uct/policy/ucb1amaf.c

   1 #include <assert.h>
   2 #include <math.h>
   3 #include <stdio.h>
   4 #include <stdlib.h>
   5 #include <string.h>
   6
   7 #include "board.h"
   8 #include "debug.h"
   9 #include "move.h"
  10 #include "random.h"
  11 #include "uct/internal.h"
  12 #include "uct/tree.h"
  13
  14 /* This implements the UCB1 policy with an extra AMAF heuristics. */
  15
  16 struct ucb1_policy_amaf {
  17         /* This is what the Modification of UCT with Patterns in Monte Carlo Go
  18          * paper calls 'p'. Original UCB has this on 2, but this seems to
  19          * produce way too wide searches; reduce this to get deeper and
  20          * narrower readouts - try 0.2. */
  21         float explore_p;
  22         /* First Play Urgency - if set to less than infinity (the MoGo paper
  23          * above reports 1.0 as the best), new branches are explored only
  24          * if none of the existing ones has higher urgency than fpu. */
  25         float fpu;
  26         int urg_randoma, urg_randomm;
  27         float explore_p_rave;
  28         int equiv_rave;
  29         bool both_colors;
  30         bool check_nakade;
  31 };
  32
  33
  34 struct tree_node *ucb1_choose(struct uct_policy *p, struct tree_node *node, struct board *b, enum stone color);
  35
  36 struct tree_node *ucb1_descend(struct uct_policy *p, struct tree *tree, struct tree_node *node, int parity, bool allow_pass);
  37
  38
  39 static inline float fast_sqrt(int x)
  40 {
  41         static const float table[] = {
  42                 0, 1, 1.41421356237309504880, 1.73205080756887729352,
  43                 2.00000000000000000000, 2.23606797749978969640,
  44                 2.44948974278317809819, 2.64575131106459059050,
  45                 2.82842712474619009760, 3.00000000000000000000,
  46                 3.16227766016837933199, 3.31662479035539984911,
  47                 3.46410161513775458705, 3.60555127546398929311,
  48                 3.74165738677394138558, 3.87298334620741688517,
  49                 4.00000000000000000000, 4.12310562561766054982,
  50                 4.24264068711928514640, 4.35889894354067355223,
  51                 4.47213595499957939281, 4.58257569495584000658,
  52                 4.69041575982342955456, 4.79583152331271954159,
  53                 4.89897948556635619639, 5.00000000000000000000,
  54                 5.09901951359278483002, 5.19615242270663188058,
  55                 5.29150262212918118100, 5.38516480713450403125,
  56                 5.47722557505166113456, 5.56776436283002192211,
  57                 5.65685424949238019520, 5.74456264653802865985,
  58                 5.83095189484530047087, 5.91607978309961604256,
  59                 6.00000000000000000000, 6.08276253029821968899,
  60                 6.16441400296897645025, 6.24499799839839820584,
  61                 6.32455532033675866399, 6.40312423743284868648,
  62                 6.48074069840786023096, 6.55743852430200065234,
  63                 6.63324958071079969822, 6.70820393249936908922,
  64                 6.78232998312526813906, 6.85565460040104412493,
  65                 6.92820323027550917410, 7.00000000000000000000,
  66                 7.07106781186547524400, 7.14142842854284999799,
  67                 7.21110255092797858623, 7.28010988928051827109,
  68                 7.34846922834953429459, 7.41619848709566294871,
  69                 7.48331477354788277116, 7.54983443527074969723,
  70                 7.61577310586390828566, 7.68114574786860817576,
  71                 7.74596669241483377035, 7.81024967590665439412,
  72                 7.87400787401181101968, 7.93725393319377177150,
  73         };
  74         //printf("sqrt %d\n", x);
  75         if (x < sizeof(table) / sizeof(*table)) {
  76                 return table[x];
  77         } else {
  78                 return sqrt(x);
  79         }
  80 }
  81
  82 /* Sylvain RAVE function */
  83 struct tree_node *
  84 ucb1srave_descend(struct uct_policy *p, struct tree *tree, struct tree_node *node, int parity, bool allow_pass)
  85 {
  86         struct ucb1_policy_amaf *b = p->data;
  87         float rave_coef = 1.0f / b->equiv_rave;
  88         float nconf = 1.f, rconf = 1.f;
  89         if (b->explore_p > 0)
  90                 nconf = sqrt(log(node->u.playouts + node->prior.playouts));
  91         if (b->explore_p_rave > 0 && node->amaf.playouts)
  92                 rconf = sqrt(log(node->amaf.playouts + node->prior.playouts));
  93
  94         // XXX: Stack overflow danger on big boards?
