/*===========================================================================*/
static int parseBackground(xmlDocPtr doc, xmlNodePtr cur, ghmm_xmlfile* f, int modelNo) {
#define CUR_PROC "parseBackground"
int error, order;
int bgNr, rev;
double *b = NULL;
char *s = NULL;
assert(f->modelType & GHMM_kDiscreteHMM);
bgNr = f->model.d[modelNo]->bp->n++;
/* get order */
order = getIntAttribute(cur, "order", &error);
if (error)
order=0;
else if (order && !(f->modelType & GHMM_kHigherOrderEmissions)) {
GHMM_LOG(LERROR, "background distribution has order > 0, but model is not higher order");
goto STOP;
}
f->model.d[modelNo]->bp->order[bgNr] = order;
/* get name */
s = (char *)getXMLCharAttribute(cur, "key", &error);
f->model.d[modelNo]->bp->name[bgNr] = s;
rev = getIntAttribute(cur, "rev", &error);
if (error)
rev = 0;
/* get distribution */
s = (char *)xmlNodeGetContent(cur);
ARRAY_MALLOC(b, pow(f->model.d[modelNo]->bp->m, order+1));
if (-1 != parseCSVList(s, pow(f->model.d[modelNo]->bp->m, order+1), b, rev))
f->model.d[modelNo]->bp->b[bgNr] = b;
else {
GHMM_LOG(LERROR, "Can not parse background CSV list.");
goto STOP;
}
free(s);
return 0;
STOP:
m_free(b);
free(s);
return -1;
#undef CUR_PROC
}
/* ========================================================================= */
static int writeAlphabet(xmlTextWriterPtr writer, ghmm_alphabet * alfa, int type) {
#define CUR_PROC "writeAlphabet"
int i;
if (0 > xmlTextWriterStartElement(writer, BAD_CAST (type == kAlphabet ? "alphabet" : "classAlphabet"))) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement");
goto STOP;;
}
if (type == kAlphabet)
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "id", "%d", alfa->id))
GHMM_LOG_PRINTF(LERROR, LOC, "failed to write id-attribute for alphabet"
"with id %d", alfa->id);
for (i=0; i<alfa->size; i++) {
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "symbol")) {
GHMM_LOG_PRINTF(LERROR, LOC, "failed to start symbol-tag no %d", i);
goto STOP;
}
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "code", "%d", i)) {
GHMM_LOG_PRINTF(LERROR, LOC, "failed to write code-attribute for symbol %s"
"with code %d", alfa->symbols[i], i);
goto STOP;
}
if (0 > xmlTextWriterWriteRaw(writer, BAD_CAST replaceXMLEntity(alfa->symbols[i]))) {
GHMM_LOG_PRINTF(LERROR, LOC, "failed to write symbol %s with code %d",
alfa->symbols[i], i);
goto STOP;
}
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG_PRINTF(LERROR, LOC, "failed to end symbol-tag no %d", i);
goto STOP;
}
}
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at ending alphabet");
goto STOP;
}
return 0;
STOP:
return -1;
#undef CUR_PROC
}
/*============================================================================*/
double ighmm_rand_get_1overa (double x, double mean, double u)
{
/* Calulates 1/a(x, mean, u), with a = the integral from x til \infty over
the Gauss density function */
# define CUR_PROC "ighmm_rand_get_1overa"
double erfc_value;
if (u <= 0.0) {
GHMM_LOG(LCONVERTED, "u <= 0.0 not allowed\n");
goto STOP;
}
#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
erfc_value = erfc ((x - mean) / sqrt (u * 2));
#else
erfc_value = ighmm_erfc ((x - mean) / sqrt (u * 2));
#endif
if (erfc_value <= DBL_MIN) {
ighmm_mes (MES_WIN, "a ~= 0.0 critical! (mue = %.2f, u =%.2f)\n", mean, u);
return (erfc_value);
}
else
return (2.0 / erfc_value);
STOP:
return (-1.0);
# undef CUR_PROC
} /* ighmm_rand_get_1overa */
/* cumalative distribution function of a-truncated N(mean, u) */
double ighmm_rand_normal_right_cdf (double x, double mean, double u, double a)
{
# define CUR_PROC "ighmm_rand_normal_right_cdf"
if (x <= a)
return (0.0);
if (u <= a) {
GHMM_LOG(LCONVERTED, "u <= a not allowed\n");
goto STOP;
}
#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
/*
Function: int erfc (double x, gsl_sf_result * result)
These routines compute the complementary error function
erfc(x) = 1 - erf(x) = 2/\sqrt(\pi) \int_x^\infty \exp(-t^2).
*/
return 1.0 + (erf ((x - mean) / sqrt (u * 2)) -
1.0) / erfc ((a - mean) / sqrt (u * 2));
#else
return 1.0 + (ighmm_erf ((x - mean) / sqrt (u * 2)) -
1.0) / ighmm_erfc ((a - mean) / sqrt (u * 2));
#endif /* Check for ISO C99 */
STOP:
return (-1.0);
# undef CUR_PROC
} /* double ighmm_rand_normal_cdf */
/*===========================================================================*/
static int parseModelType(const char * data, unsigned int size) {
#define CUR_PROC "parseModelType"
int i, noValidMo, modelType=0;
const char * end = data;
char * str;
while ((end = strchr(data, ' '))) {
modelType += matchModelType(data, end-data);
size -= (end-data)+1;
data = end+1;
}
modelType += matchModelType(data, size);
noValidMo = sizeof(validModelTypes)/sizeof(validModelTypes[0]);
for (i=0; i<noValidMo; i++) {
if (modelType == validModelTypes[i])
break;
}
if (i == noValidMo) {
str = ighmm_mprintf(NULL, 0, "%d is no known valid model type", modelType);
GHMM_LOG(LERROR, str);
m_free(str);
return -1;
}
return modelType;
#undef CUR_PROC
}
/* covariance matrix is linearized */
double ighmm_rand_binormal_density(const double *x, double *mean, double *cov)
{
# define CUR_PROC "ighmm_rand_binormal_density"
double rho;
#ifndef DO_WITH_GSL
double numerator,part1,part2,part3;
#endif
if (cov[0] <= 0.0 || cov[2 + 1] <= 0.0) {
GHMM_LOG(LCONVERTED, "variance <= 0.0 not allowed\n");
goto STOP;
}
rho = cov[1] / ( sqrt (cov[0]) * sqrt (cov[2 + 1]) );
/* The denominator is possibly < EPS??? Check that ? */
#ifdef DO_WITH_GSL
/* double gsl_ran_bivariate_gaussian_pdf (double x, double y, double sigma_x,
double sigma_y, double rho) */
return gsl_ran_bivariate_gaussian_pdf (x[0], x[1], sqrt (cov[0]),
sqrt (cov[2 + 1]), rho);
#else
part1 = (x[0] - mean[0]) / sqrt (cov[0]);
part2 = (x[1] - mean[1]) / sqrt (cov[2 + 1]);
part3 = m_sqr (part1) - 2 * part1 * part2 + m_sqr (part2);
numerator = exp ( -1 * (part3) / ( 2 * (1 - m_sqr(rho)) ) );
return (numerator / ( 2 * PI * sqrt(1 - m_sqr(rho)) ));
#endif
STOP:
return (-1.0);
# undef CUR_PROC
} /* double ighmm_rand_binormal_density */
/* ========================================================================= */
static char * doubleArrayToCSV(double * array, int size) {
#define CUR_PROC "doubleArrayToCSV"
int i, pos=0;
char *csv=NULL;
int singlelength = (2 + /* comma and space */
8 + /* 8 signifcant digits */
1 + /* sign */
5 + /* 'E' and signed mantissa */
3); /* safety */
int maxlength = size * singlelength;
ARRAY_MALLOC(csv, maxlength);
for (i=0; i < size-1 && pos + singlelength < maxlength; i++) {
pos += sprintf(csv+pos, "%.8g, ", array[i]);
}
if (i < size-1 || pos + singlelength > maxlength) {
