uct/policy/ucb1amaf.c

   1 #include <assert.h>
   2 #include <limits.h>
   3 #include <math.h>
   4 #include <stdio.h>
   5 #include <stdlib.h>
   6 #include <string.h>
   7
   8 #include "board.h"
   9 #include "debug.h"
  10 #include "move.h"
  11 #include "random.h"
  12 #include "tactics/util.h"
  13 #include "uct/internal.h"
  14 #include "uct/tree.h"
  15 #include "uct/policy/generic.h"
  16
  17 /* This implements the UCB1 policy with an extra AMAF heuristics. */
  18
  19 struct ucb1_policy_amaf {
  20         /* This is what the Modification of UCT with Patterns in Monte Carlo Go
  21          * paper calls 'p'. Original UCB has this on 2, but this seems to
  22          * produce way too wide searches; reduce this to get deeper and
  23          * narrower readouts - try 0.2. */
  24         floating_t explore_p;
  25         /* In distributed mode, encourage different slaves to work on different
  26          * parts of the tree by adding virtual wins to different nodes. */
  27         int virtual_win;
  28         int root_virtual_win;
  29         int vwin_min_playouts;
  30         /* First Play Urgency - if set to less than infinity (the MoGo paper
  31          * above reports 1.0 as the best), new branches are explored only
  32          * if none of the existing ones has higher urgency than fpu. */
  33         floating_t fpu;
  34         unsigned int equiv_rave;
  35         bool sylvain_rave;
  36         /* Give more weight to moves played earlier. */
  37         int distance_rave;
  38         /* Give 0 or negative rave bonus to ko threats before taking the ko.
  39            0=no bonus, 1=invert rave bonus, 2=double penalty, etc... */
  40         int threat_rave;
  41         /* Coefficient of local tree values embedded in RAVE. */
  42         floating_t ltree_rave;
  43         /* Coefficient of criticality embedded in RAVE. */
  44         floating_t crit_rave;
  45         int crit_min_playouts;
  46         floating_t crit_plthres_coef;
  47         bool crit_negative;
  48         bool crit_negflip;
  49         bool crit_amaf;
  50         bool crit_lvalue;
  51 };
  52
  53
  54 static inline floating_t fast_sqrt(unsigned int x)
  55 {
  56         static const floating_t table[] = {
  57                 0, 1, 1.41421356237309504880, 1.73205080756887729352,
  58                 2.00000000000000000000, 2.23606797749978969640,
  59                 2.44948974278317809819, 2.64575131106459059050,
  60                 2.82842712474619009760, 3.00000000000000000000,
  61                 3.16227766016837933199, 3.31662479035539984911,
  62                 3.46410161513775458705, 3.60555127546398929311,
  63                 3.74165738677394138558, 3.87298334620741688517,
  64                 4.00000000000000000000, 4.12310562561766054982,
  65                 4.24264068711928514640, 4.35889894354067355223,
  66                 4.47213595499957939281, 4.58257569495584000658,
  67                 4.69041575982342955456, 4.79583152331271954159,
  68                 4.89897948556635619639, 5.00000000000000000000,
  69                 5.09901951359278483002, 5.19615242270663188058,
  70                 5.29150262212918118100, 5.38516480713450403125,
  71                 5.47722557505166113456, 5.56776436283002192211,
  72                 5.65685424949238019520, 5.74456264653802865985,
  73                 5.83095189484530047087, 5.91607978309961604256,
  74                 6.00000000000000000000, 6.08276253029821968899,
  75                 6.16441400296897645025, 6.24499799839839820584,
  76                 6.32455532033675866399, 6.40312423743284868648,
  77                 6.48074069840786023096, 6.55743852430200065234,
