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         /* First Play Urgency - if set to less than infinity (the MoGo paper
  26          * above reports 1.0 as the best), new branches are explored only
  27          * if none of the existing ones has higher urgency than fpu. */
  28         floating_t fpu;
  29         unsigned int equiv_rave;
  30         bool check_nakade;
  31         bool sylvain_rave;
  32         /* Coefficient of local tree values embedded in RAVE. */
  33         floating_t ltree_rave;
  34         /* Coefficient of criticality embedded in RAVE. */
  35         floating_t crit_rave;
  36         int crit_min_playouts;
  37         bool crit_negative;
  38         bool crit_negflip;
  39         bool crit_amaf;
  40         bool crit_lvalue;
  41 };
  42
  43
  44 static inline floating_t fast_sqrt(unsigned int x)
  45 {
  46         static const floating_t table[] = {
  47                 0, 1, 1.41421356237309504880, 1.73205080756887729352,
  48                 2.00000000000000000000, 2.23606797749978969640,
  49                 2.44948974278317809819, 2.64575131106459059050,
  50                 2.82842712474619009760, 3.00000000000000000000,
  51                 3.16227766016837933199, 3.31662479035539984911,
  52                 3.46410161513775458705, 3.60555127546398929311,
  53                 3.74165738677394138558, 3.87298334620741688517,
  54                 4.00000000000000000000, 4.12310562561766054982,
  55                 4.24264068711928514640, 4.35889894354067355223,
  56                 4.47213595499957939281, 4.58257569495584000658,
  57                 4.69041575982342955456, 4.79583152331271954159,
  58                 4.89897948556635619639, 5.00000000000000000000,
  59                 5.09901951359278483002, 5.19615242270663188058,
  60                 5.29150262212918118100, 5.38516480713450403125,
  61                 5.47722557505166113456, 5.56776436283002192211,
  62                 5.65685424949238019520, 5.74456264653802865985,
  63                 5.83095189484530047087, 5.91607978309961604256,
  64                 6.00000000000000000000, 6.08276253029821968899,
  65                 6.16441400296897645025, 6.24499799839839820584,
  66                 6.32455532033675866399, 6.40312423743284868648,
  67                 6.48074069840786023096, 6.55743852430200065234,
  68                 6.63324958071079969822, 6.70820393249936908922,
  69                 6.78232998312526813906, 6.85565460040104412493,
  70                 6.92820323027550917410, 7.00000000000000000000,
  71                 7.07106781186547524400, 7.14142842854284999799,
  72                 7.21110255092797858623, 7.28010988928051827109,
  73                 7.34846922834953429459, 7.41619848709566294871,
  74                 7.48331477354788277116, 7.54983443527074969723,
  75                 7.61577310586390828566, 7.68114574786860817576,
  76                 7.74596669241483377035, 7.81024967590665439412,
  77                 7.87400787401181101968, 7.93725393319377177150,
  78         };
  79         if (x < sizeof(table) / sizeof(*table)) {
  80                 return table[x];
  81         } else {
  82                 return sqrt(x);
  83         }
  84 }
  85
  86 #define URAVE_DEBUG if (0)
  87 static floating_t inline
  88 ucb1rave_evaluate(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity)
  89 {
  90         struct ucb1_policy_amaf *b = p->data;
  91         struct tree_node *node = descent->node;
  92         struct tree_node *lnode = descent->lnode;
  93
  94         struct move_stats n = node->u, r = node->amaf;
  95         if (p->uct->amaf_prior) {
  96                 stats_merge(&r, &node->prior);
  97         } else {
  98                 stats_merge(&n, &node->prior);
  99         }
 100
 101         /* Local tree heuristics. */
 102         assert(!lnode || lnode->parent);
 103         if (p->uct->local_tree && b->ltree_rave > 0 && lnode
 104             && (p->uct->local_tree_rootchoose || lnode->parent->parent)) {
 105                 struct move_stats l = lnode->u;
 106                 l.playouts = ((floating_t) l.playouts) * b->ltree_rave / LTREE_PLAYOUTS_MULTIPLIER;
 107                 URAVE_DEBUG fprintf(stderr, "[ltree] adding [%s] %f%%%d to [%s] RAVE %f%%%d\n",
 108                         coord2sstr(node_coord(lnode), tree->board), l.value, l.playouts,
 109                         coord2sstr(node_coord(node), tree->board), r.value, r.playouts);
 110                 stats_merge(&r, &l);
 111         }
 112
 113         /* Criticality heuristics. */
 114         if (b->crit_rave > 0 && node->u.playouts > b->crit_min_playouts) {
 115                 floating_t crit = tree_node_criticality(tree, node);
 116                 if (b->crit_negative || crit > 0) {
 117                         floating_t val = 1.0f;
 118                         if (b->crit_negflip && crit < 0) {
 119                                 val = 0;
 120                                 crit = -crit;
 121                         }
 122                         struct move_stats c = {
 123                                 .value = tree_node_get_value(tree, parity, val),
 124                                 .playouts = crit * r.playouts * b->crit_rave
 125                         };
 126                         URAVE_DEBUG fprintf(stderr, "[crit] adding %f%%%d to [%s] RAVE %f%%%d\n",
 127                                 c.value, c.playouts,
 128                                 coord2sstr(node_coord(node), tree->board), r.value, r.playouts);
 129                         stats_merge(&r, &c);
 130                 }
 131         }
 132
 133
 134         floating_t value = 0;
 135         if (n.playouts) {
 136                 if (r.playouts) {
 137                         /* At the beginning, beta is at 1 and RAVE is used.