GHMM_LOG(LERROR, "writing CSV failed");
goto STOP;
} else {
pos += sprintf(csv+pos, "%.8g", array[i]);
}
/*printf("%d bytes of %d written\n", pos, maxlength);*/
return csv;
STOP:
free(csv);
return NULL;
#undef CUR_PROC
}
/*============================================================================*/
double ighmm_rand_normal_density_trunc(double x, double mean, double u,
double a)
{
# define CUR_PROC "ighmm_rand_normal_density_trunc"
#ifndef DO_WITH_GSL
double c;
#endif /* DO_WITH_GSL */
if (u <= 0.0) {
GHMM_LOG(LERROR, "u <= 0.0 not allowed");
goto STOP;
}
if (x < a)
return 0.0;
#ifdef DO_WITH_GSL
/* move mean to the right position */
return gsl_ran_gaussian_tail_pdf(x - mean, a - mean, sqrt(u));
#else
if ((c = ighmm_rand_get_1overa(a, mean, u)) == -1) {
GHMM_LOG_QUEUED(LERROR);
goto STOP;
};
return c * ighmm_rand_normal_density(x, mean, u);
#endif /* DO_WITH_GSL */
STOP:
return -1.0;
# undef CUR_PROC
} /* double ighmm_rand_normal_density_trunc */
/* write mean vector and covariance matrix as elements for multinormals */
static int writeMultiNormal(xmlTextWriterPtr writer, ghmm_c_emission *emission)
{
#define CUR_PROC "writeMultiNormal"
char *tmp=NULL;
/* writing mean vector*/
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "mean")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (mean)");
goto STOP;
}
tmp = doubleArrayToCSV(emission->mean.vec, emission->dimension);
if (tmp) {
if (0 > xmlTextWriterWriteRaw(writer, BAD_CAST tmp)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterWriteRaw while writing mean vector CSV");
goto STOP;
}
m_free(tmp);
tmp = NULL;
} else {
GHMM_LOG(LERROR, "converting array to CSV failed for mean vector");
goto STOP;
}
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement mean");
goto STOP;
}
/* writing covariance matrix*/
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "variance")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (variance)");
goto STOP;
}
tmp = doubleArrayToCSV(emission->variance.mat, emission->dimension * emission->dimension);
if (tmp) {
if (0 > xmlTextWriterWriteRaw(writer, BAD_CAST tmp)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterWriteRaw while writing variance matrix CSV");
goto STOP;
}
m_free(tmp);
tmp = NULL;
} else {
GHMM_LOG(LERROR, "converting array to CSV failed for covariance matrix");
goto STOP;
}
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement variance");
goto STOP;
}
return 0;
STOP:
free(tmp);
return -1;
#undef CUR_PROC
}
/*===========================================================================*/
static int parseCSVList(const char * data, unsigned int size, double * array, int reverse) {
#define CUR_PROC "parseCSVList"
int retval=0;
int i;
char * * next, * estr;
double tmp;
ARRAY_CALLOC(next, 1);
for (i=0; i<size; i++) {
array[i] = strtod(data, next);
if (data == *next) {
estr = ighmm_mprintf(NULL, 0, "error in parsing CSV. entry %d of %d. (%s)", i, size, *next);
GHMM_LOG(LERROR, estr);
m_free(estr);
retval=-1;
break;
}
if (next)
data = *next+1;
else
break;
}
if (i != size) {
retval=-1;
estr = ighmm_mprintf(NULL, 0, "error in parsing CSV. sizes do not match (%d != %d)", i, size);
GHMM_LOG(LERROR, estr);
m_free(estr);
}
if (reverse) {
for (i=0; i<size/2; i++) {
tmp = array[i];
array[i] = array[size-i-1];
array[size-i-1] = tmp;
}
}
STOP:
m_free(next);
return retval;
#undef CUR_PROC
}
double ighmm_rand_normal_right (double a, double mue, double u, int seed)
{
# define CUR_PROC "ighmm_rand_normal_right"
double x = -1;
double sigma;
#ifdef DO_WITH_GSL
double s;
#else
double U, Us, Us1, Feps, t, T;
#endif
if (u <= 0.0) {
GHMM_LOG(LCONVERTED, "u <= 0.0 not allowed\n");
goto STOP;
}
sigma = sqrt(u);
if (seed != 0) {
GHMM_RNG_SET (RNG, seed);
}
#ifdef DO_WITH_GSL
/* move boundary to lower values in order to achieve maximum at mue
gsl_ran_gaussian_tail(generator, lower_boundary, sigma)
*/
return mue + gsl_ran_gaussian_tail(RNG, a - mue, sqrt (u));
#else /* DO_WITH_GSL */
/* Inverse transformation with restricted sampling by Fishman */
U = GHMM_RNG_UNIFORM(RNG);
Feps = ighmm_rand_get_PHI((a-mue) / sigma);
Us = Feps + (1-Feps) * U;
Us1 = 1-Us;
t = m_min (Us, Us1);
t = sqrt (-log (t * t));
T =
sigma * (t - (C0 + t * (C1 + t * C2))
/ (1 + t * (D1 + t * (D2 + t * D3))));
if (Us < Us1)
x = mue - T;
else
x = mue + T;
#endif /* DO_WITH_GSL */
STOP:
return x;
# undef CUR_PROC
} /* randvar_normal_pos */
/**
Calculates the logarithm of sum(exp(log_a[j,a_pos])+exp(log_gamma[j,g_pos]))
which corresponds to the logarithm of the sum of a[j,a_pos]*gamma[j,g_pos]
@return ighmm_log_sum for products of a row from gamma and a row from matrix A
@param log_a: row of the transition matrix with logarithmic values (1.0 for log(0))
@param s: ghmm_dstate whose gamma-value is calculated
@param parent: a pointer to the parent hypothesis
*/
static double ighmm_log_gamma_sum (double *log_a, ghmm_dstate * s, hypoList * parent) {
#define CUR_PROC "ighmm_log_gamma_sum"
double result;
int j, j_id, k;
double max = 1.0;
int argmax = 0;
double *logP;
/* shortcut for the trivial case */
if (parent->gamma_states == 1)
for (j = 0; j < s->in_states; j++)
if (parent->gamma_id[0] == s->in_id[j])
return parent->gamma_a[0] + log_a[j];
ARRAY_MALLOC (logP, s->in_states);
/* calculate logs of a[k,l]*gamma[k,hi] as sums of logs and find maximum: */
for (j = 0; j < s->in_states; j++) {
j_id = s->in_id[j];
/* search for state j_id in the gamma list */
for (k = 0; k < parent->gamma_states; k++)
if (parent->gamma_id[k] == j_id)
break;
if (k == parent->gamma_states)
logP[j] = 1.0;
else {
logP[j] = log_a[j] + parent->gamma_a[k];
if (max == 1.0 || (logP[j] > max && logP[j] != 1.0)) {
max = logP[j];
argmax = j;
}
}
}
/* calculate max+log(1+sum[j!=argmax; exp(logP[j]-max)]) */
result = 1.0;
for (j = 0; j < s->in_states; j++)
if (j != argmax && logP[j] != 1.0)
result += exp (logP[j] - max);
result = log (result);
result += max;
free (logP);
return result;
STOP: /* Label STOP from ARRAY_[CM]ALLOC */
GHMM_LOG(LCONVERTED, "ighmm_log_gamma_sum failed\n");
exit (1);
#undef CUR_PROC
}
/*===========================================================================*/
static ghmm_alphabet * parseAlphabet(xmlDocPtr doc, xmlNodePtr cur, ghmm_xmlfile* f) {
#define CUR_PROC "parseAlphabet"
char * str;
int M, code, error;
xmlNodePtr symbol;
ghmm_alphabet * alfa;
ARRAY_CALLOC(alfa, 1);
symbol = cur->children;
M=0;
while (symbol!=NULL) {
if ((!xmlStrcmp(symbol->name, BAD_CAST "symbol"))) {
code = getIntAttribute(symbol, "code", &error);
if (error || code!=M) {