  78                 6.63324958071079969822, 6.70820393249936908922,
  79                 6.78232998312526813906, 6.85565460040104412493,
  80                 6.92820323027550917410, 7.00000000000000000000,
  81                 7.07106781186547524400, 7.14142842854284999799,
  82                 7.21110255092797858623, 7.28010988928051827109,
  83                 7.34846922834953429459, 7.41619848709566294871,
  84                 7.48331477354788277116, 7.54983443527074969723,
  85                 7.61577310586390828566, 7.68114574786860817576,
  86                 7.74596669241483377035, 7.81024967590665439412,
  87                 7.87400787401181101968, 7.93725393319377177150,
  88         };
  89         if (x < sizeof(table) / sizeof(*table)) {
  90                 return table[x];
  91         } else {
  92                 return sqrt(x);
  93         }
  94 }
  95
  96 #define URAVE_DEBUG if (0)
  97 static floating_t inline
  98 ucb1rave_evaluate(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity)
  99 {
 100         struct ucb1_policy_amaf *b = p->data;
 101         struct tree_node *node = descent->node;
 102         struct tree_node *lnode = descent->lnode;
 103
 104         struct move_stats n = node->u, r = node->amaf;
 105         if (p->uct->amaf_prior) {
 106                 stats_merge(&r, &node->prior);
 107         } else {
 108                 stats_merge(&n, &node->prior);
 109         }
 110
 111         /* Local tree heuristics. */
 112         assert(!lnode || lnode->parent);
 113         if (p->uct->local_tree && b->ltree_rave > 0 && lnode
 114             && (p->uct->local_tree_rootchoose || lnode->parent->parent)) {
 115                 struct move_stats l = lnode->u;
 116                 l.playouts = ((floating_t) l.playouts) * b->ltree_rave / LTREE_PLAYOUTS_MULTIPLIER;
 117                 URAVE_DEBUG fprintf(stderr, "[ltree] adding [%s] %f%%%d to [%s] RAVE %f%%%d\n",
 118                         coord2sstr(node_coord(lnode), tree->board), l.value, l.playouts,
 119                         coord2sstr(node_coord(node), tree->board), r.value, r.playouts);
 120                 stats_merge(&r, &l);
 121         }
 122
 123         /* Criticality heuristics. */
 124         if (b->crit_rave > 0 && (b->crit_plthres_coef > 0
 125                                  ? node->u.playouts > tree->root->u.playouts * b->crit_plthres_coef
 126                                  : node->u.playouts > b->crit_min_playouts)) {
 127                 floating_t crit = tree_node_criticality(tree, node);
 128                 if (b->crit_negative || crit > 0) {
 129                         floating_t val = 1.0f;
 130                         if (b->crit_negflip && crit < 0) {
 131                                 val = 0;
 132                                 crit = -crit;
 133                         }
 134                         struct move_stats c = {
 135                                 .value = tree_node_get_value(tree, parity, val),
 136                                 .playouts = crit * r.playouts * b->crit_rave
 137                         };
 138                         URAVE_DEBUG fprintf(stderr, "[crit] adding %f%%%d to [%s] RAVE %f%%%d\n",
 139                                 c.value, c.playouts,
 140                                 coord2sstr(node_coord(node), tree->board), r.value, r.playouts);
 141                         stats_merge(&r, &c);
 142                 }
 143         }
 144
 145
 146         floating_t value = 0;
 147         if (n.playouts) {
 148                 if (r.playouts) {
 149                         /* At the beginning, beta is at 1 and RAVE is used.
