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