str = ighmm_mprintf(NULL, 0, "non consecutive code %d == %d", code, M);
GHMM_LOG(LERROR, str);
m_free(str);
goto STOP;
} else
M++;
}
symbol=symbol->next;
}
alfa->size = M;
/*printf("Parsing alphabet with %d symbols\n", alfa->size);*/
ARRAY_MALLOC(alfa->symbols, M);
symbol = cur->children;
M=0;
while (symbol!=NULL) {
if ((!xmlStrcmp(symbol->name, BAD_CAST "symbol"))) {
alfa->symbols[M++] = (char *)xmlNodeGetContent(symbol);
/*printf("%d. symbol: %s\n", M, alfa->symbols[M-1]);*/
}
symbol=symbol->next;
}
return alfa;
STOP:
m_free(alfa->symbols);
m_free(alfa)
return NULL;
#undef CUR_PROC
}
/* ========================================================================= */
static char * strModeltype(int modelType) {
#define CUR_PROC "strModeltype"
int end;
char * mt;
ARRAY_CALLOC(mt, 200);
if (modelType > 0) {
if (modelType & GHMM_kLeftRight)
strcat(mt, "left-right ");
if (modelType & GHMM_kSilentStates)
strcat(mt, "silent ");
if (modelType & GHMM_kTiedEmissions)
strcat(mt, "tied ");
if (modelType & GHMM_kHigherOrderEmissions)
strcat(mt, "higher-order ");
if (modelType & GHMM_kBackgroundDistributions)
strcat(mt, "background ");
if (modelType & GHMM_kLabeledStates)
strcat(mt, "labeled ");
if (modelType & GHMM_kTransitionClasses)
strcat(mt, "transition-classes ");
if (modelType & GHMM_kDiscreteHMM)
strcat(mt, "discrete ");
if (modelType & GHMM_kContinuousHMM)
strcat(mt, "continuous ");
if (modelType & GHMM_kPairHMM)
strcat(mt, "pair ");
if (modelType & GHMM_kMultivariate)
strcat(mt, "multivariate ");
} else {
GHMM_LOG(LERROR, "can't write models with unspecified modeltype");
goto STOP;
}
/* overwrite the last space */
end = strlen(mt);
mt[end-1] = '\0';
return mt;
STOP:
m_free(mt);
return NULL;
#undef CUR_PROC
}
/* cumalative distribution function of N(mean, u) */
double ighmm_rand_normal_cdf (double x, double mean, double u)
{
# define CUR_PROC "ighmm_rand_normal_cdf"
if (u <= 0.0) {
GHMM_LOG(LCONVERTED, "u <= 0.0 not allowed\n");
goto STOP;
}
#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
/* PHI(x)=erf(x/sqrt(2))/2+0.5 */
return (erf ((x - mean) / sqrt (u * 2.0)) + 1.0) / 2.0;
#else
return (ighmm_erf ((x - mean) / sqrt (u * 2.0)) + 1.0) / 2.0;
#endif /* Check for ISO C99 */
STOP:
return (-1.0);
# undef CUR_PROC
} /* double ighmm_rand_normal_cdf */
/* cumalative distribution function of a uniform distribution in the range [min,max] */
double ighmm_rand_uniform_cdf (double x, double max, double min)
{
# define CUR_PROC "ighmm_rand_uniform_cdf"
if (max <= min) {
GHMM_LOG(LCONVERTED, "max <= min not allowed\n");
goto STOP;
}
if (x < min) {
return 0.0;
}
if (x >= max) {
return 1.0;
}
return (x-min)/(max-min);
STOP:
return (-1.0);
# undef CUR_PROC
} /* ighmm_rand_uniform_cdf */
/*============================================================================*/
double ighmm_rand_uniform_density (double x, double max, double min)
{
# define CUR_PROC "ighmm_rand_uniform_density"
double prob;
if (max <= min) {
GHMM_LOG(LCONVERTED, "max <= min not allowed \n");
goto STOP;
}
prob = 1.0/(max-min);
if ( (x <= max) && (x>=min) ){
return prob;
}else{
return 0.0;
}
STOP:
return (-1.0);
# undef CUR_PROC
} /* double ighmm_rand_uniform_density */
/* ========================================================================= */
static int writeBackground(xmlTextWriterPtr writer, ghmm_dbackground* bg) {
#define CUR_PROC "writeBackground"
int i;
char * tmp=NULL;
for (i=0; i<bg->n; i++) {
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "background")) {
GHMM_LOG_PRINTF(LERROR, LOC, "Error at starting backgroung %d", i);
return -1;
}
if (!(bg->name)) {
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "key", "bg_%d", i))
GHMM_LOG(LERROR, "Error at writing background key");
}
else {
if (0 > xmlTextWriterWriteAttribute(writer, BAD_CAST "key", BAD_CAST (bg->name[i])))
GHMM_LOG(LERROR, "Error at writing background key");
}
if (0 < bg->order[i])
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "order", "%d", bg->order[i]))
GHMM_LOG(LERROR, "can't write background order attribute");
tmp = doubleArrayToCSV(bg->b[i], pow(bg->m, bg->order[i]+1));
if (tmp) {
if (0 > xmlTextWriterWriteRaw(writer, BAD_CAST tmp)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterWriteRaw while writing"
"background distribution CSV");
m_free(tmp);
return -1;
}
m_free(tmp);
} else {
GHMM_LOG(LERROR, "converting array to CSV failed for background distribution");
return -1;
}
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement while ending"
"background distribution");
return -1;
}
}
return 0;
#undef CUR_PROC
}
/* inserts new hypothesis into list at position indicated by pointer plist */
static void ighmm_hlist_insert (hypoList ** plist, int newhyp,
hypoList * parlist)
{
#define CUR_PROC "ighmm_hlist_insert"
hypoList *newlist;
ARRAY_CALLOC (newlist, 1);
newlist->hyp_c = newhyp;
if (parlist)
parlist->refcount += 1;
newlist->parent = parlist;
newlist->next = *plist;
*plist = newlist;
return;
STOP: /* Label STOP from ARRAY_[CM]ALLOC */
GHMM_LOG(LCONVERTED, "ighmm_hlist_insert failed\n");
exit (1);
#undef CUR_PROC
}
/*===========================================================================*/
double ighmm_rand_uniform_cont (int seed, double max, double min)
{
# define CUR_PROC "ighmm_rand_uniform_cont"
if (max <= min) {
GHMM_LOG(LCONVERTED, "max <= min not allowed\n");
goto STOP;
}
if (seed != 0) {
GHMM_RNG_SET (RNG, seed);
}
#ifdef DO_WITH_GSL
return (double)(((double)gsl_rng_uniform (RNG)*(max-min)) + min);
#else
return (double)((GHMM_RNG_UNIFORM (RNG))*(max-min) + min );
#endif
STOP:
return (-1.0);
# undef CUR_PROC
} /* ighmm_rand_uniform_cont */
/* PROBLEM: Entries can get very small and be rounded to 0 */
int ighmm_cvector_normalize (double *v, int len)
{
#define CUR_PROC "ighmm_cvector_normalize"
int i;
double sum = 0.0;
char * estr;
for (i = 0; i < len; i++)
sum += v[i];
if (i>0 && sum<GHMM_EPS_PREC) {
estr = ighmm_mprintf(NULL, 0, "Can't normalize vector. Sum smaller than %g\n"
, GHMM_EPS_PREC);
GHMM_LOG(LWARN, estr);
m_free(estr);
return (-1);
}
for (i = 0; i < len; i++)
v[i] /= sum;
return 0;
#undef CUR_PROC
} /* ighmm_cvector_normalize */
double ighmm_rand_normal_density_approx (double x, double mean, double u)
{
# define CUR_PROC "ighmm_rand_normal_density_approx"
#ifdef HAVE_LIBPTHREAD
static pthread_mutex_t lock;
#endif /* HAVE_LIBPTHREAD */
int i;
double y, z, pdf_x;
if (u <= 0.0) {
GHMM_LOG(LCONVERTED, "u <= 0.0 not allowed\n");
goto STOP;
}
if (!pdf_stdnormal_exists) {
#ifdef HAVE_LIBPTHREAD
pthread_mutex_lock (&lock); /* Put on a lock, because the clustering is parallel */
#endif /* HAVE_LIBPTHREAD */