 150                          * At b->equiv_rate, beta is at 1/3 and gets steeper on. */
 151                         floating_t beta;
 152                         if (b->sylvain_rave) {
 153                                 beta = (floating_t) r.playouts / (r.playouts + n.playouts
 154                                         + (floating_t) n.playouts * r.playouts / b->equiv_rave);
 155                         } else {
 156                                 /* XXX: This can be cached in descend; but we don't use this by default. */
 157                                 beta = sqrt(b->equiv_rave / (3 * node->parent->u.playouts + b->equiv_rave));
 158                         }
 159
 160                         value = beta * r.value + (1.f - beta) * n.value;
 161                         URAVE_DEBUG fprintf(stderr, "\t%s value = %f * %f + (1 - %f) * %f (prior %f)\n",
 162                                 coord2sstr(node_coord(node), tree->board), beta, r.value, beta, n.value, node->prior.value);
 163                 } else {
 164                         value = n.value;
 165                         URAVE_DEBUG fprintf(stderr, "\t%s value = %f (prior %f)\n",
 166                                 coord2sstr(node_coord(node), tree->board), n.value, node->prior.value);
 167                 }
 168         } else if (r.playouts) {
 169                 value = r.value;
 170                 URAVE_DEBUG fprintf(stderr, "\t%s value = rave %f (prior %f)\n",
 171                         coord2sstr(node_coord(node), tree->board), r.value, node->prior.value);
 172         }
 173         descent->value.playouts = r.playouts + n.playouts;
 174         descent->value.value = value;
 175         return tree_node_get_value(tree, parity, value);
 176 }
 177
 178 void
 179 ucb1rave_descend(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity, bool allow_pass)
 180 {
 181         struct ucb1_policy_amaf *b = p->data;
 182         floating_t nconf = 1.f;
 183         if (b->explore_p > 0)
 184                 nconf = sqrt(log(descent->node->u.playouts + descent->node->prior.playouts));
 185         struct uct *u = p->uct;
 186         int vwin = 0;
 187         if (u->max_slaves > 0 && u->slave_index >= 0)
 188                 vwin = descent->node == tree->root ? b->root_virtual_win : b->virtual_win;
 189         int child = 0;
 190
 191         uctd_try_node_children(tree, descent, allow_pass, parity, u->tenuki_d, di, urgency) {
 192                 struct tree_node *ni = di.node;
 193                 urgency = ucb1rave_evaluate(p, tree, &di, parity);
 194
 195                 /* In distributed mode, encourage different slaves to work on different
 196                  * parts of the tree. We rely on the fact that children (if they exist)
 197                  * are the same and in the same order in all slaves. */
 198                 if (vwin > 0 && ni->u.playouts > b->vwin_min_playouts && (child - u->slave_index) % u->max_slaves == 0)
 199                         urgency += vwin / (ni->u.playouts + vwin);
 200
 201                 if (ni->u.playouts > 0 && b->explore_p > 0) {
 202                         urgency += b->explore_p * nconf / fast_sqrt(ni->u.playouts);
 203
 204                 } else if (ni->u.playouts + ni->amaf.playouts + ni->prior.playouts == 0) {
 205                         /* assert(!u->even_eqex); */
 206                         urgency = b->fpu;
 207                 }
 208         } uctd_set_best_child(di, urgency);
 209
 210         uctd_get_best_child(descent);
 211 }
 212
 213
 214 void
 215 ucb1amaf_update(struct uct_policy *p, struct tree *tree, struct tree_node *node,
 216                 enum stone node_color, enum stone player_color,
 217                 struct playout_amafmap *map, struct board *final_board,
 218                 floating_t result)
 219 {
 220         struct ucb1_policy_amaf *b = p->data;
 221         enum stone winner_color = result > 0.5 ? S_BLACK : S_WHITE;
 222
 223         /* Record of the random playout - for each intersection coord,
 224          * first_move[coord] is the index map->game of the first move
 225          * at this coordinate, or INT_MAX if the move was not played.
 226          * The parity gives the color of this move.
