Added some material from Chapter 3.

td2014 · td2014 · commit 7b8cd827eade · 2016-12-29T09:32:13.000-08:00
diff --git a/adxyz_agents_chap2.ipynb b/adxyz_agents_chap2.ipynb
@@ -1001,6 +1001,127 @@
     "class adxyz_SimpleReflexStateVacuumAgent(Agent):\n",
     "    pass"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Chapter 3"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We are looking to design agents that can solve goal seeking problems.\n",
+    "Step 1:  Define the goal, which is a state of the environment.  For example, the desired goal might be \"Car in Bucharest\" or \"Robot in square (10,10) with all squares clean\"  \n",
+    "Step 2:  Define the problem.  \n",
+    "- Define the states of the environment (atomic)\n",
+    "- Define the initial state\n",
+    "- Define legal actions\n",
+    "- Define transitions\n",
+    "- Define goal test\n",
+    "- Define path/step costs"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "graph-search:  A key algorithm for expanding the search space, that avoids redundent paths.  The search methods in this chapter are based on graph-search algorithm.\n",
+    "Each step of the algorithm does this:\n",
+    "Unexplored state -> frontier states -> explored states.\n",
+    "A state can only be in one of the three above categories."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Infrastructure for search algorithms:\n",
+    "Graphs - nodes that include references to \n",
+    "parent nodes\n",
+    "state descriptions\n",
+    "action that got from parent to child node\n",
+    "path cost (from initial state).\n",
+    "\n",
+    "Types of cost:\n",
+    "Search cost (time to determine solution)\n",
+    "Path cost (cost of actual solution - for example distance on a roadmap)\n",
+    "Total cost:  Sum of search + path cost (with appropriate scaling to put them in common units)."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Types of Search Strategies"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Algorithm evaluation criteria:\n",
+    "- Completeness (Does the algorithm find a solution - or all solutions)\n",
+    "- Optimality (Does the algorithm find the best solution)\n",
+    "- Time complexity (how long does the algorithm take to find solution)\n",
+    "- Space complexity (how much memory is used)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Uninformed search"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This includes all search algorithms that have no idea whether one choice is \"more promising\" than another non-goal state.  These algorithms generate non-goal states and test for goal states."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "- Breadth-first search:  Each node is expanded into the successor nodes one level at a time.  Uses a FIFO queue for the frontier."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Pseudo-code for BFS search:\n",
+    "\n",
+    "    unexploredNodes = {AllStates}-initialState\n",
+    "    exploredNodes = {}\n",
+    "    frontierNodes = initialState\n",
+    "    \n",
+    "    while currentNode != goalState:\n",
+    "        currentNode = frontierNodes.pop\n",
+    "        if currentNode == goalState:\n",
+    "            break\n",
+    "        else:\n",
+    "            exploredNodes.push(currentNode)\n",
+    "            for childNode in currentNode.children:\n",
+    "                frontierNodes.push(childNode)\n",
+    "                unexploredNodes.delete(childNode)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Informed search (heuristics)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": []
   }
  ],
  "metadata": {
