X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=bayeswf.cpp;h=a691b8fe915ac8f6bb8714e62bd2349c173201b1;hb=35021b4f76156feb48fd9420a9116da50acee3e1;hp=6441801e23c32c17bab637f4c8ad09b7b609b23c;hpb=72a2eedd021d38d6bb0786033c6f100062363789;p=wloh

diff --git a/bayeswf.cpp b/bayeswf.cpp
index 6441801..a691b8f 100644
--- a/bayeswf.cpp
+++ b/bayeswf.cpp
@@ -2,6 +2,8 @@
 #include <math.h>
 #include <string.h>
 #include <stdlib.h>
+#include <Eigen/Core>
+#include <Eigen/Eigenvalues>
 
 #include <map>
 #include <vector>
@@ -9,6 +11,7 @@
 #include <algorithm>
 
 using namespace std;
+using namespace Eigen;
 
 #define PRIOR_MU 1500
 #define PRIOR_WEIGHT 1.0
@@ -196,11 +199,15 @@ float compute_total_logl(float *mu, int num_players)
  *
  * Note that this does not depend on mu or the margin at all.
  */
-double hessian[MAX_PLAYERS][MAX_PLAYERS];
+Matrix<float, Dynamic, Dynamic> hessian;
 void construct_hessian(const float *mu, int num_players)
 {
-	memset(hessian, 0, sizeof(hessian));
+	hessian = Matrix<float, Dynamic, Dynamic>(num_players, num_players);
+	hessian.fill(0.0f);
 
+	for (int i = 0; i < num_players; ++i) {
+		hessian(i, i) += 1.0f / (prior_sigma * prior_sigma);
+	}
 	for (unsigned i = 0; i < all_matches.size(); ++i) {
 		const match &m = all_matches[i];
 
@@ -210,35 +217,22 @@ void construct_hessian(const float *mu, int num_players)
 		double sigma_sq = global_sigma * global_sigma;
 		float w = m.weight;
 
-		hessian[p1][p2] -= w / sigma_sq;
-		hessian[p2][p1] -= w / sigma_sq;
+		hessian(p1, p2) -= w / sigma_sq;
+		hessian(p2, p1) -= w / sigma_sq;
 
-		hessian[p1][p1] += w / sigma_sq;
-		hessian[p2][p2] += w / sigma_sq;
+		hessian(p1, p1) += w / sigma_sq;
+		hessian(p2, p2) += w / sigma_sq;
 	}
 }
 
-// Compute uncertainty (stddev) of mu estimates, which is 1/sqrt(H_ii),
+// Compute uncertainty (stddev) of mu estimates, which is sqrt((H^-1)_ii),
 // where H is the Hessian (see construct_hessian()).
 void compute_mu_uncertainty(const float *mu, int num_players)
 {
-	memset(mu_stddev, 0, sizeof(mu_stddev));
-
-	for (unsigned i = 0; i < all_matches.size(); ++i) {
-		const match &m = all_matches[i];
-
-		int p1 = m.player;
-		int p2 = m.other_player;
-
-		double sigma_sq = global_sigma * global_sigma;
-		float w = m.weight;
-
-		// Temporarily use mu_stddev to store the diagonal of the Hessian.
-		mu_stddev[p1] += w / sigma_sq;
-		mu_stddev[p2] += w / sigma_sq;
-	}
+	// FIXME: Use pseudoinverse if applicable.
+	Matrix<float, Dynamic, Dynamic> ih = hessian.inverse();
 	for (int i = 0; i < num_players; ++i) {
-		mu_stddev[i] = 1.0f / sqrt(mu_stddev[i]);
+		mu_stddev[i] = sqrt(ih(i, i));
 	}
 }
 
@@ -337,7 +331,7 @@ int main(int argc, char **argv)
 		sumdiff += (global_sigma - old_global_sigma) * (global_sigma - old_global_sigma);
 		if (sumdiff < EPSILON) {
 			//fprintf(stderr, "Converged after %d iterations. Stopping.\n", j);
-			printf("%d -1\n", j + 1);
+			printf("%d 0 -1\n", j + 1);
 			break;
 		}
 	}
@@ -345,15 +339,14 @@ int main(int argc, char **argv)
 #if DUMP_RAW
 	dump_raw(mu, num_players);
 #else
+	construct_hessian(mu, num_players);
 	compute_mu_uncertainty(mu, num_players);
 	dump_scores(players, mu, mu_stddev, num_players);
 	//fprintf(stderr, "Optimal sigma: %f (two-player: %f)\n", sigma[0], sigma[0] * sqrt(2.0f));
-	printf("%f -2\n", global_sigma / sqrt(2.0f));
-	printf("%f -3\n", prior_sigma);
+	printf("%f 0 -2\n", global_sigma / sqrt(2.0f));
+	printf("%f 0 -3\n", prior_sigma);
 
 	float total_logl = compute_total_logl(mu, num_players);
-	printf("%f -4\n", total_logl);
-
-//	construct_hessian(mu, sigma, num_players);
+	printf("%f 0 -4\n", total_logl);
 #endif
 }