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 #include "uct/policy/generic.h"
  14
  15 /* This implements the UCB1 policy with an extra AMAF heuristics. */
  16
  17 struct ucb1_policy_amaf {
  18         /* This is what the Modification of UCT with Patterns in Monte Carlo Go
  19          * paper calls 'p'. Original UCB has this on 2, but this seems to
  20          * produce way too wide searches; reduce this to get deeper and
  21          * narrower readouts - try 0.2. */
  22         floating_t explore_p;
  23         /* In distributed mode, encourage different slaves to work on different
  24          * parts of the tree by adding virtual wins to different nodes. */
  25         int virtual_win;
  26         int root_virtual_win;
  27         int vwin_min_playouts;
  28         /* First Play Urgency - if set to less than infinity (the MoGo paper
  29          * above reports 1.0 as the best), new branches are explored only
  30          * if none of the existing ones has higher urgency than fpu. */
  31         floating_t fpu;
  32         unsigned int equiv_rave;
  33         bool check_nakade;
  34         bool sylvain_rave;
  35         /* Coefficient of local tree values embedded in RAVE. */
  36         floating_t ltree_rave;
  37         /* Coefficient of criticality embedded in RAVE. */
  38         floating_t crit_rave;
  39         int crit_min_playouts;
  40         bool crit_negative;
  41         bool crit_amaf;
  42 };
  43
  44
  45 static inline floating_t fast_sqrt(unsigned int x)
  46 {
  47         static const floating_t table[] = {
  48                 0, 1, 1.41421356237309504880, 1.73205080756887729352,
  49                 2.00000000000000000000, 2.23606797749978969640,
  50                 2.44948974278317809819, 2.64575131106459059050,
  51                 2.82842712474619009760, 3.00000000000000000000,
  52                 3.16227766016837933199, 3.31662479035539984911,
  53                 3.46410161513775458705, 3.60555127546398929311,
  54                 3.74165738677394138558, 3.87298334620741688517,
  55                 4.00000000000000000000, 4.12310562561766054982,
  56                 4.24264068711928514640, 4.35889894354067355223,
  57                 4.47213595499957939281, 4.58257569495584000658,
  58                 4.69041575982342955456, 4.79583152331271954159,
  59                 4.89897948556635619639, 5.00000000000000000000,
  60                 5.09901951359278483002, 5.19615242270663188058,
  61                 5.29150262212918118100, 5.38516480713450403125,
  62                 5.47722557505166113456, 5.56776436283002192211,
  63                 5.65685424949238019520, 5.74456264653802865985,
  64                 5.83095189484530047087, 5.91607978309961604256,
  65                 6.00000000000000000000, 6.08276253029821968899,
  66                 6.16441400296897645025, 6.24499799839839820584,
  67                 6.32455532033675866399, 6.40312423743284868648,
  68                 6.48074069840786023096, 6.55743852430200065234,
  69                 6.63324958071079969822, 6.70820393249936908922,
  70                 6.78232998312526813906, 6.85565460040104412493,
  71                 6.92820323027550917410, 7.00000000000000000000,
  72                 7.07106781186547524400, 7.14142842854284999799,
  73                 7.21110255092797858623, 7.28010988928051827109,
  74                 7.34846922834953429459, 7.41619848709566294871,
  75                 7.48331477354788277116, 7.54983443527074969723,
  76                 7.61577310586390828566, 7.68114574786860817576,
  77                 7.74596669241483377035, 7.81024967590665439412,
  78                 7.87400787401181101968, 7.93725393319377177150,
  79         };
  80         if (x < sizeof(table) / sizeof(*table)) {
  81                 return table[x];
  82         } else {
  83                 return sqrt(x);
  84         }
  85 }
  86
  87 #define LTREE_DEBUG if (0)
  88 static floating_t inline
  89 ucb1rave_evaluate(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity)
  90 {
  91         struct ucb1_policy_amaf *b = p->data;
  92         struct tree_node *node = descent->node;
  93         struct tree_node *lnode = descent->lnode;
  94
  95         struct move_stats n = node->u, r = node->amaf;
  96         if (p->uct->amaf_prior) {
  97                 stats_merge(&r, &node->prior);
