create ccv_minimize_default_params

lib/ccv.h

@@ -535,6 +535,8 @@ typedef struct {
 	double sig;
 } ccv_minimize_param_t;
 
+extern const ccv_minimize_param_t ccv_minimize_default_params;
+
 typedef int(*ccv_minimize_f)(const ccv_dense_matrix_t* x, double* f, ccv_dense_matrix_t* df, void*);
 void ccv_minimize(ccv_dense_matrix_t* x, int length, double red, ccv_minimize_f func, ccv_minimize_param_t params, void* data);
 
@@ -1123,8 +1125,6 @@ typedef struct {
 		ccv_size_t base;
 		int range_through;
 		int step_through;
-		int active_set;
-		int wide_set;
 	} feature;
 	struct {
 		float angle;

lib/ccv_numeric.c

@@ -9,6 +9,15 @@
 #include "3rdparty/kissfft/kissf_fftndr.h"
 #endif
 
+const ccv_minimize_param_t ccv_minimize_default_params = {
+	.interp = 0.1,
+	.extrap = 3.0,
+	.max_iter = 20,
+	.ratio = 10.0,
+	.sig = 0.1,
+	.rho = 0.05,
+};
+
 void ccv_invert(ccv_matrix_t* a, ccv_matrix_t** b, int type)
 {
 }

lib/ccv_scd.c

@@ -454,7 +454,7 @@ static int _ccv_scd_feature_logistic_loss(const ccv_dense_matrix_t* x, double* f
 	return 0;
 }
 
-static void _ccv_scd_feature_supervised_train(gsl_rng* rng, ccv_array_t* features, int positive_count, int negative_count, double* pw, double* nw, float* fv, int active_set, int wide_set, double C)
+static void _ccv_scd_feature_supervised_train(gsl_rng* rng, ccv_array_t* features, int positive_count, int negative_count, double* pw, double* nw, float* fv, double C)
 {
 	parallel_for(i, features->rnum) {
 		if ((i + 1) % 31 == 1 || (i + 1) == features->rnum)
@@ -469,19 +469,11 @@ static void _ccv_scd_feature_supervised_train(gsl_rng* rng, ccv_array_t* feature
 			.nw = nw,
 			.fv = fv,
 		};
-		ccv_minimize_param_t params = {
-			.interp = 0.1,
-			.extrap = 3.0,
-			.max_iter = 20,
-			.ratio = 10.0,
-			.sig = 0.1,
-			.rho = 0.05,
-		};
 		ccv_dense_matrix_t* x = ccv_dense_matrix_new(1, 33, CCV_32F | CCV_C1, 0, 0);
 		int j;
 		for (j = 0; j < 33; j++)
 			x->data.f32[j] = gsl_rng_uniform_pos(rng) * 2 - 1.0;
-		ccv_minimize(x, 5, 1.0, _ccv_scd_feature_logistic_loss, params, &context);
+		ccv_minimize(x, 5, 1.0, _ccv_scd_feature_logistic_loss, ccv_minimize_default_params, &context);
 		for (j = 0; j < 32; j++)
 			feature->w[j] = x->data.f32[j];
 		feature->bias = x->data.f32[32];
@@ -882,7 +874,7 @@ ccv_scd_classifier_cascade_t* ccv_scd_classifier_cascade_new(ccv_array_t* posfil
 		{
 			ccv_scd_classifier_t* classifier = z.cascade->classifiers + z.t;
 			classifier->features = (ccv_scd_feature_t*)ccrealloc(classifier->features, sizeof(ccv_scd_feature_t) * (z.k + 1));
-			_ccv_scd_feature_supervised_train(rng, z.features, z.positives->rnum, z.negatives->rnum, z.pw, z.nw, z.fv, params.feature.active_set, params.feature.wide_set, params.C);
+			_ccv_scd_feature_supervised_train(rng, z.features, z.positives->rnum, z.negatives->rnum, z.pw, z.nw, z.fv, params.C);
 			int best_feature_no = _ccv_scd_best_feature_with_auc(z.s, z.features, z.positives->rnum, z.negatives->rnum, z.fv);
 			ccv_scd_feature_t best_feature = *(ccv_scd_feature_t*)ccv_array_get(z.features, best_feature_no);
 			for (i = 0; i < z.positives->rnum + z.negatives->rnum; i++)

test/unit/numeric.tests.c

@@ -69,14 +69,7 @@ TEST_CASE("minimize rosenbrock")
 	ccv_dense_matrix_t* x = ccv_dense_matrix_new(1, 2, CCV_64F | CCV_C1, 0, 0);
 	ccv_zero(x);
 	int steps = 0;
-	ccv_minimize_param_t params;
-	params.interp = 0.1;
-	params.extrap = 3.0;
-	params.max_iter = 20;
-	params.ratio = 10.0;
-	params.sig = 0.1;
-	params.rho = 0.05;
-	ccv_minimize(x, 25, 1.0, rosenbrock, params, &steps);
+	ccv_minimize(x, 25, 1.0, rosenbrock, ccv_minimize_default_params, &steps);
 	double dx[2] = { 1, 1 };
 	REQUIRE_ARRAY_EQ_WITH_TOLERANCE(double, x->data.f64, dx, 2, 1e-6, "the global minimal should be at (1.0, 1.0)");
 	ccv_matrix_free(x);