 227          */
 228         int first_map[board_size2(final_board)+1];
 229         int *first_move = &first_map[1]; // +1 for pass
 230
 231 #if 0
 232         struct board bb; bb.size = 9+2;
 233         for (struct tree_node *ni = node; ni; ni = ni->parent)
 234                 fprintf(stderr, "%s ", coord2sstr(node_coord(ni), &bb));
 235         fprintf(stderr, "[color %d] update result %d (color %d)\n",
 236                         node_color, result, player_color);
 237 #endif
 238
 239         /* Initialize first_move */
 240         for (int i = pass; i < board_size2(final_board); i++) first_move[i] = INT_MAX;
 241         int move;
 242         assert(map->gamelen > 0);
 243         for (move = map->gamelen - 1; move >= map->game_baselen; move--)
 244                 first_move[map->game[move]] = move;
 245
 246         while (node) {
 247                 if (!b->crit_amaf && !is_pass(node_coord(node))) {
 248                         stats_add_result(&node->winner_owner, board_local_value(b->crit_lvalue, final_board, node_coord(node), winner_color), 1);
 249                         stats_add_result(&node->black_owner, board_local_value(b->crit_lvalue, final_board, node_coord(node), S_BLACK), 1);
 250                 }
 251                 stats_add_result(&node->u, result, 1);
 252
 253                 bool capturing_ko = move + 1 < map->gamelen && map->is_ko_capture[move+1];
 254
 255                 /* This loop ignores symmetry considerations, but they should
 256                  * matter only at a point when AMAF doesn't help much. */
 257                 assert(map->game_baselen >= 0);
 258                 for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
 259                         if (is_pass(node_coord(ni))) continue;
 260
 261                         /* Use the child move only if it was first played by the same color. */
 262                         int first = first_move[node_coord(ni)];
 263                         if (first == INT_MAX) continue;
 264                         assert(first > move && first < map->gamelen);
 265                         int distance = first - (move + 1);
 266                         if (distance & 1) continue;
 267
 268                         int weight = 1;
 269                         floating_t res = result;
 270
 271                         /* Don't give amaf bonus to a ko threat before taking the ko.
 272                          * http://www.grappa.univ-lille3.fr/~coulom/Aja_PhD_Thesis.pdf
 273                          * move+1: B captures a ko
 274                          * move+2: W plays a ko threat
 275                          * move+3: B answers ko threat
 276                          * move+4: W re-captures the ko
 277                          * move+5: B plays a ko threat
 278                          * then do not give a amaf bonus to this threat at level move+1, prefer taking ko.
 279                          */
 280                         if (capturing_ko && distance == 4 && map->is_ko_capture[move+4]) {
 281                                 weight = b->threat_rave;
 282                                 res = 1.0 - res;
 283                         } else if (b->distance_rave != 0) {
 284                                 /* Give more weight to moves played earlier */
 285                                 weight += b->distance_rave * (map->gamelen - first) / (map->gamelen - move);
 286                         }
 287                         stats_add_result(&ni->amaf, res, weight);
 288
 289                         if (b->crit_amaf) {
 290                                 stats_add_result(&ni->winner_owner, board_local_value(b->crit_lvalue, final_board, node_coord(ni), winner_color), 1);
 291                                 stats_add_result(&ni->black_owner, board_local_value(b->crit_lvalue, final_board, node_coord(ni), S_BLACK), 1);
 292                         }
 293 #if 0
 294                         struct board bb; bb.size = 9+2;
 295                         fprintf(stderr, "* %s<%"PRIhash"> -> %s<%"PRIhash"> [%d/%f => %d/%f]\n",
 296                                 coord2sstr(node_coord(node), &bb), node->hash,
 297                                 coord2sstr(node_coord(ni), &bb), ni->hash,
 298                                 player_color, result, move, res);
 299 #endif
 300                 }
 301                 if (node->parent) {
 302                         assert(move >= 0 && map->game[move] == node_coord(node) && first_move[node_coord(node)] > move);
 303                         first_move[node_coord(node)] = move;
 304                         move--;
 305                 }
 306                 node = node->parent;
 307         }
