demo/boldbasedSentenceMap.R

In stnava/RKRNS: ANTsR Decoding

print("#################correlation of sentences based on bold#################")
#
# agg<-aggregate( featspace , list(Sent=fspacenames), mean )
# aggcor<-cor(t(data.matrix(agg[,2:ncol(agg)])))
# colnames(aggcor)<-rownames(aggcor)<-dmatsnames
# pdf("fspace_corr.pdf",width=32,height=32)
# pheatmap(aggcor,fontsize = 10)
# dev.off()
# pdf("fspace_corr.pdf",width=4096,height=4096)
# pheatmap(cor(featspace))
# dev.off()
#
if ( FALSE ) {
# dictionaries of all bold responses 
eanat1<-sparseDecom( featspace[ grep("red",rownames(featspace)),]    , sparseness=-0.9, nvecs=10, its=5, mycoption=1 )
eanat2<-sparseDecom( featspace[ grep("artist",rownames(featspace)),] , sparseness=-0.99, nvecs=10, its=5, mycoption=1 )
max(abs(cor(sentspace[grep("red",rownames(featspace)),],eanat1$projections[,])))
max(abs(cor(sentspace[grep("artist",rownames(featspace)),],eanat2$projections[,])))
artcor<-cor(sentspace[grep("artist",rownames(featspace)),],eanat2$projections[,])
rownames(artcor)<-paste("ESent",1:eigsentbasislength,sep='')
pheatmap(artcor)
eanatmat<- featspace[grep("artist",rownames(featspace)),] %*% as.matrix(eanat2$eig)
sentmat<-sentspace[grep("artist",rownames(featspace)),]
summary(lm(sentmat~eanatmat))
nperm<-50
mysparse<-c( 0.1 , -0.5 )
myrob<-0
simccamats<-list( eanatmat, sentmat )
simspacecca<-sparseDecom2( inmatrix=simccamats, nvecs=10, sparseness=mysparse, its=50, mycoption=1, inmask=c(NA,NA ), cthresh=c(0,10), ell1=-11 , perms=nperm, robust=0 )  # subaal

par(mfrow=c(1,1))
for ( i in 1:10 ) 
    for ( j in 4:4 ) {
        mm1<-matrix(eanat1$eig[,j],nrow=responselength)
        plot(mm1[,1],type='l');
        }
for ( i in 1:10 ) 
    for ( j in 4:6 ) {
        mm2<-matrix(eanat2$eig[,j],nrow=responselength)
        plot(mm2[,i],type='l',col='red');
        }


# dictionaries of sentence specific bold responses 
# these go into cca?
##### quick cca on fspace sentspace
nperm<-0
mysparse<-c( 0.2 , 0.5 )
myrob<-0
simccamats<-list( featspace, sentspace )
simspacecca<-sparseDecom2( inmatrix=simccamats, nvecs=4, sparseness=mysparse, its=5, mycoption=1, inmask=c(NA,NA ), cthresh=c(0,10), ell1= 1 , perms=nperm, robust=myrob )  # subaal
}