     {

-        float u2 = gaps::random::uniform(); // happens, needed to prevent bias

-        if (u2 >= upperBound)

+        if (mU2 >= upperBound)

         {

             return birth(domain);

         }

-        if (u2 < lowerBound)

+        if (mU2 < lowerBound)

         {

             return death(domain);

         }

         return false;

     }

-    else if (u1 >= bdProb)

+    else if (mU1 >= bdProb)

     {

-        return (u1 < 0.75f || mMaxAtoms < 2) ? move(domain) : exchange(domain);

+        return (mU1 < 0.75f || mMaxAtoms < 2) ? move(domain) : exchange(domain);

     }

     return false;

 }

@@ -132,7 +145,7 @@ bool ProposalQueue::birth(AtomicDomain &domain)

     mUsedIndices.insert(pos / mDimensionSize);

     mUsedPositions.insert(pos);

     domain.insert(pos, 0.f);

-    ++mMinAtoms;

+    ++mMinAtoms; // TODO could be rejected

     ++mMaxAtoms;

     return true;

 }

@@ -50,6 +50,10 @@ private:

     float mAlpha;

+    bool mUseCachedRng;

+    float mU1;

+    float mU2;

+

     float deathProb(uint64_t nAtoms) const;

     bool birth(AtomicDomain &domain);

     bool death(AtomicDomain &domain);

@@ -61,7 +65,8 @@ private:

 public:

     ProposalQueue(unsigned nBins, float alpha)

-        : mMinAtoms(0), mMaxAtoms(0), mNumBins(nBins), mAlpha(alpha)

+        : mMinAtoms(0), mMaxAtoms(0), mNumBins(nBins), mAlpha(alpha),

+        mUseCachedRng(false), mU1(0.f), mU2(0.f)

     {}

     // set parameters

@@ -79,6 +84,7 @@ public:

     // update min/max atoms

     void acceptDeath();

     void rejectDeath();

+    void acceptBirth();

     void rejectBirth();

     // serialization

@@ -141,11 +141,26 @@ ColMatrix::ColMatrix(const Rcpp::NumericMatrix &rmat)

     }

 }

-// skip over bytes per row to efficiently get to columns

-// need to be able to read columns in parallel

 ColMatrix::ColMatrix(const std::string &path)

 {

+    // get matrix dimensions

+    MatrixDimensions dim(CsvParser::getDimensions(path));

+    mNumRows = dim.nRow;

+    mNumCols = dim.nCol;

+        

+    // allocate matrix

+    for (unsigned i = 0; i < mNumCols; ++i)

+    {

+        mCols.push_back(Vector(mNumRows));

+    }

+    // populate matrix

+    CsvParser csv(path);

+    while (csv.hasNext())

+    {

+        MatrixElement e(csv.getNext());

+        this->operator()(e.row, e.col) = e.value;

+    }

 }

 ColMatrix ColMatrix::operator/(float val) const

@@ -51,6 +51,8 @@ MatrixDimensions CsvParser::getDimensions(const std::string &path)

 CsvParser::CsvParser(const std::string &path) : mCurrentRow(0), mCurrentCol(0)

 {

     mFile.open(path.c_str());

+    std::string line;

+    std::getline(mFile, line); // get rid of first line (column names)

 }

 bool CsvParser::hasNext()

@@ -65,11 +67,13 @@ MatrixElement CsvParser::getNext()

     std::getline(mFile, line, ',');

     if ((pos = line.find('\n')) != std::string::npos) // end of line

     {

-        return MatrixElement(mCurrentRow, mCurrentCol, line.substr(0, pos));

+        unsigned col = mCurrentCol;

+        mCurrentCol = 0;

+        return MatrixElement(mCurrentRow++, col, line.substr(0, pos));

     }

     else if (std::isdigit(line[0])) // data

     {

-        return MatrixElement(mCurrentRow, mCurrentCol, line);

+        return MatrixElement(mCurrentRow, mCurrentCol++, line);

     }

     else // row/col name

     {

@@ -1,5 +1,29 @@

 #ifndef __COGAPS_TSV_PARSER_H__

 #define __COGAPS_TSV_PARSER_H__

+#include "MatrixElement.h"

+

+#include <fstream>

+#include <vector>

+#include <string>

+

+class TsvParser

+{

+private:

+

+    std::ifstream mFile;

+

+    unsigned mCurrentRow;

+    unsigned mCurrentCol;

+

+public:

+

+    TsvParser(const std::string &path);

+

+    bool hasNext();

+    MatrixElement getNext();

+

+    static MatrixDimensions getDimensions(const std::string &path);

+};

 #endif

\ No newline at end of file