randvar_init_pdf_stdnormal ();
#ifdef HAVE_LIBPTHREAD
pthread_mutex_unlock (&lock); /* Take the lock off */
#endif /* HAVE_LIBPTHREAD */
}
y = 1 / sqrt (u);
z = fabs ((x - mean) * y);
i = (int) (z * X_FAKT_PDF);
/* linear interpolation: */
if (i >= PDFLEN - 1) {
i = PDFLEN - 1;
pdf_x = y * pdf_stdnormal[i];
}
else
pdf_x = y * (pdf_stdnormal[i] +
(z - i * X_STEP_PDF) *
(pdf_stdnormal[i + 1] - pdf_stdnormal[i]) / X_STEP_PDF);
return (pdf_x);
STOP:
return (-1.0);
# undef CUR_PROC
} /* double ighmm_rand_normal_density_approx */
/*============================================================================*/
double ighmm_rand_normal_density (double x, double mean, double u)
{
# define CUR_PROC "ighmm_rand_normal_density"
#ifndef DO_WITH_GSL
double expo;
#endif
if (u <= 0.0) {
GHMM_LOG(LCONVERTED, "u <= 0.0 not allowed\n");
goto STOP;
}
/* The denominator is possibly < EPS??? Check that ? */
#ifdef DO_WITH_GSL
/* double gsl_ran_gaussian_pdf (double x, double sigma) */
return gsl_ran_gaussian_pdf (x - mean, sqrt (u));
#else
expo = exp (-1 * m_sqr (mean - x) / (2 * u));
return (1 / (sqrt (2 * PI * u)) * expo);
#endif
STOP:
return (-1.0);
# undef CUR_PROC
} /* double ighmm_rand_normal_density */
/* ========================================================================= */
static int writeContinuousStateContents(xmlTextWriterPtr writer, ghmm_xmlfile* f,
int moNo, int sNo) {
#define CUR_PROC "writeContinuousStateContents"
int i;
ghmm_cstate *state = f->model.c[moNo]->s + sNo;
int allFixed = state->fix;
ghmm_c_emission *emission;
/* writing continuous distribution */
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "mixture")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (mixture)");
goto STOP;
}
if (f->model.c[moNo]->s[sNo].fix)
allFixed = 1;
for(i=0; i < f->model.c[moNo]->s[sNo].M; i++){
emission = f->model.c[moNo]->s[sNo].e+i;
switch (emission->type) {
case normal:
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "normal")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (normal)");
goto STOP;
}
WRITE_DOUBLE_ATTRIBUTE(writer, "mean", emission->mean.val);
WRITE_DOUBLE_ATTRIBUTE(writer, "variance", emission->variance.val);
break;
case multinormal:
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "multinormal")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (multinormal)");
goto STOP;
}
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "dimension",
"%d", emission->dimension)) {
GHMM_LOG(LERROR, "failed to write dimension attribute");
goto STOP;
}
break;
case normal_left:
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "normalLeftTail")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (normalLeftTail)");
goto STOP;
}
WRITE_DOUBLE_ATTRIBUTE(writer, "mean", emission->mean.val);
WRITE_DOUBLE_ATTRIBUTE(writer, "variance", emission->variance.val);
WRITE_DOUBLE_ATTRIBUTE(writer, "max", emission->min);
break;
case normal_right:
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "normalRightTail")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (normalRightTail)");
goto STOP;
}
WRITE_DOUBLE_ATTRIBUTE(writer, "mean", emission->mean.val);
WRITE_DOUBLE_ATTRIBUTE(writer, "variance", emission->variance.val);
WRITE_DOUBLE_ATTRIBUTE(writer, "min", emission->max);
break;
case uniform:
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "uniform")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (uniform)");
goto STOP;
}
WRITE_DOUBLE_ATTRIBUTE(writer, "min", emission->min);
WRITE_DOUBLE_ATTRIBUTE(writer, "max", emission->max);
break;
default:
GHMM_LOG_PRINTF(LERROR, LOC, "invalid density %d at position %d", emission->type, i);
goto STOP;
}
/*optional values */
if (allFixed || emission->fixed) {
if (0 > xmlTextWriterWriteAttribute(writer, BAD_CAST "fixed", BAD_CAST "1")) {
GHMM_LOG(LERROR, "failed to set fixed attribute");
goto STOP;
}
}
if (state->M > 1) {
WRITE_DOUBLE_ATTRIBUTE(writer, "prior", state->c[i]);
}
/* write mean vector and covariance matrix as childs for multinormal */
if (emission->type == multinormal) {
if (0 > writeMultiNormal(writer, emission)) {
GHMM_LOG(LERROR, "failed to write mean and covariance childs");
goto STOP;
}
}
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement (all densities)");
goto STOP;
}
}
/* end mixture tag */
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement (mixture)");
goto STOP;
}
/* writing positions */
if ((state->xPosition > 0) && (state->yPosition > 0)) {
if (xmlTextWriterStartElement(writer, BAD_CAST "position") < 0) {
GHMM_LOG(LERROR, "failed to start position element (position)");
goto STOP;
}
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "x", "%d",
state->xPosition)) {
GHMM_LOG(LERROR, "failed to write x position");
goto STOP;
}
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "y", "%d",
state->yPosition)) {
GHMM_LOG(LERROR, "failed to write y position");
goto STOP;
}
if (xmlTextWriterEndElement(writer) < 0) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement (position)");
goto STOP;
}
}
return 0;
STOP:
return -1;
#undef CUR_PROC
}
/*===========================================================================*/
static int parseState(xmlDocPtr doc, xmlNodePtr cur, ghmm_xmlfile* f, int * inDegree, int * outDegree, int modelNo) {
#define CUR_PROC "parseState"
int i, error, order=0, state=-1442, fixed=-985, tied=-9354, M, aprox, label;
int curX=0, curY=0;
double pi, prior;
double *emissions = NULL;
char *desc = NULL;
char *s = NULL, *estr;
int rev, stateFixed=1;
ghmm_cstate *newcstate;
ghmm_c_emission *emission;
xmlNodePtr elem, child, multichild;
state = getIntAttribute(cur, "id", &error);
pi = getDoubleAttribute(cur, "initial", &error);
if (error) {
estr = ighmm_mprintf(NULL, 0, "can't read required intial probability for"
"state %d", state);
GHMM_LOG(LERROR, estr);
goto STOP;
} else
desc = xmlGetProp(cur, BAD_CAST "desc");
elem = cur->children;
while (elem!=NULL) {
/* ======== silent state ============================================== */
if ((!xmlStrcmp(elem->name, BAD_CAST "silent"))) {
switch (f->modelType & PTR_TYPE_MASK) {
case (GHMM_kDiscreteHMM):
f->model.d[modelNo]->silent[state] = 1;
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
f->model.ds[modelNo]->silent[state] = 1;
break;
case (GHMM_kDiscreteHMM+GHMM_kPairHMM):
case (GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses):
f->model.dp[modelNo]->silent[state] = 1;
break;
default:
GHMM_LOG(LERROR, "invalid modelType");
goto STOP;
}
}
/* ======== discrete state (possible higher order) ==================== */
if ((!xmlStrcmp(elem->name, BAD_CAST "discrete"))) {
