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         /* First Play Urgency - if set to less than infinity (the MoGo paper
  24          * above reports 1.0 as the best), new branches are explored only
  25          * if none of the existing ones has higher urgency than fpu. */
  26         floating_t fpu;
  27         unsigned int equiv_rave;
  28         bool check_nakade;
  29         bool sylvain_rave;
  30         /* Coefficient of local tree values embedded in RAVE. */
  31         floating_t ltree_rave;
  32         /* Coefficient of criticality embedded in RAVE. */
  33         floating_t crit_rave;
  34         int crit_min_playouts;
  35         bool crit_negative;
  36         bool crit_amaf;
  37 };
  38
  39
  40 static inline floating_t fast_sqrt(unsigned int x)
  41 {
  42         static const floating_t table[] = {
  43                 0, 1, 1.41421356237309504880, 1.73205080756887729352,
  44                 2.00000000000000000000, 2.23606797749978969640,
  45                 2.44948974278317809819, 2.64575131106459059050,
  46                 2.82842712474619009760, 3.00000000000000000000,
  47                 3.16227766016837933199, 3.31662479035539984911,
  48                 3.46410161513775458705, 3.60555127546398929311,
  49                 3.74165738677394138558, 3.87298334620741688517,
  50                 4.00000000000000000000, 4.12310562561766054982,
  51                 4.24264068711928514640, 4.35889894354067355223,
  52                 4.47213595499957939281, 4.58257569495584000658,
  53                 4.69041575982342955456, 4.79583152331271954159,
  54                 4.89897948556635619639, 5.00000000000000000000,
  55                 5.09901951359278483002, 5.19615242270663188058,
  56                 5.29150262212918118100, 5.38516480713450403125,
  57                 5.47722557505166113456, 5.56776436283002192211,
  58                 5.65685424949238019520, 5.74456264653802865985,
  59                 5.83095189484530047087, 5.91607978309961604256,
  60                 6.00000000000000000000, 6.08276253029821968899,
  61                 6.16441400296897645025, 6.24499799839839820584,
  62                 6.32455532033675866399, 6.40312423743284868648,
  63                 6.48074069840786023096, 6.55743852430200065234,
  64                 6.63324958071079969822, 6.70820393249936908922,
  65                 6.78232998312526813906, 6.85565460040104412493,
  66                 6.92820323027550917410, 7.00000000000000000000,
  67                 7.07106781186547524400, 7.14142842854284999799,
  68                 7.21110255092797858623, 7.28010988928051827109,
  69                 7.34846922834953429459, 7.41619848709566294871,
  70                 7.48331477354788277116, 7.54983443527074969723,
  71                 7.61577310586390828566, 7.68114574786860817576,
  72                 7.74596669241483377035, 7.81024967590665439412,
  73                 7.87400787401181101968, 7.93725393319377177150,
  74         };
  75         if (x < sizeof(table) / sizeof(*table)) {
  76                 return table[x];
  77         } else {
  78                 return sqrt(x);
  79         }
  80 }
  81
  82 #define LTREE_DEBUG if (0)
  83 static floating_t inline
  84 ucb1rave_evaluate(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity)
  85 {
  86         struct ucb1_policy_amaf *b = p->data;
  87         struct tree_node *node = descent->node;
  88         struct tree_node *lnode = descent->lnode;
  89
  90         struct move_stats n = node->u, r = node->amaf;
  91         if (p->uct->amaf_prior) {
  92                 stats_merge(&r, &node->prior);
  93         } else {
  94                 stats_merge(&n, &node->prior);
  95         }
  96
  97         /* Local tree heuristics. */
  98         if (p->uct->local_tree && b->ltree_rave > 0 && lnode) {
  99                 struct move_stats l = lnode->u;
 100                 l.playouts = ((floating_t) l.playouts) * b->ltree_rave / LTREE_PLAYOUTS_MULTIPLIER;
 101                 LTREE_DEBUG fprintf(stderr, "[ltree] adding [%s] %f%%%d to [%s] RAVE %f%%%d\n",
 102                         coord2sstr(lnode->coord, tree->board), l.value, l.playouts,
 103                         coord2sstr(node->coord, tree->board), r.value, r.playouts);
 104                 stats_merge(&r, &l);
 105         }
 106
 107         /* Criticality heuristics. */
 108         if (b->crit_rave > 0 && node->u.playouts > b->crit_min_playouts) {
 109                 floating_t crit = tree_node_criticality(tree, node);
 110                 if (b->crit_negative || crit > 0) {
 111                         struct move_stats c = {
 112                                 .value = tree_node_get_value(tree, parity, 1.0f),
 113                                 .playouts = crit * r.playouts * b->crit_rave
 114                         };
 115                         LTREE_DEBUG fprintf(stderr, "[crit] adding %f%%%d to [%s] RAVE %f%%%d\n",
 116                                 c.value, c.playouts,
 117                                 coord2sstr(node->coord, tree->board), r.value, r.playouts);
 118                         stats_merge(&r, &c);
 119                 }
 120         }
 121
 122
 123         floating_t value = 0;
 124         if (n.playouts) {
 125                 if (r.playouts) {
 126                         /* At the beginning, beta is at 1 and RAVE is used.
