int main(int argc, char* argv[])
{
int num_haplo;
double curr_loglik, prev_loglik, delta;
struct tree *tree1, *tree2;
struct haplotype **results;
/* check for too few arguments */
if(argc != 6) {
printf("Error: wrong number of arguments - \"BEAGLEboth <input file 1> <input file 2> <output file for> <output file rev> <output file both>\".\n");
return 1;
}
tree1 = read_input(argv[1]);
tree2 = read_input(argv[2]);
prep_tree(tree1, tree2);
prep_tree(tree2, tree1);
print_tree(argv[3], tree1, 0);
print_tree(argv[4], tree2, 1);
prev_loglik = loglik(tree1);
do {
calc_tree(tree1, tree2);
calc_tree(tree2, tree1);
curr_loglik = loglik(tree1);
delta = (curr_loglik-prev_loglik)/curr_loglik;
prev_loglik = curr_loglik;
} while(delta > TOL);
calc_tree(tree1, tree2);
calc_tree(tree2, tree1);
print_tree(argv[5], tree1, 0);
return 0;
}
main(int argc, char *argv[])
{int *permC, *permV, *permV2;
int loopz, loopa, loopc, cluster,loopvertexk,loopvertexj,j,i,k,Current,start,outloopc, newcluster;
int V,vertexk,vertexj,cl,errors,numlinks;
float meanfield0,meanfield1,logfn1,logfn0,loglik1,loglik0;
float tmp;
static float zma[MAXV][MAXC],zm[MAXV][MAXC], zmat[MAXV][MAXC];
static float zmatmp[MAXV][MAXC],zmtmp[MAXV][MAXC], matmp[MAXC];
static float ma[MAXC], atmp[MAXC], manew[MAXC], floatval[MAXV*MAXC];
int c,cc;
static bool A[MAXV][MAXV], AA[MAXV][MAXV];
float T1, T0, logprior0, logprior1, logpost0, logpost1;
char LinksInputFile[40], tmps[4];
int C=atoi(argv[4]);
int Zloops=atoi(argv[5]);
int Aloops=atoi(argv[6]);
float beta=atof(argv[7]);
float Abeta=atof(argv[8]);
float Bbeta=atof(argv[9]);
// float Threshold=atof(argv[10]);
float Threshold=THRESHOLD;
int Vtoupdate=atoi(argv[10]);
bool DISP_ERRORS=atoi(argv[11]);
int BurninSteps=atoi(argv[12]);
float BurninBeta=atof(argv[13]);
// char OutputFile=argv[2];
/* get the adjacency matrixc */
FILE *fpLinkFile, *fpOutputFile, *fpInitialFile;
int counter, vertex;
int ii[MAXLINKS],jj[MAXLINKS];
float fl;
srand(time(0));
bool dontstart=0;
bool randominit=1;
bool singleinit=0;
fpLinkFile = fopen(argv[1], "r");
if(fpLinkFile==NULL){
printf("Error: can't open LinkFile %s\n",argv[1]);
dontstart=1;}
if (strcmp(argv[3],"random")==0)
{singleinit=0; randominit=1;printf("\nRandom initialisation.\n");}
if (strcmp(argv[3],"single")==0)
{singleinit=1; randominit=1;printf("\nReplicated single cluster initialisation.\n");}
if (randominit==0)
{fpInitialFile = fopen(argv[3], "r");
if(fpInitialFile==NULL){
printf("Error: can't open InitialFile %s\n",argv[3]);
dontstart=1;
}}
if (dontstart==1){return 1;}
else
/* start the analysis : */
{
counter=0;
while(!feof(fpLinkFile)){
fscanf(fpLinkFile, "%d %d", &ii[counter], &jj[counter]);
counter++;
}
fclose(fpLinkFile);
numlinks=counter;
printf("LinkFile read successfully.\n");
/* find the number of vertices : */
int nums[2];
nums[0]=max_value(ii,numlinks);