 138                          * At b->equiv_rate, beta is at 1/3 and gets steeper on. */
 139                         floating_t beta;
 140                         if (b->sylvain_rave) {
 141                                 beta = (floating_t) r.playouts / (r.playouts + n.playouts
 142                                         + (floating_t) n.playouts * r.playouts / b->equiv_rave);
 143                         } else {
 144                                 /* XXX: This can be cached in descend; but we don't use this by default. */
 145                                 beta = sqrt(b->equiv_rave / (3 * node->parent->u.playouts + b->equiv_rave));
 146                         }
 147
 148                         value = beta * r.value + (1.f - beta) * n.value;
 149                         URAVE_DEBUG fprintf(stderr, "\t%s value = %f * %f + (1 - %f) * %f (prior %f)\n",
 150                                 coord2sstr(node_coord(node), tree->board), beta, r.value, beta, n.value, node->prior.value);
 151                 } else {
 152                         value = n.value;
 153                         URAVE_DEBUG fprintf(stderr, "\t%s value = %f (prior %f)\n",
 154                                 coord2sstr(node_coord(node), tree->board), n.value, node->prior.value);
 155                 }
 156         } else if (r.playouts) {
 157                 value = r.value;
 158                 URAVE_DEBUG fprintf(stderr, "\t%s value = rave %f (prior %f)\n",
 159                         coord2sstr(node_coord(node), tree->board), r.value, node->prior.value);
 160         }
 161         descent->value.playouts = r.playouts + n.playouts;
 162         descent->value.value = value;
 163         return tree_node_get_value(tree, parity, value);
 164 }
 165
 166 void
 167 ucb1rave_descend(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity, bool allow_pass)
 168 {
 169         struct ucb1_policy_amaf *b = p->data;
 170         floating_t nconf = 1.f;
 171         if (b->explore_p > 0)
 172                 nconf = sqrt(log(descent->node->u.playouts + descent->node->prior.playouts));
 173
 174         uctd_try_node_children(tree, descent, allow_pass, parity, p->uct->tenuki_d, di, urgency) {
 175                 struct tree_node *ni = di.node;
 176                 urgency = ucb1rave_evaluate(p, tree, &di, parity);
 177
 178                 if (ni->u.playouts > 0 && b->explore_p > 0) {
 179                         urgency += b->explore_p * nconf / fast_sqrt(ni->u.playouts);
 180
 181                 } else if (ni->u.playouts + ni->amaf.playouts + ni->prior.playouts == 0) {
 182                         /* assert(!u->even_eqex); */
 183                         urgency = b->fpu;
 184                 }
 185         } uctd_set_best_child(di, urgency);
 186
 187         uctd_get_best_child(descent);
 188 }
 189
 190
 191 void
 192 ucb1amaf_update(struct uct_policy *p, struct tree *tree, struct tree_node *node,
 193                 enum stone node_color, enum stone player_color,
 194                 struct playout_amafmap *map, struct board *final_board,
 195                 floating_t result)
 196 {
 197         struct ucb1_policy_amaf *b = p->data;
 198         enum stone winner_color = result > 0.5 ? S_BLACK : S_WHITE;
 199
 200         /* Record of the random playout - for each intersection coord,
 201          * first_move[coord] is the index map->game of the first move
 202          * at this coordinate, or INT_MAX if the move was not played.
 203          * The parity gives the color of this move.
