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
  14 /* This implements the UCB1 policy with an extra AMAF heuristics. */
  15
  16 struct ucb1_policy_amaf {
  17         /* This is what the Modification of UCT with Patterns in Monte Carlo Go
  18          * paper calls 'p'. Original UCB has this on 2, but this seems to
  19          * produce way too wide searches; reduce this to get deeper and
  20          * narrower readouts - try 0.2. */
  21         float explore_p;
  22         /* First Play Urgency - if set to less than infinity (the MoGo paper
  23          * above reports 1.0 as the best), new branches are explored only
  24          * if none of the existing ones has higher urgency than fpu. */
  25         float fpu;
  26         int urg_randoma, urg_randomm;
  27         int equiv_rave;
  28         bool both_colors;
  29         bool check_nakade;
  30         bool sylvain_rave;
  31 };
  32
  33
  34 struct tree_node *ucb1_choose(struct uct_policy *p, struct tree_node *node, struct board *b, enum stone color);
  35
  36 struct tree_node *ucb1_descend(struct uct_policy *p, struct tree *tree, struct tree_node *node, int parity, bool allow_pass);
  37
  38
  39 static inline float fast_sqrt(int x)
  40 {
  41         static const float table[] = {
  42                 0, 1, 1.41421356237309504880, 1.73205080756887729352,
  43                 2.00000000000000000000, 2.23606797749978969640,
  44                 2.44948974278317809819, 2.64575131106459059050,
  45                 2.82842712474619009760, 3.00000000000000000000,
  46                 3.16227766016837933199, 3.31662479035539984911,
  47                 3.46410161513775458705, 3.60555127546398929311,
  48                 3.74165738677394138558, 3.87298334620741688517,
  49                 4.00000000000000000000, 4.12310562561766054982,
  50                 4.24264068711928514640, 4.35889894354067355223,
  51                 4.47213595499957939281, 4.58257569495584000658,
  52                 4.69041575982342955456, 4.79583152331271954159,
  53                 4.89897948556635619639, 5.00000000000000000000,
  54                 5.09901951359278483002, 5.19615242270663188058,
  55                 5.29150262212918118100, 5.38516480713450403125,
  56                 5.47722557505166113456, 5.56776436283002192211,
  57                 5.65685424949238019520, 5.74456264653802865985,
  58                 5.83095189484530047087, 5.91607978309961604256,
  59                 6.00000000000000000000, 6.08276253029821968899,
  60                 6.16441400296897645025, 6.24499799839839820584,
  61                 6.32455532033675866399, 6.40312423743284868648,
  62                 6.48074069840786023096, 6.55743852430200065234,
  63                 6.63324958071079969822, 6.70820393249936908922,
  64                 6.78232998312526813906, 6.85565460040104412493,
  65                 6.92820323027550917410, 7.00000000000000000000,
  66                 7.07106781186547524400, 7.14142842854284999799,
  67                 7.21110255092797858623, 7.28010988928051827109,
  68                 7.34846922834953429459, 7.41619848709566294871,
  69                 7.48331477354788277116, 7.54983443527074969723,
  70                 7.61577310586390828566, 7.68114574786860817576,
  71                 7.74596669241483377035, 7.81024967590665439412,
  72                 7.87400787401181101968, 7.93725393319377177150,
  73         };
  74         //printf("sqrt %d\n", x);
  75         if (x < sizeof(table) / sizeof(*table)) {
  76                 return table[x];
  77         } else {
  78                 return sqrt(x);
  79         }
  80 }
  81
  82 struct tree_node *
  83 ucb1rave_descend(struct uct_policy *p, struct tree *tree, struct tree_node *node, int parity, bool allow_pass)
  84 {
  85         struct ucb1_policy_amaf *b = p->data;
  86         float beta = 0;
  87         float nconf = 1.f;
  88         if (b->explore_p > 0)
  89                 nconf = sqrt(log(node->u.playouts + node->prior.playouts));
  90
  91         if (!b->sylvain_rave)
  92                 beta = sqrt(b->equiv_rave / (3 * node->u.playouts + b->equiv_rave));
  93
  94         // XXX: Stack overflow danger on big boards?