  98         } else {
  99                 stats_merge(&n, &node->prior);
 100         }
 101
 102         /* Local tree heuristics. */
 103         assert(!lnode || lnode->parent);
 104         if (p->uct->local_tree && b->ltree_rave > 0 && lnode
 105             && (p->uct->local_tree_rootchoose || lnode->parent->parent)) {
 106                 struct move_stats l = lnode->u;
 107                 l.playouts = ((floating_t) l.playouts) * b->ltree_rave / LTREE_PLAYOUTS_MULTIPLIER;
 108                 LTREE_DEBUG fprintf(stderr, "[ltree] adding [%s] %f%%%d to [%s] RAVE %f%%%d\n",
 109                         coord2sstr(lnode->coord, tree->board), l.value, l.playouts,
 110                         coord2sstr(node->coord, tree->board), r.value, r.playouts);
 111                 stats_merge(&r, &l);
 112         }
 113
 114         /* Criticality heuristics. */
 115         if (b->crit_rave > 0 && node->u.playouts > b->crit_min_playouts) {
 116                 floating_t crit = tree_node_criticality(tree, node);
 117                 if (b->crit_negative || crit > 0) {
 118                         struct move_stats c = {
 119                                 .value = tree_node_get_value(tree, parity, 1.0f),
 120                                 .playouts = crit * r.playouts * b->crit_rave
 121                         };
 122                         LTREE_DEBUG fprintf(stderr, "[crit] adding %f%%%d to [%s] RAVE %f%%%d\n",
 123                                 c.value, c.playouts,
 124                                 coord2sstr(node->coord, tree->board), r.value, r.playouts);
 125                         stats_merge(&r, &c);
 126                 }
 127         }
 128
 129
 130         floating_t value = 0;
 131         if (n.playouts) {
 132                 if (r.playouts) {
 133                         /* At the beginning, beta is at 1 and RAVE is used.
 134                          * At b->equiv_rate, beta is at 1/3 and gets steeper on. */
 135                         floating_t beta;
 136                         if (b->sylvain_rave) {
 137                                 beta = (floating_t) r.playouts / (r.playouts + n.playouts
 138                                         + (floating_t) n.playouts * r.playouts / b->equiv_rave);
 139                         } else {
 140                                 /* XXX: This can be cached in descend; but we don't use this by default. */
 141                                 beta = sqrt(b->equiv_rave / (3 * node->parent->u.playouts + b->equiv_rave));
 142                         }
 143
 144                         value = beta * r.value + (1.f - beta) * n.value;
 145                 } else {
 146                         value = n.value;
 147                 }
 148         } else if (r.playouts) {
 149                 value = r.value;
 150         }
 151         descent->value.playouts = r.playouts + n.playouts;
 152         descent->value.value = value;
 153         return tree_node_get_value(tree, parity, value);
 154 }
 155
 156 void
 157 ucb1rave_descend(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity, bool allow_pass)
 158 {
 159         struct ucb1_policy_amaf *b = p->data;
 160         floating_t nconf = 1.f;
 161         if (b->explore_p > 0)
 162                 nconf = sqrt(log(descent->node->u.playouts + descent->node->prior.playouts));
 163         struct uct *u = p->uct;
 164         int vwin = 0;
 165         if (u->max_slaves > 0 && u->slave_index >= 0)
 166                 vwin = descent->node == tree->root ? b->root_virtual_win : b->virtual_win;
 167         int child = 0;
 168
 169         uctd_try_node_children(tree, descent, allow_pass, parity, u->tenuki_d, di, urgency) {
 170                 struct tree_node *ni = di.node;
 171                 urgency = ucb1rave_evaluate(p, tree, &di, parity);
 172
 173                 /* In distributed mode, encourage different slaves to work on different
 174                  * parts of the tree. We rely on the fact that children (if they exist)
 175                  * are the same and in the same order in all slaves. */
 176                 if (vwin > 0 && ni->u.playouts > b->vwin_min_playouts && (child - u->slave_index) % u->max_slaves == 0)
 177                         urgency += vwin / (ni->u.playouts + vwin);
 178
 179                 if (ni->u.playouts > 0 && b->explore_p > 0) {