 204          */
 205         int first_map[board_size2(final_board)+1];
 206         int *first_move = &first_map[1]; // +1 for pass
 207
 208 #if 0
 209         struct board bb; bb.size = 9+2;
 210         for (struct tree_node *ni = node; ni; ni = ni->parent)
 211                 fprintf(stderr, "%s ", coord2sstr(node_coord(ni), &bb));
 212         fprintf(stderr, "[color %d] update result %d (color %d)\n",
 213                         node_color, result, player_color);
 214 #endif
 215
 216         /* Initialize first_move */
 217         for (int i = pass; i < board_size2(final_board); i++) first_move[i] = INT_MAX;
 218         int move;
 219         assert(map->gamelen > 0);
 220         for (move = map->gamelen - 1; move >= map->game_baselen; move--)
 221                 first_move[map->game[move]] = move;
 222
 223         while (node) {
 224                 if (!b->crit_amaf && !is_pass(node_coord(node))) {
 225                         stats_add_result(&node->winner_owner, board_local_value(b->crit_lvalue, final_board, node_coord(node), winner_color), 1);
 226                         stats_add_result(&node->black_owner, board_local_value(b->crit_lvalue, final_board, node_coord(node), S_BLACK), 1);
 227                 }
 228                 stats_add_result(&node->u, result, 1);
 229
 230                 /* This loop ignores symmetry considerations, but they should
 231                  * matter only at a point when AMAF doesn't help much. */
 232                 assert(map->game_baselen >= 0);
 233                 for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
 234                         /* Use the child move only if it was first played by the same color. */
 235                         int first = first_move[node_coord(ni)];
 236                         if (first == INT_MAX || (first & 1) == (move & 1))
 237                                 continue;
 238
 239                         if (b->crit_amaf && !is_pass(node_coord(node))) {
 240                                 stats_add_result(&ni->winner_owner, board_local_value(b->crit_lvalue, final_board, node_coord(ni), winner_color), 1);
 241                                 stats_add_result(&ni->black_owner, board_local_value(b->crit_lvalue, final_board, node_coord(ni), S_BLACK), 1);
 242                         }
 243                         stats_add_result(&ni->amaf, result, 1);
 244 #if 0
 245                         struct board bb; bb.size = 9+2;
 246                         fprintf(stderr, "* %s<%"PRIhash"> -> %s<%"PRIhash"> [%d/%f => %d/%f]\n",
 247                                 coord2sstr(node_coord(node), &bb), node->hash,
 248                                 coord2sstr(node_coord(ni), &bb), ni->hash,
 249                                 player_color, result, move, result);
 250 #endif
 251                 }
 252                 if (node->parent) {
 253                         assert(move >= 0 && map->game[move] == node_coord(node));
 254                         first_move[node_coord(node)] = move;
 255                         move--;
 256                 }
 257                 node = node->parent;
 258         }
 259 }
 260
 261
 262 struct uct_policy *
 263 policy_ucb1amaf_init(struct uct *u, char *arg)
 264 {
 265         struct uct_policy *p = calloc2(1, sizeof(*p));
 266         struct ucb1_policy_amaf *b = calloc2(1, sizeof(*b));
 267         p->uct = u;
 268         p->data = b;
 269         p->choose = uctp_generic_choose;
 270         p->winner = uctp_generic_winner;
 271         p->evaluate = ucb1rave_evaluate;
 272         p->descend = ucb1rave_descend;
 273         p->update = ucb1amaf_update;
 274         p->wants_amaf = true;
 275
 276         b->explore_p = 0; // 0.02 can be also good on 19x19 with prior=eqex=40
 277         b->equiv_rave = 3000;
 278         b->fpu = INFINITY;
 279         b->check_nakade = true;
 280         b->sylvain_rave = true;
 281         b->ltree_rave = 0.75f;
 282
 283         b->crit_rave = 1.0f;
 284         b->crit_min_playouts = 2000;
 285         b->crit_negative = 1;
 286         b->crit_amaf = 0;
 287
 288         if (arg) {
 289                 char *optspec, *next = arg;
 290                 while (*next) {
 291                         optspec = next;
 292                         next += strcspn(next, ":");
 293                         if (*next) { *next++ = 0; } else { *next = 0; }
 294
 295                         char *optname = optspec;
 296                         char *optval = strchr(optspec, '=');
 297                         if (optval) *optval++ = 0;
 298
 299                         if (!strcasecmp(optname, "explore_p")) {
 300                                 b->explore_p = atof(optval);
 301                         } else if (!strcasecmp(optname, "fpu") && optval) {
 302                                 b->fpu = atof(optval);
 303                         } else if (!strcasecmp(optname, "equiv_rave") && optval) {
 304                                 b->equiv_rave = atof(optval);
 305                         } else if (!strcasecmp(optname, "sylvain_rave")) {
 306                                 b->sylvain_rave = !optval || *optval == '1';
 307                         } else if (!strcasecmp(optname, "check_nakade")) {
 308                                 b->check_nakade = !optval || *optval == '1';
 309                         } else if (!strcasecmp(optname, "ltree_rave") && optval) {
 310                                 b->ltree_rave = atof(optval);
 311                         } else if (!strcasecmp(optname, "crit_rave") && optval) {
 312                                 b->crit_rave = atof(optval);
 313                         } else if (!strcasecmp(optname, "crit_min_playouts") && optval) {
 314                                 b->crit_min_playouts = atoi(optval);
 315                         } else if (!strcasecmp(optname, "crit_negative")) {
 316                                 b->crit_negative = !optval || *optval == '1';
 317                         } else if (!strcasecmp(optname, "crit_negflip")) {
 318                                 b->crit_negflip = !optval || *optval == '1';
 319                         } else if (!strcasecmp(optname, "crit_amaf")) {
 320                                 b->crit_amaf = !optval || *optval == '1';
 321                         } else if (!strcasecmp(optname, "crit_lvalue")) {
 322                                 b->crit_lvalue = !optval || *optval == '1';
 323                         } else {
 324                                 fprintf(stderr, "ucb1amaf: Invalid policy argument %s or missing value\n",
 325                                         optname);
 326                                 exit(1);
 327                         }
 328                 }
 329         }
 330
 331         return p;
 332 }