assert((f->modelType & GHMM_kDiscreteHMM) && ((f->modelType & GHMM_kPairHMM) == 0));
/* fixed is a propety of the distribution and optional */
fixed = getIntAttribute(elem, "fixed", &error);
if (error)
fixed = 0;
/* order is optional for discrete */
if (f->modelType & GHMM_kHigherOrderEmissions) {
order = getIntAttribute(elem, "order", &error);
if (error)
order = 0;
}
rev = getIntAttribute(cur, "rev", &error);
if (error)
rev = 0;
/* parsing emission probabilities */
s = (char *)xmlNodeGetContent(elem);
switch (f->modelType & PTR_TYPE_MASK) {
case (GHMM_kDiscreteHMM):
f->model.d[modelNo]->s[state].desc = desc;
f->model.d[modelNo]->s[state].pi = pi;
f->model.d[modelNo]->s[state].fix = fixed;
if (f->modelType & GHMM_kHigherOrderEmissions) {
f->model.d[modelNo]->order[state] = order;
if (f->model.d[modelNo]->maxorder < order) {
f->model.d[modelNo]->maxorder = order;
estr = ighmm_mprintf(NULL, 0, "Updated maxorder to %d\n",
f->model.d[modelNo]->maxorder);
GHMM_LOG(LDEBUG, estr);
m_free(estr);
}
}
ARRAY_MALLOC(emissions, pow(f->model.d[modelNo]->M, order+1));
parseCSVList(s, pow(f->model.d[modelNo]->M, order+1), emissions, rev);
free(f->model.d[modelNo]->s[state].b);
f->model.d[modelNo]->s[state].b = emissions;
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
f->model.ds[modelNo]->s[state].desc = desc;
f->model.ds[modelNo]->s[state].pi = pi;
f->model.ds[modelNo]->s[state].fix = fixed;
if (f->modelType & GHMM_kHigherOrderEmissions)
f->model.ds[modelNo]->order[state] = order;
ARRAY_MALLOC(emissions, pow(f->model.ds[modelNo]->M, order+1));
parseCSVList(s, pow(f->model.ds[modelNo]->M, order+1), emissions, rev);
f->model.ds[modelNo]->s[state].b = emissions;
break;
default:
GHMM_LOG(LERROR, "invalid modelType");
goto STOP;
}
m_free(s);
}
/* ======== continuous state ========================================== */
if ((!xmlStrcmp(elem->name, BAD_CAST "mixture"))) {
assert(f->modelType & GHMM_kContinuousHMM);
M = 0;
child = elem->children;
while (child != NULL) {
if ((!xmlStrcmp(child->name, BAD_CAST "normal")) ||
(!xmlStrcmp(child->name, BAD_CAST "normalLeftTail")) ||
(!xmlStrcmp(child->name, BAD_CAST "normalRightTail")) ||
(!xmlStrcmp(child->name, BAD_CAST "multinormal")) ||
(!xmlStrcmp(child->name, BAD_CAST "uniform"))){
M ++;
}
child = child->next;
}
ghmm_cstate_alloc(f->model.c[modelNo]->s + state, M, inDegree[state], outDegree[state], f->model.c[modelNo]->cos);
newcstate = f->model.c[modelNo]->s + state;
newcstate->desc = desc;
newcstate->M = M;
newcstate->pi = pi;
if( f->model.c[modelNo]->M < M)
f->model.c[modelNo]->M = M;
child = elem->children;
i = 0;
while (child != NULL) {
emission = newcstate->e+i;
/* common attributes */
if ((!xmlStrcmp(child->name, BAD_CAST "normal")) ||
(!xmlStrcmp(child->name, BAD_CAST "normalLeftTail")) ||
(!xmlStrcmp(child->name, BAD_CAST "normalRightTail")) ||
(!xmlStrcmp(child->name, BAD_CAST "multinormal")) ||
(!xmlStrcmp(child->name, BAD_CAST "uniform"))){
fixed = getIntAttribute(child, "fixed", &error);
if (error)
fixed = 0;
stateFixed = fixed && stateFixed;
/* allocate emission */
emission->fixed = fixed;
prior = getDoubleAttribute(child, "prior", &error);
if (error)
prior = 1.0;
newcstate->c[i] = prior;
}
/* child is not a density, continue with the next child */
else {
child = child->next;
continue;
}
/* density type dependent attributes */
if ((!xmlStrcmp(child->name, BAD_CAST "normal"))) {
emission->mean.val = getDoubleAttribute(child, "mean", &error);
emission->variance.val = getDoubleAttribute(child, "variance", &error);
/* should the normal distribution be approximated? */
aprox = getIntAttribute(child, "approx", &error);
if (error)
aprox = 0;
emission->type = aprox ? normal_approx : normal;
emission->dimension = 1;
if (f->model.c[modelNo]->dim > 1) {
GHMM_LOG(LERROR, "All emissions must have same dimension.");
goto STOP;
}
}
if ((!xmlStrcmp(child->name, BAD_CAST "normalLeftTail"))) {
emission->mean.val = getDoubleAttribute(child, "mean", &error);
emission->variance.val = getDoubleAttribute(child, "variance", &error);
emission->min = getDoubleAttribute(child, "max", &error);
emission->type = normal_left;
emission->dimension = 1;
if (f->model.c[modelNo]->dim > 1) {
GHMM_LOG(LERROR, "All emissions must have same dimension.");
goto STOP;
}
}
if ((!xmlStrcmp(child->name, BAD_CAST "normalRightTail"))) {
emission->mean.val = getDoubleAttribute(child, "mean", &error);
emission->variance.val = getDoubleAttribute(child, "variance", &error);
emission->max = getDoubleAttribute(child, "min", &error);
emission->type = normal_right;
emission->dimension = 1;
if (f->model.c[modelNo]->dim > 1) {
GHMM_LOG(LERROR, "All emissions must have same dimension.");
goto STOP;
}
}
if ((!xmlStrcmp(child->name, BAD_CAST "uniform"))) {
emission->max = getDoubleAttribute(child, "max", &error);
emission->min = getDoubleAttribute(child, "min", &error);
emission->type = uniform;
emission->dimension = 1;
if (f->model.c[modelNo]->dim > 1) {
GHMM_LOG(LERROR, "All emissions must have same dimension.");
goto STOP;
}
}
if ((!xmlStrcmp(child->name, BAD_CAST "multinormal"))) {
emission->type = multinormal;
emission->dimension = getIntAttribute(child, "dimension", &error);
/* check that all emissions in all states have same dimension or
set when first emission is read*/
if (f->model.c[modelNo]->dim <= 1)
f->model.c[modelNo]->dim = emission->dimension;
else if (f->model.c[modelNo]->dim != emission->dimension) {
GHMM_LOG(LERROR, "All emissions must have same dimension.");
goto STOP;
}
if (0 != ghmm_c_emission_alloc(emission, emission->dimension)) {
GHMM_LOG(LERROR, "Can not allocate multinormal emission.");
goto STOP;
}
multichild = child->children;
while (multichild != NULL) {
if ((!xmlStrcmp(multichild->name, BAD_CAST "mean"))) {
s = (char *)xmlNodeGetContent(multichild);
if (-1 == parseCSVList(s, emission->dimension, emission->mean.vec, 0)) {
GHMM_LOG(LERROR, "Can not parse mean CSV list.");
goto STOP;
}
}
if ((!xmlStrcmp(multichild->name, BAD_CAST "variance"))) {
s = (char *)xmlNodeGetContent(multichild);
if (-1 == parseCSVList(s, emission->dimension * emission->dimension,
emission->variance.mat, 0)) {
GHMM_LOG(LERROR, "Can not parse variance CSV list.");
goto STOP;
}
if (0 != ighmm_invert_det(emission->sigmainv, &emission->det,
emission->dimension, emission->variance.mat))
{
GHMM_LOG(LERROR, "Can not calculate inverse of covariance matrix.");
goto STOP;
}
if (0 != ighmm_cholesky_decomposition(emission->sigmacd,
emission->dimension,
emission->variance.mat))
{
GHMM_LOG(LERROR, "Can not calculate cholesky decomposition of covariance matrix.");