 180                         urgency += b->explore_p * nconf / fast_sqrt(ni->u.playouts);
 181
 182                 } else if (ni->u.playouts + ni->amaf.playouts + ni->prior.playouts == 0) {
 183                         /* assert(!u->even_eqex); */
 184                         urgency = b->fpu;
 185                 }
 186         } uctd_set_best_child(di, urgency);
 187
 188         uctd_get_best_child(descent);
 189 }
 190
 191
 192 void
 193 ucb1amaf_update(struct uct_policy *p, struct tree *tree, struct tree_node *node,
 194                 enum stone node_color, enum stone player_color,
 195                 struct playout_amafmap *map, struct board *final_board,
 196                 floating_t result)
 197 {
 198         struct ucb1_policy_amaf *b = p->data;
 199         enum stone winner_color = result > 0.5 ? S_BLACK : S_WHITE;
 200         enum stone child_color = stone_other(node_color);
 201
 202 #if 0
 203         struct board bb; bb.size = 9+2;
 204         for (struct tree_node *ni = node; ni; ni = ni->parent)
 205                 fprintf(stderr, "%s ", coord2sstr(ni->coord, &bb));
 206         fprintf(stderr, "[color %d] update result %d (color %d)\n",
 207                         node_color, result, player_color);
 208 #endif
 209
 210         while (node) {
 211                 if (node->parent == NULL)
 212                         assert(tree->root_color == stone_other(child_color));
 213
 214                 if (!b->crit_amaf && !is_pass(node->coord)) {
 215                         stats_add_result(&node->winner_owner, board_at(final_board, node->coord) == winner_color ? 1.0 : 0.0, 1);
 216                         stats_add_result(&node->black_owner, board_at(final_board, node->coord) == S_BLACK ? 1.0 : 0.0, 1);
 217                 }
 218                 stats_add_result(&node->u, result, 1);
 219                 if (amaf_nakade(map->map[node->coord]))
 220                         amaf_op(map->map[node->coord], -);
 221
 222                 /* This loop ignores symmetry considerations, but they should
 223                  * matter only at a point when AMAF doesn't help much. */
 224                 assert(map->game_baselen >= 0);
 225                 for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
 226                         enum stone amaf_color = map->map[ni->coord];
 227                         assert(amaf_color != S_OFFBOARD);
 228                         if (amaf_color == S_NONE)
 229                                 continue;
 230                         if (amaf_nakade(map->map[ni->coord])) {
 231                                 if (!b->check_nakade)
 232                                         continue;
 233                                 unsigned int i;
 234                                 for (i = map->game_baselen; i < map->gamelen; i++)
 235                                         if (map->game[i].coord == ni->coord
 236                                             && map->game[i].color == child_color)
 237                                                 break;
 238                                 if (i == map->gamelen)
 239                                         continue;
 240                                 amaf_color = child_color;
 241                         }
 242
 243                         floating_t nres = result;
 244                         if (amaf_color != child_color) {
 245                                 continue;
 246                         }
 247                         /* For child_color != player_color, we still want
 248                          * to record the result unmodified; in that case,
 249                          * we will correctly negate them at the descend phase. */
 250
 251                         if (b->crit_amaf && !is_pass(node->coord)) {
 252                                 stats_add_result(&ni->winner_owner, board_at(final_board, ni->coord) == winner_color ? 1.0 : 0.0, 1);
 253                                 stats_add_result(&ni->black_owner, board_at(final_board, ni->coord) == S_BLACK ? 1.0 : 0.0, 1);
 254                         }
 255                         stats_add_result(&ni->amaf, nres, 1);
 256
 257 #if 0
 258                         struct board bb; bb.size = 9+2;
 259                         fprintf(stderr, "* %s<%"PRIhash"> -> %s<%"PRIhash"> [%d/%f => %d/%f]\n",
 260                                 coord2sstr(node->coord, &bb), node->hash,