 127                          * At b->equiv_rate, beta is at 1/3 and gets steeper on. */
 128                         floating_t beta;
 129                         if (b->sylvain_rave) {
 130                                 beta = (floating_t) r.playouts / (r.playouts + n.playouts
 131                                         + (floating_t) n.playouts * r.playouts / b->equiv_rave);
 132                         } else {
 133                                 /* XXX: This can be cached in descend; but we don't use this by default. */
 134                                 beta = sqrt(b->equiv_rave / (3 * node->parent->u.playouts + b->equiv_rave));
 135                         }
 136
 137                         value = beta * r.value + (1.f - beta) * n.value;
 138                 } else {
 139                         value = n.value;
 140                 }
 141         } else if (r.playouts) {
 142                 value = r.value;
 143         }
 144         descent->value.playouts = r.playouts + n.playouts;
 145         descent->value.value = value;
 146         return tree_node_get_value(tree, parity, value);
 147 }
 148
 149 void
 150 ucb1rave_descend(struct uct_policy *p, struct tree *tree, struct uct_descent *descent, int parity, bool allow_pass)
 151 {
 152         struct ucb1_policy_amaf *b = p->data;
 153         floating_t nconf = 1.f;
 154         if (b->explore_p > 0)
 155                 nconf = sqrt(log(descent->node->u.playouts + descent->node->prior.playouts));
 156
 157         uctd_try_node_children(tree, descent, allow_pass, parity, p->uct->tenuki_d, di, urgency) {
 158                 struct tree_node *ni = di.node;
 159                 urgency = ucb1rave_evaluate(p, tree, &di, parity);
 160
 161                 if (ni->u.playouts > 0 && b->explore_p > 0) {
 162                         urgency += b->explore_p * nconf / fast_sqrt(ni->u.playouts);
 163
 164                 } else if (ni->u.playouts + ni->amaf.playouts + ni->prior.playouts == 0) {
 165                         /* assert(!u->even_eqex); */
 166                         urgency = b->fpu;
 167                 }
 168         } uctd_set_best_child(di, urgency);
 169
 170         uctd_get_best_child(descent);
 171 }
 172
 173
 174 void
 175 ucb1amaf_update(struct uct_policy *p, struct tree *tree, struct tree_node *node,
 176                 enum stone node_color, enum stone player_color,
 177                 struct playout_amafmap *map, struct board *final_board,
 178                 floating_t result)
 179 {
 180         struct ucb1_policy_amaf *b = p->data;
 181         enum stone winner_color = result > 0.5 ? S_BLACK : S_WHITE;
 182         enum stone child_color = stone_other(node_color);
 183
 184 #if 0
 185         struct board bb; bb.size = 9+2;
 186         for (struct tree_node *ni = node; ni; ni = ni->parent)
 187                 fprintf(stderr, "%s ", coord2sstr(ni->coord, &bb));
 188         fprintf(stderr, "[color %d] update result %d (color %d)\n",
 189                         node_color, result, player_color);
 190 #endif
 191
 192         while (node) {
 193                 if (node->parent == NULL)
 194                         assert(tree->root_color == stone_other(child_color));
 195
 196                 if (!b->crit_amaf) {
 197                         stats_add_result(&node->winner_owner, board_at(final_board, node->coord) == winner_color ? 1.0 : 0.0, 1);
 198                         stats_add_result(&node->black_owner, board_at(final_board, node->coord) == S_BLACK ? 1.0 : 0.0, 1);
 199                 }
 200                 stats_add_result(&node->u, result, 1);
 201                 if (amaf_nakade(map->map[node->coord]))
 202                         amaf_op(map->map[node->coord], -);
 203
 204                 /* This loop ignores symmetry considerations, but they should
 205                  * matter only at a point when AMAF doesn't help much. */
 206                 assert(map->game_baselen >= 0);
 207                 for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
 208                         enum stone amaf_color = map->map[ni->coord];
 209                         assert(amaf_color != S_OFFBOARD);
 210                         if (amaf_color == S_NONE)
 211                                 continue;
 212                         if (amaf_nakade(map->map[ni->coord])) {
 213                                 if (!b->check_nakade)
 214                                         continue;
 215                                 unsigned int i;
 216                                 for (i = map->game_baselen; i < map->gamelen; i++)
 217                                         if (map->game[i].coord == ni->coord
 218                                             && map->game[i].color == child_color)
 219                                                 break;
 220                                 if (i == map->gamelen)
 221                                         continue;