  95         struct tree_node *nbest[512] = { node->children }; int nbests = 1;
  96         float best_urgency = -9999;
  97
  98         for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
  99                 /* Do not consider passing early. */
 100                 if (likely(!allow_pass) && unlikely(is_pass(ni->coord)))
 101                         continue;
 102
 103                 /* TODO: Exploration? */
 104
 105                 int ngames = ni->u.playouts;
 106                 int nwins = ni->u.wins;
 107                 int rgames = ni->amaf.playouts;
 108                 int rwins = ni->amaf.wins;
 109                 if (p->uct->amaf_prior) {
 110                         rgames += ni->prior.playouts;
 111                         rwins += ni->prior.wins;
 112                 } else {
 113                         ngames += ni->prior.playouts;
 114                         nwins += ni->prior.wins;
 115                 }
 116                 if (tree_parity(tree, parity) < 0) {
 117                         nwins = ngames - nwins;
 118                         rwins = rgames - rwins;
 119                 }
 120                 float nval = 0, rval = 0;
 121                 if (ngames) {
 122                         nval = (float) nwins / ngames;
 123                         if (b->explore_p > 0)
 124                                 nval += b->explore_p * nconf / fast_sqrt(ngames);
 125                 }
 126                 if (rgames) {
 127                         rval = (float) rwins / rgames;
 128                         if (b->explore_p_rave > 0 && !is_pass(ni->coord))
 129                                 rval += b->explore_p_rave * rconf / fast_sqrt(rgames);
 130                 }
 131
 132                 float urgency;
 133                 if (ngames) {
 134                         if (rgames) {
 135                                 /* At the beginning, beta is at 1 and RAVE is used.
 136                                  * At b->equiv_rate, beta is at 1/3 and gets steeper on. */
 137                                 float beta = (float) rgames / (rgames + ngames + rave_coef * ngames * rgames);
 138 #if 0
 139                                 //if (node->coord == 7*11+4) // D7
 140                                 fprintf(stderr, "[beta %f = %d / (%d + %d + %f)]\n",
 141                                         beta, rgames, rgames, ngames, rave_coef * ngames * rgames);
 142 #endif
 143                                 urgency = beta * rval + (1 - beta) * nval;
 144                         } else {
 145                                 urgency = nval;
 146                         }
 147                 } else if (rgames) {
 148                         urgency = rval;
 149                 } else {
 150                         /* assert(!u->even_eqex); */
 151                         urgency = b->fpu;
 152                 }
 153
 154 #if 0
 155                 struct board bb; bb.size = 11;
 156                 //if (node->coord == 7*11+4) // D7
 157                 fprintf(stderr, "%s<%lld>-%s<%lld> urgency %f (r %d / %d, n %d / %d)\n",
 158                         coord2sstr(ni->parent->coord, &bb), ni->parent->hash,
 159                         coord2sstr(ni->coord, &bb), ni->hash, urgency,
 160                         rwins, rgames, nwins, ngames);
 161 #endif
 162                 if (b->urg_randoma)
 163                         urgency += (float)(fast_random(b->urg_randoma) - b->urg_randoma / 2) / 1000;
 164                 if (b->urg_randomm)
 165                         urgency *= (float)(fast_random(b->urg_randomm) + 5) / b->urg_randomm;
 166
 167                 if (urgency - best_urgency > __FLT_EPSILON__) { // urgency > best_urgency
 168                         best_urgency = urgency; nbests = 0;
 169                 }
 170                 if (urgency - best_urgency > -__FLT_EPSILON__) { // urgency >= best_urgency
 171                         /* We want to always choose something else than a pass
 172                          * in case of a tie. pass causes degenerative behaviour. */
 173                         if (nbests == 1 && is_pass(nbest[0]->coord)) {
 174                                 nbests--;
 175                         }
 176                         nbest[nbests++] = ni;
 177                 }
 178         }
 179 #if 0
 180         struct board bb; bb.size = 11;
 181         fprintf(stderr, "[%s %d: ", coord2sstr(node->coord, &bb), nbests);
 182         for (int zz = 0; zz < nbests; zz++)
 183                 fprintf(stderr, "%s", coord2sstr(nbest[zz]->coord, &bb));
 184         fprintf(stderr, "]\n");
 185 #endif
 186         return nbest[fast_random(nbests)];
 187 }
 188
 189 static void
 190 update_node(struct uct_policy *p, struct tree_node *node, int result)
 191 {
 192         node->u.playouts++;
 193         node->u.wins += result;
 194         tree_update_node_value(node, p->uct->amaf_prior);
 195 }
 196
 197 static void
 198 update_node_amaf(struct uct_policy *p, struct tree_node *node, int result)
 199 {
 200         node->amaf.playouts++;
 201         node->amaf.wins += result;
 202         tree_update_node_rvalue(node, p->uct->amaf_prior);
 203 }
 204
 205 void
 206 ucb1amaf_update(struct uct_policy *p, struct tree *tree, struct tree_node *node, enum stone node_color, enum stone player_color, struct playout_amafmap *map, int result)
 207 {
 208         struct ucb1_policy_amaf *b = p->data;
 209         enum stone child_color = stone_other(node_color);
 210
 211 #if 0
 212         struct board bb; bb.size = 9+2;
 213         for (struct tree_node *ni = node; ni; ni = ni->parent)
 214                 fprintf(stderr, "%s ", coord2sstr(ni->coord, &bb));