  95         struct tree_node *nbest[512] = { node->children }; int nbests = 1;
  96         float best_urgency = -9999;
  97
  98         for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
  99                 /* Do not consider passing early. */
 100                 if (unlikely(!allow_pass && is_pass(ni->coord)))
 101                         continue;
 102
 103                 /* TODO: Exploration? */
 104
 105                 struct move_stats n = ni->u, r = ni->amaf;
 106                 if (p->uct->amaf_prior) {
 107                         stats_merge(&r, &ni->prior);
 108                 } else {
 109                         stats_merge(&n, &ni->prior);
 110                 }
 111                 if (tree_parity(tree, parity) < 0) {
 112                         stats_reverse_parity(&n);
 113                         stats_reverse_parity(&r);
 114                 }
 115
 116                 float urgency;
 117                 if (n.playouts) {
 118                         if (r.playouts) {
 119                                 /* At the beginning, beta is at 1 and RAVE is used.
 120                                  * At b->equiv_rate, beta is at 1/3 and gets steeper on. */
 121                                 if (b->sylvain_rave)
 122                                         beta = (float) r.playouts / (r.playouts + n.playouts + n.playouts * r.playouts / b->equiv_rave);
 123 #if 0
 124                                 //if (node->coord == 7*11+4) // D7
 125                                 fprintf(stderr, "[beta %f = %d / (%d + %d + %f)]\n",
 126                                         beta, rgames, rgames, ngames, ngames * rgames / b->equiv_rave);
 127 #endif
 128                                 urgency = beta * r.value + (1.f - beta) * n.value;
 129                         } else {
 130                                 urgency = n.value;
 131                         }
 132
 133                         if (b->explore_p > 0)
 134                                 urgency += b->explore_p * nconf / fast_sqrt(n.playouts);
 135                 } else if (r.playouts) {
 136                         urgency = r.value;
 137                 } else {
 138                         /* assert(!u->even_eqex); */
 139                         urgency = b->fpu;
 140                 }
 141
 142 #if 0
 143                 struct board bb; bb.size = 11;
 144                 //if (node->coord == 7*11+4) // D7
 145                 fprintf(stderr, "%s<%lld>-%s<%lld> urgency %f (r %d / %d + e = %f, n %d / %d + e = %f)\n",
 146                         coord2sstr(ni->parent->coord, &bb), ni->parent->hash,
 147                         coord2sstr(ni->coord, &bb), ni->hash, urgency,
 148                         rwins, rgames, rval, nwins, ngames, nval);
 149 #endif
 150                 if (b->urg_randoma)
 151                         urgency += (float)(fast_random(b->urg_randoma) - b->urg_randoma / 2) / 1000;
 152                 if (b->urg_randomm)
 153                         urgency *= (float)(fast_random(b->urg_randomm) + 5) / b->urg_randomm;
 154
 155                 if (urgency - best_urgency > __FLT_EPSILON__) { // urgency > best_urgency
 156                         best_urgency = urgency; nbests = 0;
 157                 }
 158                 if (urgency - best_urgency > -__FLT_EPSILON__) { // urgency >= best_urgency
 159                         /* We want to always choose something else than a pass
 160                          * in case of a tie. pass causes degenerative behaviour. */
 161                         if (nbests == 1 && is_pass(nbest[0]->coord)) {
 162                                 nbests--;
 163                         }
 164                         nbest[nbests++] = ni;
 165                 }
 166         }
 167 #if 0
 168         struct board bb; bb.size = 11;
 169         fprintf(stderr, "RESULT [%s %d: ", coord2sstr(node->coord, &bb), nbests);
 170         for (int zz = 0; zz < nbests; zz++)
 171                 fprintf(stderr, "%s", coord2sstr(nbest[zz]->coord, &bb));
 172         fprintf(stderr, "]\n");
 173 #endif
 174         return nbest[fast_random(nbests)];
 175 }
 176
 177 void
 178 ucb1amaf_update(struct uct_policy *p, struct tree *tree, struct tree_node *node, enum stone node_color, enum stone player_color, struct playout_amafmap *map, int result)
 179 {
 180         struct ucb1_policy_amaf *b = p->data;
 181         enum stone child_color = stone_other(node_color);
 182
 183 #if 0
 184         struct board bb; bb.size = 9+2;
 185         for (struct tree_node *ni = node; ni; ni = ni->parent)
 186                 fprintf(stderr, "%s ", coord2sstr(ni->coord, &bb));
 187         fprintf(stderr, "[color %d] update result %d (color %d)\n",
 188                         node_color, result, player_color);
 189 #endif
 190
 191         while (node) {
 192                 if (node->parent == NULL)
 193                         assert(tree->root_color == stone_other(child_color));
 194
 195                 stats_add_result(&node->u, result, 1);
 196                 if (amaf_nakade(map->map[node->coord]))
 197                         amaf_op(map->map[node->coord], -);
 198
