Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

ksvm lambda fix #1556

Merged
merged 1 commit into from
Aug 2, 2018
Merged

ksvm lambda fix #1556

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 17 additions & 10 deletions vowpalwabbit/kernel_svm.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -406,7 +406,11 @@ void predict(svm_params& params, svm_example** ec_arr, float* scores, size_t n)
for(size_t i = 0; i < n; i++)
{
ec_arr[i]->compute_kernels(params);
scores[i] = dense_dot(ec_arr[i]->krow.begin(), model->alpha, model->num_support)/params.lambda;
//cout<<"size of krow = "<<ec_arr[i]->krow.size()<<endl;
if(ec_arr[i]->krow.size() > 0)
scores[i] = dense_dot(ec_arr[i]->krow.begin(), model->alpha, model->num_support)/params.lambda;
else
scores[i] = 0;
}
}

Expand Down Expand Up @@ -517,7 +521,7 @@ bool update(svm_params& params, size_t pos)

float proj = alphaKi*ld.label;
float ai = (params.lambda - proj)/inprods[pos];
//params.all->opts_n_args.trace_message<<model->num_support<<" "<<pos<<" "<<proj<<" "<<alphaKi<<" "<<alpha_old<<" "<<ld.label<<" "<<model->delta[pos]<<" " << endl;
//cout<<model->num_support<<" "<<pos<<" "<<proj<<" "<<alphaKi<<" "<<alpha_old<<" "<<ld.label<<" "<<model->delta[pos]<<" " << ai<<" "<<params.lambda<<endl;

if(ai > fec->ex.l.simple.weight)
ai = fec->ex.l.simple.weight;
Expand Down Expand Up @@ -731,13 +735,13 @@ void train(svm_params& params)
if(model_pos >= 0)
{
bool overshoot = update(params, model_pos);
//params.all->opts_n_args.trace_message<<model_pos<<":alpha = "<<model->alpha[model_pos]<<endl;
//cout<<model_pos<<":alpha = "<<model->alpha[model_pos]<<endl;

double* subopt = calloc_or_throw<double>(model->num_support);
for(size_t j = 0; j < params.reprocess; j++)
{
if(model->num_support == 0) break;
//params.all->opts_n_args.trace_message<<"reprocess: ";
//cout<<"reprocess: ";
int randi = 1;
if (merand48(params.all->random_state) < 0.5)
randi = 0;
Expand All @@ -748,8 +752,8 @@ void train(svm_params& params)
{
if(!overshoot && max_pos == (size_t)model_pos && max_pos > 0 && j == 0)
params.all->opts_n_args.trace_message<<"Shouldn't reprocess right after process!!!"<<endl;
//params.all->opts_n_args.trace_message<<max_pos<<" "<<subopt[max_pos]<<endl;
// params.all->opts_n_args.trace_message<<params.model->support_vec[0]->example_counter<<endl;
//cout<<max_pos<<" "<<subopt[max_pos]<<endl;
//cout<<params.model->support_vec[0]->example_counter<<endl;
if(max_pos*model->num_support <= params.maxcache)
make_hot_sv(params, max_pos);
update(params, max_pos);
Expand All @@ -761,8 +765,8 @@ void train(svm_params& params)
update(params, rand_pos);
}
}
//params.all->opts_n_args.trace_message<<endl;
// params.all->opts_n_args.trace_message<<params.model->support_vec[0]->example_counter<<endl;
//cout<<endl;
//cout<<params.model->support_vec[0]->example_counter<<endl;
free(subopt);
}
}
Expand Down Expand Up @@ -792,6 +796,7 @@ void learn(svm_params& params, single_learner&, example& ec)
float score = 0;
predict(params, &sec, &score, 1);
ec.pred.scalar = score;
//cout<<"Score = "<<score<<endl;
ec.loss = max(0.f, 1.f - score*ec.l.simple.label);
params.loss_sum += ec.loss;
if(params.all->training && ec.example_counter % 100 == 0)
Expand Down Expand Up @@ -864,8 +869,8 @@ LEARNER::base_learner* kernel_svm_setup(arguments& arg)
("subsample", params->subsample, (size_t)1, "number of items to subsample from the pool")
.keep("kernel", kernel_type, (string)"linear", "type of kernel (rbf or linear (default))")
.keep("bandwidth", bandwidth, 1.f, "bandwidth of rbf kernel")
.keep("degree", degree, 2, "degree of poly kernel")
.keep("lambda", params->lambda, "saving regularization for test time").missing())
.keep("degree", degree, 2, "degree of poly kernel").missing())
//.keep("lambda", params->lambda, "saving regularization for test time").missing())
return nullptr;

string loss_function = "hinge";
Expand All @@ -891,6 +896,8 @@ LEARNER::base_learner* kernel_svm_setup(arguments& arg)
params->subsample = (size_t)ceil(params->pool_size / arg.all->all_reduce->total);

params->lambda = arg.all->l2_lambda;
if(params->lambda == 0.)
params->lambda = 1.;
params->all->opts_n_args.trace_message<<"Lambda = "<<params->lambda<<endl;
params->all->opts_n_args.trace_message<<"Kernel = "<<kernel_type<<endl;

Expand Down