 215         fprintf(stderr, "[color %d] update result %d (color %d)\n",
 216                         node_color, result, player_color);
 217 #endif
 218
 219         while (node) {
 220                 if (node->parent == NULL)
 221                         assert(tree->root_color == stone_other(child_color));
 222
 223                 update_node(p, node, result);
 224                 if (amaf_nakade(map->map[node->coord]))
 225                         amaf_op(map->map[node->coord], -);
 226
 227                 /* This loop ignores symmetry considerations, but they should
 228                  * matter only at a point when AMAF doesn't help much. */
 229                 for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
 230                         assert(map->map[ni->coord] != S_OFFBOARD);
 231                         if (map->map[ni->coord] == S_NONE)
 232                                 continue;
 233                         assert(map->game_baselen >= 0);
 234                         enum stone amaf_color = map->map[ni->coord];
 235                         if (amaf_nakade(map->map[ni->coord])) {
 236                                 if (!b->check_nakade)
 237                                         continue;
 238                                 /* We don't care to implement both_colors
 239                                  * properly since it sucks anyway. */
 240                                 int i;
 241                                 for (i = map->game_baselen; i < map->gamelen; i++)
 242                                         if (map->game[i].coord == ni->coord
 243                                             && map->game[i].color == child_color)
 244                                                 break;
 245                                 if (i == map->gamelen)
 246                                         continue;
 247                                 amaf_color = child_color;
 248                         }
 249
 250                         int nres = result;
 251                         if (amaf_color != child_color) {
 252                                 if (!b->both_colors)
 253                                         continue;
 254                                 nres = !nres;
 255                         }
 256                         /* For child_color != player_color, we still want
 257                          * to record the result unmodified; in that case,
 258                          * we will correctly negate them at the descend phase. */
 259
 260                         update_node_amaf(p, ni, nres);
 261
 262 #if 0
 263                         fprintf(stderr, "* %s<%lld> -> %s<%lld> [%d %d => %d/%d]\n", coord2sstr(node->coord, &bb), node->hash, coord2sstr(ni->coord, &bb), ni->hash, player_color, child_color, result);
 264 #endif
 265                 }
 266
 267                 if (!is_pass(node->coord)) {
 268                         map->game_baselen--;
 269                 }
 270                 node = node->parent; child_color = stone_other(child_color);
 271         }
 272 }
 273
 274
 275 struct uct_policy *
 276 policy_ucb1amaf_init(struct uct *u, char *arg)
 277 {
 278         struct uct_policy *p = calloc(1, sizeof(*p));
 279         struct ucb1_policy_amaf *b = calloc(1, sizeof(*b));
 280         p->uct = u;
 281         p->data = b;
 282         p->descend = ucb1srave_descend;
 283         p->choose = ucb1_choose;
 284         p->update = ucb1amaf_update;
 285         p->wants_amaf = true;
 286
 287         // RAVE: 0.2vs0: 40% (+-7.3) 0.1vs0: 54.7% (+-3.5)
 288         b->explore_p = 0.1;
 289         b->explore_p_rave = 0.01;
 290         b->equiv_rave = 3000;
 291         b->fpu = INFINITY;
 292         b->check_nakade = true;
 293
 294         if (arg) {
 295                 char *optspec, *next = arg;
 296                 while (*next) {
 297                         optspec = next;
 298                         next += strcspn(next, ":");
 299                         if (*next) { *next++ = 0; } else { *next = 0; }
 300
 301                         char *optname = optspec;
 302                         char *optval = strchr(optspec, '=');
 303                         if (optval) *optval++ = 0;
 304
 305                         if (!strcasecmp(optname, "explore_p")) {
 306                                 b->explore_p = atof(optval);
 307                         } else if (!strcasecmp(optname, "fpu") && optval) {
 308                                 b->fpu = atof(optval);
 309                         } else if (!strcasecmp(optname, "urg_randoma") && optval) {
 310                                 b->urg_randoma = atoi(optval);
 311                         } else if (!strcasecmp(optname, "urg_randomm") && optval) {
 312                                 b->urg_randomm = atoi(optval);
 313                         } else if (!strcasecmp(optname, "explore_p_rave") && optval) {
 314                                 b->explore_p_rave = atof(optval);
 315                         } else if (!strcasecmp(optname, "equiv_rave") && optval) {
 316                                 b->equiv_rave = atof(optval);
 317                         } else if (!strcasecmp(optname, "both_colors")) {
 318                                 b->both_colors = true;
 319                         } else if (!strcasecmp(optname, "check_nakade")) {
 320                                 b->check_nakade = !optval || *optval == '1';
 321                         } else {
 322                                 fprintf(stderr, "ucb1: Invalid policy argument %s or missing value\n", optname);
 323                         }
 324                 }
 325         }
 326
 327         if (b->explore_p_rave < 0) b->explore_p_rave = b->explore_p;
 328
 329         return p;
 330 }