goto STOP;
}
}
multichild = multichild->next;
}
}
i++;
child = child->next;
}
newcstate->fix = stateFixed;
}
/* ======== pair hmm state ============================================ */
if ((!xmlStrcmp(elem->name, BAD_CAST "pair"))) {
}
/* -------- background name ------------------------------------------ */
if ((!xmlStrcmp(elem->name, BAD_CAST "backgroundKey"))) {
assert(f->modelType & GHMM_kBackgroundDistributions);
s = (char *)xmlNodeGetContent(elem);
for (i=0; i<f->model.d[modelNo]->bp->n; i++) {
if (0 == strcmp(s, f->model.d[modelNo]->bp->name[i])) {
if (order != f->model.d[modelNo]->bp->order[i]) {
estr = ighmm_mprintf(NULL, 0, "order of background %s and state %d"
" does not match",
f->model.d[modelNo]->bp->name[i], state);
GHMM_LOG(LERROR, estr);
m_free(estr);
goto STOP;
} else {
f->model.d[modelNo]->background_id[state] = i;
break;
}
}
}
if (i == f->model.d[modelNo]->bp->n) {
estr = ighmm_mprintf(NULL, 0, "can't find background with name %s in"
" state %d", s, state);
GHMM_LOG(LERROR, estr);
m_free(estr);
goto STOP;
}
m_free(s);
}
/* -------- tied to --------------------------------------------------- */
if ((!xmlStrcmp(elem->name, BAD_CAST "class"))) {
assert(f->modelType & GHMM_kLabeledStates);
s = (char *)xmlNodeGetContent(elem);
label = atoi(s);
m_free(s);
if ((f->modelType & PTR_TYPE_MASK) == GHMM_kDiscreteHMM) {
if (f->model.d[modelNo]->label_alphabet->size > label)
f->model.d[modelNo]->label[state] = label;
else
GHMM_LOG(LWARN, "Invalid label");
}
}
/* -------- tied to --------------------------------------------------- */
if ((!xmlStrcmp(elem->name, BAD_CAST "tiedTo"))) {
assert(f->modelType & GHMM_kTiedEmissions);
s = (char *)xmlNodeGetContent(elem);
tied = atoi(s);
if (state>=tied) {
f->model.d[modelNo]->tied_to[state] = tied;
if (f->model.d[modelNo]->tied_to[tied] != tied) {
estr = ighmm_mprintf(NULL, 0, "state %d not tied to tie group leader", state);
GHMM_LOG(LERROR, estr);
m_free(estr);
goto STOP;
}
} else {
estr = ighmm_mprintf(NULL, 0, "state %d tiedTo (%d) is invalid", state, tied);
GHMM_LOG(LERROR, estr);
m_free(estr);
goto STOP;
}
m_free(s);
}
/* -------- position for graphical editing ---------------------------- */
if ((!xmlStrcmp(elem->name, BAD_CAST "position"))) {
curX = getIntAttribute(elem, "x", &error);
if (error)
GHMM_LOG(LWARN, "failed to read x position");
curY = getIntAttribute(elem, "y", &error);
if (error)
GHMM_LOG(LWARN, "failed to read y position");
switch (f->modelType & PTR_TYPE_MASK) {
case GHMM_kDiscreteHMM:
f->model.d[modelNo]->s[state].xPosition = curX;
f->model.d[modelNo]->s[state].yPosition = curY;
break;
case GHMM_kDiscreteHMM+GHMM_kTransitionClasses:
f->model.ds[modelNo]->s[state].xPosition = curX;
f->model.ds[modelNo]->s[state].yPosition = curY;
break;
case GHMM_kDiscreteHMM+GHMM_kPairHMM:
case GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses:
f->model.dp[modelNo]->s[state].xPosition = curX;
f->model.dp[modelNo]->s[state].yPosition = curY;
break;
case GHMM_kContinuousHMM:
case GHMM_kContinuousHMM+GHMM_kTransitionClasses:
case (GHMM_kContinuousHMM+GHMM_kMultivariate):
case (GHMM_kContinuousHMM+GHMM_kMultivariate+GHMM_kTransitionClasses):
f->model.c[modelNo]->s[state].xPosition = curX;
f->model.c[modelNo]->s[state].yPosition = curY;
break;
default:
GHMM_LOG(LERROR, "invalid modelType");
goto STOP;
}
}
elem = elem->next;
}
return 0;
STOP:
m_free(s);
m_free(desc);
m_free(emissions)
return -1;
#undef CUR_PROC
}
/* ========================================================================= */
static int writeState(xmlTextWriterPtr writer, ghmm_xmlfile* f, int moNo, int sNo) {
#define CUR_PROC "writeState"
int rc;
double w_pi;
char *w_desc=NULL;
/* start state */
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "state")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (state)");
goto STOP;
}
/* write id attribute */
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "id", "%d", sNo))
GHMM_LOG(LERROR, "failed to write statte id attribute");
/* read state attribute from different model types */
switch (f->modelType & PTR_TYPE_MASK) {
case GHMM_kDiscreteHMM:
w_pi = f->model.d[moNo]->s[sNo].pi;
w_desc = f->model.d[moNo]->s[sNo].desc;
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
w_pi = f->model.ds[moNo]->s[sNo].pi;
w_desc = f->model.ds[moNo]->s[sNo].desc;
break;
case (GHMM_kDiscreteHMM+GHMM_kPairHMM):
case (GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses):
/*
w_pi = f->model.d[moNo]->s[sNo].pi;
w_desc = f->model.d[moNo]->s[sNo];
*/
break;
case GHMM_kContinuousHMM:
case (GHMM_kContinuousHMM+GHMM_kTransitionClasses):
case (GHMM_kContinuousHMM+GHMM_kMultivariate):
case (GHMM_kContinuousHMM+GHMM_kMultivariate+GHMM_kTransitionClasses):
w_pi = f->model.c[moNo]->s[sNo].pi;
w_desc = f->model.c[moNo]->s[sNo].desc;
break;
default:
GHMM_LOG(LCRITIC, "invalid modelType");}
/* write initial probability as attribute */
WRITE_DOUBLE_ATTRIBUTE(writer, "initial", w_pi);
/* write state description */
if (w_desc) {
if (xmlTextWriterWriteAttribute(writer, BAD_CAST "desc", BAD_CAST replaceXMLEntity(w_desc)))
GHMM_LOG(LERROR, "writing state description failed");
}
/* write state contents for different model types */
switch (f->modelType & PTR_TYPE_MASK) {
case GHMM_kDiscreteHMM:
rc = writeDiscreteStateContents(writer, f, moNo, sNo);
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
rc = writeDiscreteSwitchingStateContents(writer, f, moNo, sNo);
break;
case (GHMM_kDiscreteHMM+GHMM_kPairHMM):
case (GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses):
/*
rc = writeDiscretePairStateContents(writer, f, moNo, sNo);
*/
break;
case GHMM_kContinuousHMM:
case (GHMM_kContinuousHMM+GHMM_kTransitionClasses):
case (GHMM_kContinuousHMM+GHMM_kMultivariate):
case (GHMM_kContinuousHMM+GHMM_kMultivariate+GHMM_kTransitionClasses):
rc = writeContinuousStateContents(writer, f, moNo, sNo);
break;
default:
GHMM_LOG(LCRITIC, "invalid modelType");
goto STOP;
}
if (rc) {
GHMM_LOG_PRINTF(LERROR, LOC, "writing state contents failed. model_type = %s",
strModeltype(f->modelType & PTR_TYPE_MASK));
goto STOP;
}
/* end state*/
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement (state)");
goto STOP;
}
return 0;
STOP:
return -1;
#undef CUR_PROC
}
/* ========================================================================= */
static int writeTransition(xmlTextWriterPtr writer, ghmm_xmlfile* f, int moNo,
int sNo) {
#define CUR_PROC "writeTransition"
int cos, i, j;
int out_states, * out_id;
double * * out_a;
double * w_out_a;
char * tmp;
/* write state contents for different model types */