nums[1]=max_value(jj,numlinks);
V = max_value(nums,2);
printf("Number of vertices =%d\n",V);
if (Vtoupdate==0){Vtoupdate=V;}
printf("Will update at each iteration %d randomly selected vertices\n",Vtoupdate);
if (randominit==0){
cluster=0;
while(!feof(fpInitialFile)){
for (vertex=0;vertex<V;vertex++){
fscanf(fpInitialFile, "%f ",&zm[vertex][cluster]);
}
++cluster;
}
fclose(fpInitialFile);
printf("InitFile read successfully.\n");
if (C==0){C=cluster;}}
/* setup mean alpha */
for (i=0;i<C;i++){ma[i]=1.0;}
/* initialise Z */
for (i=0;i<C;i++)
{ for (j=0;j<V;j++)
{if (randominit==1){zm[j][i]=1.0*(rand()>0.5*RAND_MAX);} /* initial to 0.05 not important */
zma[j][i]=ma[i]*zm[j][i];
}
}
/* setup adj matrix based on links */
for (i=0;i<V;i++)
{ for (j=0;j<i;j++)
{
A[j][i]=0; A[i][j]=0;
}
A[i][i]=1;
}
for (counter=0;counter<numlinks;counter++)
{A[ii[counter]-1][jj[counter]-1]=1;A[jj[counter]-1][ii[counter]-1]=1;}
if (singleinit==1)
{printf("\nFinding first a single cluster based on random initialisation....\n");
updateZ(argv[2],zm,zma,ma,A,BurninSteps,1,V,V,BurninBeta,Threshold,DISP_ERRORS);
for (i=1;i<C;i++)
{ for (j=0;j<V;j++)
{zm[j][i]=zm[k][0];zma[j][i]=ma[i]*zm[j][i];
}
}
}
printf("\n\nFinding the clique matrix marginals p(z[vertex][cluster]=1|adjacency matrix):\n");
if (BurninSteps>0){updateZ(argv[2],zm,zma,ma,A,BurninSteps,C,V,Vtoupdate,BurninBeta,Threshold,DISP_ERRORS);}
for (loopa=0;loopa<=Aloops;loopa++)
{if (loopa==0) {updateZ(argv[2],zm,zma,ma,A,Zloops,C,V,Vtoupdate,beta,Threshold,DISP_ERRORS);}
else
{updateZ(argv[2],zm,zma,ma,A,Zloops,C,V,Vtoupdate,beta,Threshold,DISP_ERRORS);}
if (loopa<Aloops){
/* update alpha */
printf("\n[%d]Updating number of clusters",loopa);
for (i=0;i<V;i++){
for (cluster=0;cluster<C;cluster++){zmat[i][cluster]=zma[i][cluster];}}
for (c=0;c<C;c++){manew[c]=0.0;}
for (cluster=0;cluster<C;cluster++)
{
for (i=0;i<V;i++){zmat[i][cluster]=zm[i][cluster];}
loglik1=loglik(zmat,A,V,C,beta,Threshold);
for (i=0;i<V;i++){zmat[i][cluster]=0.0;}
loglik0=loglik(zmat,A,V,C,beta,Threshold);
for (i=0;i<V;i++){zmat[i][cluster]=zma[i][cluster];}
for (c=0;c<C;c++){atmp[c]=ma[c];}
atmp[cluster]=1.0;
T1=0.0;
for (c=0;c<C;c++){T1+=atmp[c];}
T0=1.0*C-T1;
logprior1=betalog(Abeta+T1,Bbeta+T0);
atmp[cluster]=0.0;
T1=0.0;
for (c=0;c<C;c++){T1+=atmp[c];}
T0=1.0*C-T1;
logprior0=betalog(Abeta+T1,Bbeta+T0);
logpost1=logprior1+loglik1;
logpost0=logprior0+loglik0;
ma[cluster]=1.0/(1.0+exp(logpost0-logpost1));
}
cluster=0; newcluster=0;
for (cluster=0;cluster<C;cluster++){
matmp[cluster]=ma[cluster];
for (i=0;i<V;i++){
zmatmp[i][cluster]=zma[i][cluster];
zmtmp[i][cluster]=zm[i][cluster];
}
}
for (cluster=0;cluster<C;cluster++){
if (ma[cluster]>SMALLVALUE){
for (i=0;i<V;i++){
zma[i][newcluster]=zmatmp[i][cluster];
zm[i][newcluster]=zmtmp[i][cluster];
}
ma[newcluster]=matmp[cluster];
newcluster++;
}
}
C=newcluster;
}
}
}
}
