Skip to content

Commit

Permalink
Reformatting using astyle
Browse files Browse the repository at this point in the history
  • Loading branch information
@pmelsted
pmelsted committed Feb 6, 2015
1 parent 7c03b47 commit 70a6dd5
Show file tree
Hide file tree
Showing 18 changed files with 1,485 additions and 1,510 deletions.
1 change: 1 addition & 0 deletions .gitignore
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,3 +14,4 @@ src/kallisto
src/*.a
CTestTestfile.cmake
*~
*.orig
8 changes: 8 additions & 0 deletions astyle.txt
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,8 @@
close-templates
add-brackets
align-reference=type
align-pointer=name
style=google
indent=spaces=2
keep-one-line-blocks
keep-one-line-statements
264 changes: 132 additions & 132 deletions src/EMAlgorithm.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,153 +14,153 @@ const double TOLERANCE = 1e-5;
template <typename Index>
struct EMAlgorithm {

// counts is vector from collector, with indices corresponding to ec ids
// TODO: initialize alpha a bit more intelligently
// TODO: refactor to remove dependence on Index
EMAlgorithm(const ProgramOptions& opt, const Index& idx,
const std::vector<int>& counts,
const std::vector<double>& eff_lens,
const WeightMap& wm) :
idx_(idx),
num_trans_(idx.num_trans),
counts_(counts),
eff_lens_(eff_lens),
weight_map_(wm),
alpha_(idx.num_trans, 1.0/idx.num_trans), // uniform distribution over transcripts
rho_(idx.num_trans, 0.0),
rho_set_(false)
{}

~EMAlgorithm() {}

void run(size_t n_iter = 500) {
std::vector<double> next_alpha(alpha_.size(), 0.0);

assert(weight_map_.size() == counts_.size());

double denom;

std::cout << "[em]\tfishing for the right mixture (. = 50 rounds)" <<
std::endl;

for (auto i = 0; i < n_iter; ++i) {
if (i % 50 == 0) {
std::cout << ".";
if (i % 500 == 0 && i > 0) {
std::cout << std::endl;
}
}

for (auto& ec_kv : idx_.ecmap ) {
denom = 0.0;

// first, compute the denominator: a normalizer
// iterate over transcripts in EC map
auto w_search = weight_map_.find(ec_kv.first);

// everything in ecmap should be in weight_map
assert( w_search != weight_map_.end() );
assert( w_search->second.size() == ec_kv.second.size() );

for (auto t_it = 0; t_it < ec_kv.second.size(); ++t_it) {
denom += alpha_[ec_kv.second[t_it]] * w_search->second[t_it];
}

if (denom < TOLERANCE) {
continue;
}

/* std::cout << "denom: " << denom << std::endl; */

// compute the update step
for (auto t_it = 0; t_it < ec_kv.second.size(); ++t_it) {
next_alpha[ec_kv.second[t_it]] += counts_[ec_kv.first] *
(w_search->second[t_it] * alpha_[ec_kv.second[t_it]] / denom);
}
}

// TODO: check for relative difference for convergence in EM

// reassign alpha_ to next_alpha
std::copy(next_alpha.begin(), next_alpha.end(), alpha_.begin());

// clear all next_alpha values 0 for next iteration
std::fill(next_alpha.begin(), next_alpha.end(), 0.0);
}

std::cout << std::endl;
std::cout.flush();
}

void compute_rho() {
if (rho_set_) {
// rho has already been set, let's clear it
std::fill(rho_.begin(), rho_.end(), 0.0);
// counts is vector from collector, with indices corresponding to ec ids
// TODO: initialize alpha a bit more intelligently
// TODO: refactor to remove dependence on Index
EMAlgorithm(const ProgramOptions& opt, const Index& idx,
const std::vector<int>& counts,
const std::vector<double>& eff_lens,
const WeightMap& wm) :
idx_(idx),
num_trans_(idx.num_trans),
counts_(counts),
eff_lens_(eff_lens),
weight_map_(wm),
alpha_(idx.num_trans, 1.0/idx.num_trans), // uniform distribution over transcripts
rho_(idx.num_trans, 0.0),
rho_set_(false)
{}

