Refactor MO and CC and define matching cMABs

### Changes: * Introduced `ModelMO` and `ModelCC` classes for multi-objective and cost control modeling. * Updated `BaseMab` to support `ModelMO` and `ModelCC`. * Added `BaseCmabBernoulliMO`, `CmabBernoulliMO`, and `CmabBernoulliMOCC` for multi-objective CMAB. * Edited `BaseSmabBernoulliMO`, `SmabBernoulliMO`, and `SmabBernoulliMOCC` for multi-objective SMAB. * Enhanced validation and utility functions to handle multi-objective models. * Updated documentation to reflect new multi-objective capabilities.
PlaytikaOSS · Sep 30, 2024 · 069bd42 · 069bd42
1 parent cddda5c
commit 069bd42
Show file tree

Hide file tree

Showing 12 changed files with 1,176 additions and 312 deletions.
diff --git a/docs/tutorials/mab.ipynb b/docs/tutorials/mab.ipynb
@@ -102,14 +102,14 @@
        "</pre>\n"
       ],
       "text/plain": [
-       "\u001b[1;35mSmabBernoulli\u001b[0m\u001b[1m(\u001b[0m\n",
-       "    \u001b[33mactions\u001b[0m=\u001b[1m{\u001b[0m\n",
-       "        \u001b[32m'a1'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a2'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a3'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m\n",
-       "    \u001b[1m}\u001b[0m,\n",
-       "    \u001b[33mstrategy\u001b[0m=\u001b[1;35mClassicBandit\u001b[0m\u001b[1m(\u001b[0m\u001b[1m)\u001b[0m\n",
-       "\u001b[1m)\u001b[0m\n"
+       "\u001B[1;35mSmabBernoulli\u001B[0m\u001B[1m(\u001B[0m\n",
+       "    \u001B[33mactions\u001B[0m=\u001B[1m{\u001B[0m\n",
+       "        \u001B[32m'a1'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a2'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a3'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m\n",
+       "    \u001B[1m}\u001B[0m,\n",
+       "    \u001B[33mstrategy\u001B[0m=\u001B[1;35mClassicBandit\u001B[0m\u001B[1m(\u001B[0m\u001B[1m)\u001B[0m\n",
+       "\u001B[1m)\u001B[0m\n"
       ]
      },
      "metadata": {},
@@ -169,14 +169,14 @@
        "</pre>\n"
       ],
       "text/plain": [
-       "\u001b[1;35mSmabBernoulli\u001b[0m\u001b[1m(\u001b[0m\n",
-       "    \u001b[33mactions\u001b[0m=\u001b[1m{\u001b[0m\n",
-       "        \u001b[32m'a1'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a3'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a2'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m\n",
-       "    \u001b[1m}\u001b[0m,\n",
-       "    \u001b[33mstrategy\u001b[0m=\u001b[1;35mClassicBandit\u001b[0m\u001b[1m(\u001b[0m\u001b[1m)\u001b[0m\n",
-       "\u001b[1m)\u001b[0m\n"
+       "\u001B[1;35mSmabBernoulli\u001B[0m\u001B[1m(\u001B[0m\n",
+       "    \u001B[33mactions\u001B[0m=\u001B[1m{\u001B[0m\n",
+       "        \u001B[32m'a1'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a3'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a2'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m\n",
+       "    \u001B[1m}\u001B[0m,\n",
+       "    \u001B[33mstrategy\u001B[0m=\u001B[1;35mClassicBandit\u001B[0m\u001B[1m(\u001B[0m\u001B[1m)\u001B[0m\n",
+       "\u001B[1m)\u001B[0m\n"
       ]
      },
      "metadata": {},
@@ -422,14 +422,14 @@
        "</pre>\n"
       ],
       "text/plain": [
-       "\u001b[1;35mSmabBernoulli\u001b[0m\u001b[1m(\u001b[0m\n",
-       "    \u001b[33mactions\u001b[0m=\u001b[1m{\u001b[0m\n",
-       "        \u001b[32m'a1'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m1\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a3'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m2\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m1\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a2'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m3\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m3\u001b[0m\u001b[1m)\u001b[0m\n",
-       "    \u001b[1m}\u001b[0m,\n",
-       "    \u001b[33mstrategy\u001b[0m=\u001b[1;35mClassicBandit\u001b[0m\u001b[1m(\u001b[0m\u001b[1m)\u001b[0m\n",
-       "\u001b[1m)\u001b[0m\n"
+       "\u001B[1;35mSmabBernoulli\u001B[0m\u001B[1m(\u001B[0m\n",
+       "    \u001B[33mactions\u001B[0m=\u001B[1m{\u001B[0m\n",
+       "        \u001B[32m'a1'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m1\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a3'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m2\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m1\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a2'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m3\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m3\u001B[0m\u001B[1m)\u001B[0m\n",
+       "    \u001B[1m}\u001B[0m,\n",
+       "    \u001B[33mstrategy\u001B[0m=\u001B[1;35mClassicBandit\u001B[0m\u001B[1m(\u001B[0m\u001B[1m)\u001B[0m\n",
+       "\u001B[1m)\u001B[0m\n"
       ]
      },
      "metadata": {},
@@ -494,14 +494,14 @@
        "</pre>\n"
       ],
       "text/plain": [
-       "\u001b[1;35mSmabBernoulli\u001b[0m\u001b[1m(\u001b[0m\n",
-       "    \u001b[33mactions\u001b[0m=\u001b[1m{\u001b[0m\n",
-       "        \u001b[32m'a1'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m337\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m369\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a3'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m4448\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m4315\u001b[0m\u001b[1m)\u001b[0m,\n",
-       "        \u001b[32m'a2'\u001b[0m: \u001b[1;35mBeta\u001b[0m\u001b[1m(\u001b[0m\u001b[33mn_successes\u001b[0m=\u001b[1;36m246\u001b[0m, \u001b[33mn_failures\u001b[0m=\u001b[1;36m296\u001b[0m\u001b[1m)\u001b[0m\n",
-       "    \u001b[1m}\u001b[0m,\n",
-       "    \u001b[33mstrategy\u001b[0m=\u001b[1;35mClassicBandit\u001b[0m\u001b[1m(\u001b[0m\u001b[1m)\u001b[0m\n",
-       "\u001b[1m)\u001b[0m\n"
+       "\u001B[1;35mSmabBernoulli\u001B[0m\u001B[1m(\u001B[0m\n",
+       "    \u001B[33mactions\u001B[0m=\u001B[1m{\u001B[0m\n",
+       "        \u001B[32m'a1'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m337\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m369\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a3'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m4448\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m4315\u001B[0m\u001B[1m)\u001B[0m,\n",
+       "        \u001B[32m'a2'\u001B[0m: \u001B[1;35mBeta\u001B[0m\u001B[1m(\u001B[0m\u001B[33mn_successes\u001B[0m=\u001B[1;36m246\u001B[0m, \u001B[33mn_failures\u001B[0m=\u001B[1;36m296\u001B[0m\u001B[1m)\u001B[0m\n",
+       "    \u001B[1m}\u001B[0m,\n",
+       "    \u001B[33mstrategy\u001B[0m=\u001B[1;35mClassicBandit\u001B[0m\u001B[1m(\u001B[0m\u001B[1m)\u001B[0m\n",
+       "\u001B[1m)\u001B[0m\n"
       ]
      },
      "metadata": {},