Algorithm-LibLinear

 view release on metacpan or  search on metacpan

src/liblinear/linear.cpp  view on Meta::CPAN

			int end = fold_start[i+1];

			param_tmp->init_sol = prev_w[i];
			struct model *submodel = train(&subprob[i],param_tmp);

			int total_w_size;
			if(submodel->nr_class == 2)
				total_w_size = subprob[i].n;
			else
				total_w_size = subprob[i].n * submodel->nr_class;

			if(prev_w[i] == NULL)
			{
				prev_w[i] = Malloc(double, total_w_size);
				for(j=0; j<total_w_size; j++)
					prev_w[i][j] = submodel->w[j];
			}
			else if(num_unchanged_w >= 0)
			{
				double norm_w_diff = 0;
				for(j=0; j<total_w_size; j++)
				{
					norm_w_diff += (submodel->w[j] - prev_w[i][j])*(submodel->w[j] - prev_w[i][j]);
					prev_w[i][j] = submodel->w[j];
				}
				norm_w_diff = sqrt(norm_w_diff);

				if(norm_w_diff > 1e-15)
					num_unchanged_w = -1;
			}
			else
			{
				for(j=0; j<total_w_size; j++)
					prev_w[i][j] = submodel->w[j];
			}

			for(j=begin; j<end; j++)
				target[perm[j]] = predict(submodel,prob->x[perm[j]]);

			free_and_destroy_model(&submodel);
		}
		set_print_string_function(default_print_string);

		if(param_tmp->solver_type == L2R_LR || param_tmp->solver_type == L2R_L2LOSS_SVC)
		{
			int total_correct = 0;
			for(i=0; i<prob->l; i++)
				if(target[i] == prob->y[i])
					++total_correct;
			double current_rate = (double)total_correct/prob->l;
			if(current_rate > *best_score)
			{
				*best_C = param_tmp->C;
				*best_score = current_rate;
			}

			info("log2c=%7.2f\trate=%g\n",log(param_tmp->C)/log(2.0),100.0*current_rate);
		}
		else if(param_tmp->solver_type == L2R_L2LOSS_SVR)
		{
			double total_error = 0.0;
			for(i=0; i<prob->l; i++)
			{
				double y = prob->y[i];
				double v = target[i];
				total_error += (v-y)*(v-y);
			}
			double current_error = total_error/prob->l;
			if(current_error < *best_score)
			{
				*best_C = param_tmp->C;
				*best_score = current_error;
			}

			info("log2c=%7.2f\tp=%7.2f\tMean squared error=%g\n",log(param_tmp->C)/log(2.0),param_tmp->p,current_error);
		}

		num_unchanged_w++;
		if(num_unchanged_w == 5)
			break;
		param_tmp->C = param_tmp->C*ratio;
	}

	if(param_tmp->C > max_C)
		info("WARNING: maximum C reached.\n");
	free(target);
	for(i=0; i<nr_fold; i++)
		free(prev_w[i]);
	free(prev_w);
}


//
// Interface functions
//
model* train(const problem *prob, const parameter *param)
{
	int i,j;
	int l = prob->l;
	int n = prob->n;
	int w_size = prob->n;
	model *model_ = Malloc(model,1);

	if(prob->bias>=0)
		model_->nr_feature=n-1;
	else
		model_->nr_feature=n;
	model_->param = *param;
	model_->bias = prob->bias;

	if(check_regression_model(model_))
	{
		model_->w = Malloc(double, w_size);

		if(param->init_sol != NULL)
			for(i=0;i<w_size;i++)
				model_->w[i] = param->init_sol[i];
		else
			for(i=0;i<w_size;i++)
				model_->w[i] = 0;

		model_->nr_class = 2;
		model_->label = NULL;
		train_one(prob, param, model_->w, 0, 0);
	}
	else if(check_oneclass_model(model_))
	{
		model_->w = Malloc(double, w_size);
		model_->nr_class = 2;
		model_->label = NULL;
		solve_oneclass_svm(prob, param, model_->w, &(model_->rho));
	}
	else
	{
		int nr_class;

src/liblinear/linear.cpp  view on Meta::CPAN

};

int save_model(const char *model_file_name, const struct model *model_)
{
	int i;
	int nr_feature=model_->nr_feature;
	int n;
	const parameter& param = model_->param;

	if(model_->bias>=0)
		n=nr_feature+1;
	else
		n=nr_feature;
	int w_size = n;
	FILE *fp = fopen(model_file_name,"w");
	if(fp==NULL) return -1;

	char *old_locale = setlocale(LC_ALL, NULL);
	if (old_locale)
	{
		old_locale = strdup(old_locale);
	}
	setlocale(LC_ALL, "C");

	int nr_w;
	if(model_->nr_class==2 && model_->param.solver_type != MCSVM_CS)
		nr_w=1;
	else
		nr_w=model_->nr_class;

	fprintf(fp, "solver_type %s\n", solver_type_table[param.solver_type]);
	fprintf(fp, "nr_class %d\n", model_->nr_class);

	if(model_->label)
	{
		fprintf(fp, "label");
		for(i=0; i<model_->nr_class; i++)
			fprintf(fp, " %d", model_->label[i]);
		fprintf(fp, "\n");
	}

	fprintf(fp, "nr_feature %d\n", nr_feature);

	fprintf(fp, "bias %.17g\n", model_->bias);

	if(check_oneclass_model(model_))
		fprintf(fp, "rho %.17g\n", model_->rho);