 199                 /* This loop ignores symmetry considerations, but they should
 200                  * matter only at a point when AMAF doesn't help much. */
 201                 for (struct tree_node *ni = node->children; ni; ni = ni->sibling) {
 202                         assert(map->map[ni->coord] != S_OFFBOARD);
 203                         if (map->map[ni->coord] == S_NONE)
 204                                 continue;
 205                         assert(map->game_baselen >= 0);
 206                         enum stone amaf_color = map->map[ni->coord];
 207                         if (amaf_nakade(map->map[ni->coord])) {
 208                                 if (!b->check_nakade)
 209                                         continue;
 210                                 /* We don't care to implement both_colors
 211                                  * properly since it sucks anyway. */
 212                                 int i;
 213                                 for (i = map->game_baselen; i < map->gamelen; i++)
 214                                         if (map->game[i].coord == ni->coord
 215                                             && map->game[i].color == child_color)
 216                                                 break;
 217                                 if (i == map->gamelen)
 218                                         continue;
 219                                 amaf_color = child_color;
 220                         }
 221
 222                         int nres = result;
 223                         if (amaf_color != child_color) {
 224                                 if (!b->both_colors)
 225                                         continue;
 226                                 nres = !nres;
 227                         }
 228                         /* For child_color != player_color, we still want
 229                          * to record the result unmodified; in that case,
 230                          * we will correctly negate them at the descend phase. */
 231
 232                         stats_add_result(&ni->amaf, nres, 1);
 233
 234 #if 0
 235                         fprintf(stderr, "* %s<%lld> -> %s<%lld> [%d %d => %d/%d]\n", coord2sstr(node->coord, &bb), node->hash, coord2sstr(ni->coord, &bb), ni->hash, player_color, child_color, result);
 236 #endif
 237                 }
 238
 239                 if (!is_pass(node->coord)) {
 240                         map->game_baselen--;
 241                 }
 242                 node = node->parent; child_color = stone_other(child_color);
 243         }
 244 }
 245
 246
 247 struct uct_policy *
 248 policy_ucb1amaf_init(struct uct *u, char *arg)
 249 {
 250         struct uct_policy *p = calloc(1, sizeof(*p));
 251         struct ucb1_policy_amaf *b = calloc(1, sizeof(*b));
 252         p->uct = u;
 253         p->data = b;
 254         p->descend = ucb1rave_descend;
 255         p->choose = ucb1_choose;
 256         p->update = ucb1amaf_update;
 257         p->wants_amaf = true;
 258
 259         // RAVE: 0.2vs0: 40% (+-7.3) 0.1vs0: 54.7% (+-3.5)
 260         b->explore_p = 0.1;
 261         b->equiv_rave = 3000;
 262         b->fpu = INFINITY;
 263         b->check_nakade = true;
 264         b->sylvain_rave = true;
 265
 266         if (arg) {
 267                 char *optspec, *next = arg;
 268                 while (*next) {
 269                         optspec = next;
 270                         next += strcspn(next, ":");
 271                         if (*next) { *next++ = 0; } else { *next = 0; }
 272
 273                         char *optname = optspec;
 274                         char *optval = strchr(optspec, '=');
 275                         if (optval) *optval++ = 0;
 276
 277                         if (!strcasecmp(optname, "explore_p")) {
 278                                 b->explore_p = atof(optval);
 279                         } else if (!strcasecmp(optname, "fpu") && optval) {
 280                                 b->fpu = atof(optval);
 281                         } else if (!strcasecmp(optname, "urg_randoma") && optval) {
 282                                 b->urg_randoma = atoi(optval);
 283                         } else if (!strcasecmp(optname, "urg_randomm") && optval) {
 284                                 b->urg_randomm = atoi(optval);
 285                         } else if (!strcasecmp(optname, "equiv_rave") && optval) {
 286                                 b->equiv_rave = atof(optval);
 287                         } else if (!strcasecmp(optname, "both_colors")) {
 288                                 b->both_colors = true;
 289                         } else if (!strcasecmp(optname, "sylvain_rave")) {
 290                                 b->sylvain_rave = !optval || *optval == '1';
 291                         } else if (!strcasecmp(optname, "check_nakade")) {
 292                                 b->check_nakade = !optval || *optval == '1';
 293                         } else {
 294                                 fprintf(stderr, "ucb1: Invalid policy argument %s or missing value\n", optname);
 295                                 exit(1);
 296                         }
 297                 }
 298         }
 299
 300         return p;
 301 }