switch (f->modelType & PTR_TYPE_MASK) {
case GHMM_kDiscreteHMM:
out_states = f->model.d[moNo]->s[sNo].out_states;
out_id = f->model.d[moNo]->s[sNo].out_id;
out_a = &(f->model.d[moNo]->s[sNo].out_a);
cos = 1;
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
out_states = f->model.ds[moNo]->s[sNo].out_states;
out_id = f->model.ds[moNo]->s[sNo].out_id;
out_a = f->model.ds[moNo]->s[sNo].out_a;
cos = f->model.ds[moNo]->cos;
break;
case (GHMM_kDiscreteHMM+GHMM_kPairHMM):
case (GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses):
/*
out_states = f->model.dp[moNo]->s[sNo].out_states;
out_id = f->model.dp[moNo]->s[sNo].out_id;
out_a = f->model.dp[moNo]->s[sNo].out_a;
cos = f->model.dp[moNo]->cos;
*/
break;
case GHMM_kContinuousHMM:
case (GHMM_kContinuousHMM+GHMM_kTransitionClasses):
case (GHMM_kContinuousHMM+GHMM_kMultivariate):
case (GHMM_kContinuousHMM+GHMM_kMultivariate+GHMM_kTransitionClasses):
out_states = f->model.c[moNo]->s[sNo].out_states;
out_id = f->model.c[moNo]->s[sNo].out_id;
out_a = f->model.c[moNo]->s[sNo].out_a;
cos = f->model.c[moNo]->cos;
break;
default:
GHMM_LOG(LCRITIC, "invalid modelType");}
ARRAY_MALLOC(w_out_a, cos);
for (i=0; i<out_states; i++) {
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "transition")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (transition)");
goto STOP;
}
/* write source id (current state attribute */
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "source", "%d", sNo))
GHMM_LOG(LERROR, "failed to write transition source attribute");
/* write target id as attribute */
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "target", "%d", out_id[i]))
GHMM_LOG(LERROR, "failed to write transition target attribute");
for (j=0; j<cos; j++)
w_out_a[j] = out_a[j][i];
tmp = doubleArrayToCSV(w_out_a, cos);
if (tmp) {
if (0 > xmlTextWriterWriteElement(writer, BAD_CAST "probability", BAD_CAST tmp)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterWriteElement (transition probabilities)");
m_free(tmp);
goto STOP;
}
m_free(tmp);
} else {
GHMM_LOG(LERROR, "converting transition probabilities array to CSV failed");
goto STOP;
}
/* end transition */
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement (transition)");
goto STOP;
}
}
return 0;
STOP:
return -1;
#undef CUR_PROC
}
/* ========================================================================= */
static int writeHMM(xmlTextWriterPtr writer, ghmm_xmlfile* f, int number) {
#define CUR_PROC "writeHMM"
int rc=0, i, N;
int w_cos;
double w_prior;
char *w_name;
char * w_type;
/* start HMM */
if (0 > xmlTextWriterStartElement(writer, BAD_CAST "HMM")) {
GHMM_LOG(LERROR, "Error at xmlTextWriterStartElement (HMM)");
goto STOP;;
}
/* write HMM attributes applicable */
switch (f->modelType & PTR_TYPE_MASK) {
case GHMM_kDiscreteHMM:
w_name = f->model.d[number]->name;
w_type = strModeltype(f->model.d[number]->model_type);
w_prior = f->model.d[number]->prior;
N = f->model.d[number]->N;
w_cos = 1;
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
w_name = f->model.ds[number]->name;
w_type = strModeltype(f->model.ds[number]->model_type);
w_prior = f->model.ds[number]->prior;
N = f->model.ds[number]->N;
w_cos = 0;
break;
case (GHMM_kDiscreteHMM+GHMM_kPairHMM):
case (GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses):
/*
w_name = f->model.dp[number]->name;
w_type = strModeltype(f->model.dp[number]->model_type);
w_prior = f->model.dp[number]->prior;
N = f->model.dp[number]->N;
w_cos = 0;
*/
break;
case GHMM_kContinuousHMM:
case (GHMM_kContinuousHMM+GHMM_kMultivariate):
case (GHMM_kContinuousHMM+GHMM_kTransitionClasses):
case (GHMM_kContinuousHMM+GHMM_kMultivariate+GHMM_kTransitionClasses):
w_name = f->model.c[number]->name;
if (f->model.c[number]->model_type)
w_type = strModeltype(f->model.c[number]->model_type);
else
w_type = strModeltype(f->modelType);
w_prior = f->model.c[number]->prior;
N = f->model.c[number]->N;
w_cos = f->model.c[number]->cos;
break;
default:
GHMM_LOG(LERROR, "invalid modelType");
goto STOP;}
if (w_name) {
if (xmlTextWriterWriteAttribute(writer, BAD_CAST "name", w_name))
GHMM_LOG(LERROR, "writing HMM name failed");
}
if (xmlTextWriterWriteAttribute(writer, BAD_CAST "type", BAD_CAST w_type))
GHMM_LOG(LERROR, "writing HMM type failed");
if (w_prior >= 0.0) {
WRITE_DOUBLE_ATTRIBUTE(writer, "prior", w_prior);
}
if (w_cos > 1)
if (0 > xmlTextWriterWriteFormatAttribute(writer, BAD_CAST "transitionClasses",
"%d", w_cos))
GHMM_LOG(LERROR, "failed to write no of transitionClasses");
/* write alphabet if applicable */
switch (f->modelType & (GHMM_kDiscreteHMM + GHMM_kTransitionClasses
+ GHMM_kPairHMM)) {
case GHMM_kDiscreteHMM:
rc = writeAlphabet(writer, f->model.d[number]->alphabet, kAlphabet);
break;
case (GHMM_kDiscreteHMM+GHMM_kTransitionClasses):
/*rc = writeAlphabet(writer, f->model.ds[number]->alphabet, kAlphabet);*/
break;
case (GHMM_kDiscreteHMM+GHMM_kPairHMM):
case (GHMM_kDiscreteHMM+GHMM_kPairHMM+GHMM_kTransitionClasses):
/*rc = writeAlphabet(writer, f->model.dp[number]->alphabets[0], kAlphabet);
if (rc) {
GHMM_LOG(LERROR, "writing first alphabet of discrete pair HMM failed");
goto STOP;
}
rc = writeAlphabet(writer, f->model.dp[number]->alphabets[1], kAlphabet);*/
break;
}
if (rc) {
GHMM_LOG_PRINTF(LERROR, LOC, "writing alphabet for HMM %d (type %s) failed",
number, strModeltype(f->modelType));
goto STOP;
}
/* write label alphabet if applicable */
if ((f->modelType & PTR_TYPE_MASK) == GHMM_kDiscreteHMM
&& f->modelType & GHMM_kLabeledStates) {
if (writeAlphabet(writer, f->model.d[number]->label_alphabet, kLabelAlphabet))
GHMM_LOG(LERROR, "writing of label alphabet failed");
}
/* write background distributions if applicable */
if ((f->modelType & PTR_TYPE_MASK) == GHMM_kDiscreteHMM
&& f->modelType & GHMM_kBackgroundDistributions) {
if (writeBackground(writer, f->model.d[number]->bp))
GHMM_LOG(LERROR, "writing of background distributions failed");
}
/* write all states */
for (i=0; i<N; i++)
if (writeState(writer, f, number, i)) {
GHMM_LOG_PRINTF(LERROR, LOC, "writing of state %d in HMM %d failed", i, number);
goto STOP;
}
/* write all outgoing transitions */
for (i=0; i<N; i++)
if (writeTransition(writer, f, number, i)) {
GHMM_LOG_PRINTF(LERROR, LOC, "writing transitions of state %d in HMM %d failed",
i, number);
goto STOP;
}
/*end HMM*/
if (0 > xmlTextWriterEndElement(writer)) {
GHMM_LOG(LERROR, "Error at xmlTextWriterEndElement (HMM)");
goto STOP;
}
return 0;
STOP:
return -1;
#undef CUR_PROC
}
/**
Trains the ghmm_dmodel with a set of annotated sequences till convergence using
gradient descent.