 222                                 amaf_color = child_color;
 223                         }
 224
 225                         floating_t nres = result;
 226                         if (amaf_color != child_color) {
 227                                 continue;
 228                         }
 229                         /* For child_color != player_color, we still want
 230                          * to record the result unmodified; in that case,
 231                          * we will correctly negate them at the descend phase. */
 232
 233                         if (b->crit_amaf) {
 234                                 stats_add_result(&ni->winner_owner, board_at(final_board, ni->coord) == winner_color ? 1.0 : 0.0, 1);
 235                                 stats_add_result(&ni->black_owner, board_at(final_board, ni->coord) == S_BLACK ? 1.0 : 0.0, 1);
 236                         }
 237                         stats_add_result(&ni->amaf, nres, 1);
 238
 239 #if 0
 240                         struct board bb; bb.size = 9+2;
 241                         fprintf(stderr, "* %s<%"PRIhash"> -> %s<%"PRIhash"> [%d/%f => %d/%f]\n",
 242                                 coord2sstr(node->coord, &bb), node->hash,
 243                                 coord2sstr(ni->coord, &bb), ni->hash,
 244                                 player_color, result, child_color, nres);
 245 #endif
 246                 }
 247
 248                 if (!is_pass(node->coord)) {
 249                         map->game_baselen--;
 250                 }
 251                 node = node->parent; child_color = stone_other(child_color);
 252         }
 253 }
 254
 255
 256 struct uct_policy *
 257 policy_ucb1amaf_init(struct uct *u, char *arg)
 258 {
 259         struct uct_policy *p = calloc2(1, sizeof(*p));
 260         struct ucb1_policy_amaf *b = calloc2(1, sizeof(*b));
 261         p->uct = u;
 262         p->data = b;
 263         p->choose = uctp_generic_choose;
 264         p->winner = uctp_generic_winner;
 265         p->evaluate = ucb1rave_evaluate;
 266         p->descend = ucb1rave_descend;
 267         p->update = ucb1amaf_update;
 268         p->wants_amaf = true;
 269
 270         b->explore_p = 0; // 0.02 can be also good on 19x19 with prior=eqex=40
 271         b->equiv_rave = 3000;
 272         b->fpu = INFINITY;
 273         b->check_nakade = true;
 274         b->sylvain_rave = true;
 275         b->ltree_rave = 0.75f;
 276         b->crit_rave = 2.0f;
 277         b->crit_min_playouts = 32;
 278         b->crit_negative = 1;
 279         b->crit_amaf = 1;
 280
 281         if (arg) {
 282                 char *optspec, *next = arg;
 283                 while (*next) {
 284                         optspec = next;
 285                         next += strcspn(next, ":");
 286                         if (*next) { *next++ = 0; } else { *next = 0; }
 287
 288                         char *optname = optspec;
 289                         char *optval = strchr(optspec, '=');
 290                         if (optval) *optval++ = 0;
 291
 292                         if (!strcasecmp(optname, "explore_p")) {
 293                                 b->explore_p = atof(optval);
 294                         } else if (!strcasecmp(optname, "fpu") && optval) {
 295                                 b->fpu = atof(optval);
 296                         } else if (!strcasecmp(optname, "equiv_rave") && optval) {
 297                                 b->equiv_rave = atof(optval);
 298                         } else if (!strcasecmp(optname, "sylvain_rave")) {
 299                                 b->sylvain_rave = !optval || *optval == '1';
 300                         } else if (!strcasecmp(optname, "check_nakade")) {
 301                                 b->check_nakade = !optval || *optval == '1';
 302                         } else if (!strcasecmp(optname, "ltree_rave") && optval) {
 303                                 b->ltree_rave = atof(optval);
 304                         } else if (!strcasecmp(optname, "crit_rave") && optval) {
 305                                 b->crit_rave = atof(optval);
 306                         } else if (!strcasecmp(optname, "crit_min_playouts") && optval) {
 307                                 b->crit_min_playouts = atoi(optval);
 308                         } else if (!strcasecmp(optname, "crit_negative")) {
 309                                 b->crit_negative = !optval || *optval == '1';
 310                         } else if (!strcasecmp(optname, "crit_amaf")) {
 311                                 b->crit_amaf = !optval || *optval == '1';
 312                         } else {
 313                                 fprintf(stderr, "ucb1amaf: Invalid policy argument %s or missing value\n",
 314                                         optname);
 315                                 exit(1);
 316                         }
 317                 }
 318         }
 319
 320         return p;
 321 }