 261                                 coord2sstr(ni->coord, &bb), ni->hash,
 262                                 player_color, result, child_color, nres);
 263 #endif
 264                 }
 265
 266                 if (!is_pass(node->coord)) {
 267                         map->game_baselen--;
 268                 }
 269                 node = node->parent; child_color = stone_other(child_color);
 270         }
 271 }
 272
 273
 274 struct uct_policy *
 275 policy_ucb1amaf_init(struct uct *u, char *arg)
 276 {
 277         struct uct_policy *p = calloc2(1, sizeof(*p));
 278         struct ucb1_policy_amaf *b = calloc2(1, sizeof(*b));
 279         p->uct = u;
 280         p->data = b;
 281         p->choose = uctp_generic_choose;
 282         p->winner = uctp_generic_winner;
 283         p->evaluate = ucb1rave_evaluate;
 284         p->descend = ucb1rave_descend;
 285         p->update = ucb1amaf_update;
 286         p->wants_amaf = true;
 287
 288         b->explore_p = 0; // 0.02 can be also good on 19x19 with prior=eqex=40
 289         b->equiv_rave = 3000;
 290         b->fpu = INFINITY;
 291         b->check_nakade = true;
 292         b->sylvain_rave = true;
 293         b->ltree_rave = 0.75f;
 294
 295         b->crit_rave = 1.0f;
 296         b->crit_min_playouts = 2000;
 297         b->crit_negative = 1;
 298         b->crit_amaf = 0;
 299
 300         b->root_virtual_win = -1;
 301         b->vwin_min_playouts = 1000;
 302
 303         if (arg) {
 304                 char *optspec, *next = arg;
 305                 while (*next) {
 306                         optspec = next;
 307                         next += strcspn(next, ":");
 308                         if (*next) { *next++ = 0; } else { *next = 0; }
 309
 310                         char *optname = optspec;
 311                         char *optval = strchr(optspec, '=');
 312                         if (optval) *optval++ = 0;
 313
 314                         if (!strcasecmp(optname, "explore_p")) {
 315                                 b->explore_p = atof(optval);
 316                         } else if (!strcasecmp(optname, "fpu") && optval) {
 317                                 b->fpu = atof(optval);
 318                         } else if (!strcasecmp(optname, "equiv_rave") && optval) {
 319                                 b->equiv_rave = atof(optval);
 320                         } else if (!strcasecmp(optname, "sylvain_rave")) {
 321                                 b->sylvain_rave = !optval || *optval == '1';
 322                         } else if (!strcasecmp(optname, "check_nakade")) {
 323                                 b->check_nakade = !optval || *optval == '1';
 324                         } else if (!strcasecmp(optname, "ltree_rave") && optval) {
 325                                 b->ltree_rave = atof(optval);
 326                         } else if (!strcasecmp(optname, "crit_rave") && optval) {
 327                                 b->crit_rave = atof(optval);
 328                         } else if (!strcasecmp(optname, "crit_min_playouts") && optval) {
 329                                 b->crit_min_playouts = atoi(optval);
 330                         } else if (!strcasecmp(optname, "crit_negative")) {
 331                                 b->crit_negative = !optval || *optval == '1';
 332                         } else if (!strcasecmp(optname, "crit_amaf")) {
 333                                 b->crit_amaf = !optval || *optval == '1';
 334                         } else if (!strcasecmp(optname, "virtual_win") && optval) {
 335                                 b->virtual_win = atoi(optval);
 336                         } else if (!strcasecmp(optname, "root_virtual_win") && optval) {
 337                                 b->root_virtual_win = atoi(optval);
 338                         } else if (!strcasecmp(optname, "vwin_min_playouts") && optval) {
 339                                 b->vwin_min_playouts = atoi(optval);
 340                         } else {
 341                                 fprintf(stderr, "ucb1amaf: Invalid policy argument %s or missing value\n",
 342                                         optname);
 343                                 exit(1);
 344                         }
 345                 }
 346         }
 347         if (b->root_virtual_win < 0)
 348                 b->root_virtual_win = b->virtual_win;
 349
 350         return p;
 351 }