Model must not have silent states. (checked in Python wrapper)
@return trained model/NULL pointer success/error
@param mo: pointer to a ghmm_dmodel
@param sq: struct of annotated sequences
@param eta: intial parameter eta (learning rate)
@param no_steps number of training steps
*/
ghmm_dmodel* ghmm_dmodel_label_gradient_descent (ghmm_dmodel* mo, ghmm_dseq * sq, double eta, int no_steps)
{
#define CUR_PROC "ghmm_dmodel_label_gradient_descent"
char * str;
int runs = 0;
double cur_perf, last_perf;
ghmm_dmodel *last;
last = ghmm_dmodel_copy(mo);
last_perf = compute_performance (last, sq);
while (eta > GHMM_EPS_PREC && runs < no_steps) {
runs++;
if (-1 == gradient_descent_onestep(mo, sq, eta)) {
ghmm_dmodel_free(&last);
return NULL;
}
cur_perf = compute_performance(mo, sq);
if (last_perf < cur_perf) {
/* if model is degenerated, lower eta and try again */
if (cur_perf > 0.0) {
str = ighmm_mprintf(NULL, 0, "current performance = %g", cur_perf);
GHMM_LOG(LINFO, str);
m_free(str);
ghmm_dmodel_free(&mo);
mo = ghmm_dmodel_copy(last);
eta *= .5;
}
else {
/* Improvement insignificant, assume convergence */
if (fabs (last_perf - cur_perf) < cur_perf * (-1e-8)) {
ghmm_dmodel_free(&last);
str = ighmm_mprintf(NULL, 0, "convergence after %d steps.", runs);
GHMM_LOG(LINFO, str);
m_free(str);
return 0;
}
if (runs < 175 || 0 == runs % 50) {
str = ighmm_mprintf(NULL, 0, "Performance: %g\t improvement: %g\t step %d", cur_perf,
cur_perf - last_perf, runs);
GHMM_LOG(LINFO, str);
m_free(str);
}
/* significant improvement, next iteration */
ghmm_dmodel_free(&last);
last = ghmm_dmodel_copy(mo);
last_perf = cur_perf;
eta *= 1.07;
}
}
/* no improvement */
else {
if (runs < 175 || 0 == runs % 50) {
str = ighmm_mprintf(NULL, 0, "Performance: %g\t !IMPROVEMENT: %g\t step %d", cur_perf,
cur_perf - last_perf, runs);
GHMM_LOG(LINFO, str);
m_free(str);
}
/* try another training step */
runs++;
eta *= .85;
if (-1 == gradient_descent_onestep(mo, sq, eta)) {
ghmm_dmodel_free(&last);
return NULL;
}
cur_perf = compute_performance (mo, sq);
str = ighmm_mprintf(NULL, 0, "Performance: %g\t ?Improvement: %g\t step %d", cur_perf,
cur_perf - last_perf, runs);
GHMM_LOG(LINFO, str);
m_free(str);
/* improvement, save and proceed with next iteration */
if (last_perf < cur_perf && cur_perf < 0.0) {
ghmm_dmodel_free (&last);
last = ghmm_dmodel_copy(mo);
last_perf = cur_perf;
}
/* still no improvement, revert to saved model */
else {
runs--;
ghmm_dmodel_free(&mo);
mo = ghmm_dmodel_copy(last);
eta *= .9;
}
}
}
ghmm_dmodel_free(&last);
return mo;
#undef CUR_PROC
}
/*============================================================================*/
int ghmm_cmodel_logp_joint(ghmm_cmodel *mo, const double *O, int len,
const int *S, int slen, double *log_p)
{
# define CUR_PROC "ghmm_cmodel_logp_joint"
int prevstate, state, state_pos=0, pos=0, j, osc=0;
int dim = mo->dim;
prevstate = state = S[0];
*log_p = log(mo->s[state].pi);
if (!(mo->model_type & GHMM_kSilentStates) || 1 /* XXX !mo->silent[state] */ )
{
*log_p += log(ghmm_cmodel_calc_b(mo->s+state, O+pos));
pos+=dim;
}
for (state_pos=1; state_pos < slen || pos+dim <= len; state_pos++) {
state = S[state_pos];
for (j=0; j < mo->s[state].in_states; ++j) {
if (prevstate == mo->s[state].in_id[j])
break;
}
if (mo->cos > 1) {
if (!mo->class_change->get_class) {
GHMM_LOG(LERROR, "get_class not initialized");
goto STOP;
}
osc = mo->class_change->get_class(mo, O, mo->class_change->k, pos);
if (osc >= mo->cos) {
GHMM_LOG_PRINTF(LERROR, LOC, "get_class returned index %d "
"but model has only %d classes!", osc, mo->cos);
goto STOP;
}
}
if (j == mo->s[state].in_states ||
fabs(mo->s[state].in_a[osc][j]) < GHMM_EPS_PREC) {
GHMM_LOG_PRINTF(LERROR, LOC, "Sequence can't be built. There is no "
"transition from state %d to %d.", prevstate, state);
goto STOP;
}
*log_p += log(mo->s[state].in_a[osc][j]);
if (!(mo->model_type & GHMM_kSilentStates) || 1 /* XXX !mo->silent[state] */) {
*log_p += log(ghmm_cmodel_calc_b(mo->s+state, O+pos));
pos+=dim;
}
prevstate = state;
}
if (pos < len)
GHMM_LOG_PRINTF(LINFO, LOC, "state sequence too short! processed only %d symbols", pos/dim);
if (state_pos < slen)
GHMM_LOG_PRINTF(LINFO, LOC, "sequence too short! visited only %d states", state_pos);
return 0;
STOP:
return -1;
# undef CUR_PROC
}
