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