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