	fprintf(fp, "w\n");
	for(i=0; i<w_size; i++)
	{
		int j;
		for(j=0; j<nr_w; j++)
			fprintf(fp, "%.17g ", model_->w[i*nr_w+j]);
		fprintf(fp, "\n");
	}

	setlocale(LC_ALL, old_locale);
	free(old_locale);

	if (ferror(fp) != 0 || fclose(fp) != 0) return -1;
	else return 0;
}

//
// FSCANF helps to handle fscanf failures.
// Its do-while block avoids the ambiguity when
// if (...)
//    FSCANF();
// is used
//
#define FSCANF(_stream, _format, _var)do\
{\
	if (fscanf(_stream, _format, _var) != 1)\
	{\
		fprintf(stderr, "ERROR: fscanf failed to read the model\n");\
		EXIT_LOAD_MODEL()\
	}\
}while(0)
// EXIT_LOAD_MODEL should NOT end with a semicolon.
#define EXIT_LOAD_MODEL()\
{\
	setlocale(LC_ALL, old_locale);\
	free(model_->label);\
	free(model_);\
	free(old_locale);\
	return NULL;\
}
struct model *load_model(const char *model_file_name)
{
	FILE *fp = fopen(model_file_name,"r");
	if(fp==NULL) return NULL;

	int i;
	int nr_feature;
	int n;
	int nr_class;
	double bias;
	double rho;
	model *model_ = Malloc(model,1);
	parameter& param = model_->param;
	// parameters for training only won't be assigned, but arrays are assigned as NULL for safety
	param.nr_weight = 0;
	param.weight_label = NULL;
	param.weight = NULL;
	param.init_sol = NULL;

	model_->label = NULL;

	char *old_locale = setlocale(LC_ALL, NULL);
	if (old_locale)
	{
		old_locale = strdup(old_locale);
	}
	setlocale(LC_ALL, "C");

	char cmd[81];
	while(1)
	{
		FSCANF(fp,"%80s",cmd);
		if(strcmp(cmd,"solver_type")==0)

src/liblinear/linear.cpp  view on Meta::CPAN

		{
			FSCANF(fp,"%d",&nr_class);
			model_->nr_class=nr_class;
		}
		else if(strcmp(cmd,"nr_feature")==0)
		{
			FSCANF(fp,"%d",&nr_feature);
			model_->nr_feature=nr_feature;
		}
		else if(strcmp(cmd,"bias")==0)
		{
			FSCANF(fp,"%lf",&bias);
			model_->bias=bias;
		}
		else if(strcmp(cmd,"rho")==0)
		{
			FSCANF(fp,"%lf",&rho);
			model_->rho=rho;
		}
		else if(strcmp(cmd,"w")==0)
		{
			break;
		}
		else if(strcmp(cmd,"label")==0)
		{
			int nr_class = model_->nr_class;
			model_->label = Malloc(int,nr_class);
			for(int i=0;i<nr_class;i++)
				FSCANF(fp,"%d",&model_->label[i]);
		}
		else
		{
			fprintf(stderr,"unknown text in model file: [%s]\n",cmd);
			EXIT_LOAD_MODEL()
		}
	}

	nr_feature=model_->nr_feature;
	if(model_->bias>=0)
		n=nr_feature+1;
	else
		n=nr_feature;
	int w_size = n;
	int nr_w;
	if(nr_class==2 && param.solver_type != MCSVM_CS)
		nr_w = 1;
	else
		nr_w = nr_class;

	model_->w=Malloc(double, w_size*nr_w);
	for(i=0; i<w_size; i++)
	{
		int j;
		for(j=0; j<nr_w; j++)
			FSCANF(fp, "%lf ", &model_->w[i*nr_w+j]);
	}

	setlocale(LC_ALL, old_locale);
	free(old_locale);

	if (ferror(fp) != 0 || fclose(fp) != 0) return NULL;

	return model_;
}

int get_nr_feature(const model *model_)
{
	return model_->nr_feature;
}

int get_nr_class(const model *model_)
{
	return model_->nr_class;
}

void get_labels(const model *model_, int* label)
{
	if (model_->label != NULL)
		for(int i=0;i<model_->nr_class;i++)
			label[i] = model_->label[i];
}

// use inline here for better performance (around 20% faster than the non-inline one)
static inline double get_w_value(const struct model *model_, int idx, int label_idx)
{
	int nr_class = model_->nr_class;
	int solver_type = model_->param.solver_type;
	const double *w = model_->w;

	if(idx < 0 || idx > model_->nr_feature)
		return 0;
	if(check_regression_model(model_) || check_oneclass_model(model_))
		return w[idx];
	else
	{
		if(label_idx < 0 || label_idx >= nr_class)
			return 0;
		if(nr_class == 2 && solver_type != MCSVM_CS)
		{
			if(label_idx == 0)
				return w[idx];
			else
				return -w[idx];
		}
		else
			return w[idx*nr_class+label_idx];
	}
}

// feat_idx: starting from 1 to nr_feature
// label_idx: starting from 0 to nr_class-1 for classification models;
//            for regression and one-class SVM models, label_idx is
//            ignored.
double get_decfun_coef(const struct model *model_, int feat_idx, int label_idx)
{
	if(feat_idx > model_->nr_feature)
		return 0;
	return get_w_value(model_, feat_idx-1, label_idx);
}

double get_decfun_bias(const struct model *model_, int label_idx)