 308 }
 309
 310
 311 struct uct_policy *
 312 policy_ucb1amaf_init(struct uct *u, char *arg, struct board *board)
 313 {
 314         struct uct_policy *p = calloc2(1, sizeof(*p));
 315         struct ucb1_policy_amaf *b = calloc2(1, sizeof(*b));
 316         p->uct = u;
 317         p->data = b;
 318         p->choose = uctp_generic_choose;
 319         p->winner = uctp_generic_winner;
 320         p->evaluate = ucb1rave_evaluate;
 321         p->descend = ucb1rave_descend;
 322         p->update = ucb1amaf_update;
 323         p->wants_amaf = true;
 324
 325         b->explore_p = 0;
 326         b->equiv_rave = board_large(board) ? 4000 : 3000;
 327         b->fpu = INFINITY;
 328         b->sylvain_rave = true;
 329         b->distance_rave = 3;
 330         b->threat_rave = 0;
 331         b->ltree_rave = 0.75f;
 332
 333         b->crit_rave = 1.1f;
 334         b->crit_min_playouts = 2000;
 335         b->crit_negative = 1;
 336         b->crit_amaf = 0;
 337
 338         b->root_virtual_win = -1;
 339         b->vwin_min_playouts = 1000;
 340
 341         if (arg) {
 342                 char *optspec, *next = arg;
 343                 while (*next) {
 344                         optspec = next;
 345                         next += strcspn(next, ":");
 346                         if (*next) { *next++ = 0; } else { *next = 0; }
 347
 348                         char *optname = optspec;
 349                         char *optval = strchr(optspec, '=');
 350                         if (optval) *optval++ = 0;
 351
 352                         if (!strcasecmp(optname, "explore_p")) {
 353                                 b->explore_p = atof(optval);
 354                         } else if (!strcasecmp(optname, "fpu") && optval) {
 355                                 b->fpu = atof(optval);
 356                         } else if (!strcasecmp(optname, "equiv_rave") && optval) {
 357                                 b->equiv_rave = atof(optval);
 358                         } else if (!strcasecmp(optname, "sylvain_rave")) {
 359                                 b->sylvain_rave = !optval || *optval == '1';
 360                         } else if (!strcasecmp(optname, "distance_rave") && optval) {
 361                                 b->distance_rave = atoi(optval);
 362                         } else if (!strcasecmp(optname, "threat_rave") && optval) {
 363                                 b->threat_rave = atoi(optval);
 364                         } else if (!strcasecmp(optname, "ltree_rave") && optval) {
 365                                 b->ltree_rave = atof(optval);
 366                         } else if (!strcasecmp(optname, "crit_rave") && optval) {
 367                                 b->crit_rave = atof(optval);
 368                         } else if (!strcasecmp(optname, "crit_min_playouts") && optval) {
 369                                 b->crit_min_playouts = atoi(optval);
 370                         } else if (!strcasecmp(optname, "crit_plthres_coef") && optval) {
 371                                 b->crit_plthres_coef = atof(optval);
 372                         } else if (!strcasecmp(optname, "crit_negative")) {
 373                                 b->crit_negative = !optval || *optval == '1';
 374                         } else if (!strcasecmp(optname, "crit_negflip")) {
 375                                 b->crit_negflip = !optval || *optval == '1';
 376                         } else if (!strcasecmp(optname, "crit_amaf")) {
 377                                 b->crit_amaf = !optval || *optval == '1';
 378                         } else if (!strcasecmp(optname, "crit_lvalue")) {
 379                                 b->crit_lvalue = !optval || *optval == '1';
 380                         } else if (!strcasecmp(optname, "virtual_win") && optval) {
 381                                 b->virtual_win = atoi(optval);
 382                         } else if (!strcasecmp(optname, "root_virtual_win") && optval) {
 383                                 b->root_virtual_win = atoi(optval);
 384                         } else if (!strcasecmp(optname, "vwin_min_playouts") && optval) {
 385                                 b->vwin_min_playouts = atoi(optval);
 386                         } else {
 387                                 fprintf(stderr, "ucb1amaf: Invalid policy argument %s or missing value\n",
 388                                         optname);
 389                                 exit(1);
 390                         }
 391                 }
 392         }
 393         if (b->root_virtual_win < 0)
 394                 b->root_virtual_win = b->virtual_win;
 395
 396         return p;
 397 }