~EMAlgorithm() {}

void run(size_t n_iter = 500) {
std::vector<double> next_alpha(alpha_.size(), 0.0);

assert(weight_map_.size() == counts_.size());

double denom;

std::cout << "[em]\tfishing for the right mixture (. = 50 rounds)" <<
std::endl;

for (auto i = 0; i < n_iter; ++i) {
if (i % 50 == 0) {
std::cout << ".";
if (i % 500 == 0 && i > 0) {
std::cout << std::endl;
}
}

double total {0.0};
for (auto i = 0; i < alpha_.size(); ++i) {
// TODO: consider what the right tolerance is
if (eff_lens_[i] < TOLERANCE) {
continue;
}
rho_[i] = alpha_[i] / eff_lens_[i];
total += rho_[i];
for (auto& ec_kv : idx_.ecmap ) {
denom = 0.0;

// first, compute the denominator: a normalizer
// iterate over transcripts in EC map
auto w_search = weight_map_.find(ec_kv.first);

// everything in ecmap should be in weight_map
assert( w_search != weight_map_.end() );
assert( w_search->second.size() == ec_kv.second.size() );

for (auto t_it = 0; t_it < ec_kv.second.size(); ++t_it) {
denom += alpha_[ec_kv.second[t_it]] * w_search->second[t_it];
}

if (denom < TOLERANCE) {
continue;
}

for (auto& r : rho_) {
r /= total;
/* std::cout << "denom: " << denom << std::endl; */

// compute the update step
for (auto t_it = 0; t_it < ec_kv.second.size(); ++t_it) {
next_alpha[ec_kv.second[t_it]] += counts_[ec_kv.first] *
(w_search->second[t_it] * alpha_[ec_kv.second[t_it]] / denom);
}
}

// TODO: check for relative difference for convergence in EM

// reassign alpha_ to next_alpha
std::copy(next_alpha.begin(), next_alpha.end(), alpha_.begin());

rho_set_ = true;
// clear all next_alpha values 0 for next iteration
std::fill(next_alpha.begin(), next_alpha.end(), 0.0);
}

void write(const std::string& dir_out) const {
const std::string out_fname = "/expression.txt";
std::cout << std::endl;
std::cout.flush();
}

std::ofstream out;
out.open(dir_out + out_fname, std::ios::out);
void compute_rho() {
if (rho_set_) {
// rho has already been set, let's clear it
std::fill(rho_.begin(), rho_.end(), 0.0);
}

if (!out.is_open()) {
std::cerr << "Error opening '" << dir_out + out_fname << "'" <<
std::endl;
exit(1);
}
double total {0.0};
for (auto i = 0; i < alpha_.size(); ++i) {
// TODO: consider what the right tolerance is
if (eff_lens_[i] < TOLERANCE) {
continue;
}
rho_[i] = alpha_[i] / eff_lens_[i];
total += rho_[i];
}

for (auto& r : rho_) {
r /= total;
}

rho_set_ = true;
}

out.precision(15);

out <<
"target_id" << "\t" <<
"kallisto_id" << "\t" <<
"rho" << "\t" <<
"tpm" << "\t" <<
"est_counts" <<
std::endl;

const double MILLION = 1e6;

for (auto i = 0; i < rho_.size(); ++i) {
out <<
idx_.target_names_[i] << "\t" <<
i << "\t" <<
rho_[i] << "\t" <<
rho_[i] * MILLION << "\t" <<
alpha_[i] <<
void write(const std::string& dir_out) const {
const std::string out_fname = "/expression.txt";

std::ofstream out;
out.open(dir_out + out_fname, std::ios::out);

if (!out.is_open()) {
std::cerr << "Error opening '" << dir_out + out_fname << "'" <<
std::endl;
}
exit(1);
}

out.flush();
out.close();
out.precision(15);

out <<
"target_id" << "\t" <<
"kallisto_id" << "\t" <<
"rho" << "\t" <<
"tpm" << "\t" <<
"est_counts" <<
std::endl;

const double MILLION = 1e6;

for (auto i = 0; i < rho_.size(); ++i) {
out <<
idx_.target_names_[i] << "\t" <<
i << "\t" <<
rho_[i] << "\t" <<
rho_[i] * MILLION << "\t" <<
alpha_[i] <<
std::endl;
}

int num_trans_;
const Index &idx_;
const std::vector<int>& counts_;
const std::vector<double>& eff_lens_;
const WeightMap& weight_map_;
std::vector<double> alpha_;
std::vector<double> rho_;
bool rho_set_;
out.flush();
out.close();
}

int num_trans_;
const Index& idx_;
const std::vector<int>& counts_;
const std::vector<double>& eff_lens_;
const WeightMap& weight_map_;
std::vector<double> alpha_;
std::vector<double> rho_;
bool rho_set_;
};

#endif // KALLISTO_EMALGORITHM_H
Loading

0 comments on commit 70a6dd5

Please sign in to comment.