Adding horizon notebook, to be adjusted to fit new example format

nasa · Aug 31, 2023 · db5aae4 · db5aae4
1 parent cdf22b7
commit db5aae4
Showing 1 changed file with 48 additions and 24 deletions.
diff --git a/examples/horizon.ipynb b/examples/horizon.ipynb
@@ -4,7 +4,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Welcome to ProgPy's Horizon Example Notebook! "
+    "# Welcome to ProgPy's Horizon Example Notebook"
    ]
   },
   {
@@ -26,7 +26,7 @@
     "\n",
     "The results of this prediction will be:\n",
     "- Predicted future values (inputs, states, outputs, event_states) with uncertainty from prediction\n",
-    "- Time event is predicted to occur (with uncertainty)"
+    "- Time the event is predicted to occur (with uncertainty)"
    ]
   },
   {
@@ -70,7 +70,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "To do prediction, we need a distribution of samples. Here, we'll create a distribution around the initial state using a Multivariate Normal Distribution. We'll use 1000 samples for prediction. "
+    "To do prediction, we need a distribution of samples. Here, we'll create a distribution around the initial state using a Multivariate Normal Distribution. We'll use 500 samples for prediction. "
    ]
   },
   {
@@ -79,7 +79,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "NUM_SAMPLES = 1000\n",
+    "NUM_SAMPLES = 500\n",
     "x = MultivariateNormalDist(initial_state.keys(), initial_state.values(), np.diag([x_i*0.01 for x_i in initial_state.values()]))"
    ]
   },
@@ -96,20 +96,18 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "mc = predictors.MonteCarlo(m)"
+    "mc = MonteCarlo(m)"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Now, let's perform a prediction. We given the `predict` method the following arguments:\n",
-    "- Samples ()\n",
-    "- Number of samples \n",
+    "Now, let's perform a prediction. We give the `predict` method the following arguments:\n",
+    "- Distribution of initial samples\n",
+    "- Number of samples for prediction \n",
     "- Step size for the prediction \n",
-    "- \n",
-    "\n",
-    "Finally, the predict function requires a future loading function to be passed. This function is used to generate future inputs to the model. Since our model does not have any inputs, we'll pass a function that returns an empty set."
+    "- Prediction horizon, i.e. time value to predict to\n"
    ]
   },
   {
@@ -118,22 +116,18 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "PREDICTION_HORIZON = 5\n",
-    "samples = filt.x\n",
-    "STEP_SIZE = 0.001\n",
-    "\n",
-    "def future_loading(t, x = None):\n",
-    "    return {}\n",
+    "PREDICTION_HORIZON = 7.7\n",
+    "STEP_SIZE = 0.01\n",
     "\n",
-    "# Making Prediction\n",
-    "mc_results = mc.predict(samples, future_loading, dt=STEP_SIZE, horizon = PREDICTION_HORIZON)\n"
+    "# Make Prediction\n",
+    "mc_results = mc.predict(x, n_samples=NUM_SAMPLES, dt=STEP_SIZE, horizon = PREDICTION_HORIZON)\n"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "#### Let's see the results of the predicted time of event!"
+    "Let's see the results of the predicted time of event. We'll plot histograms of the distribution of times where `falling` and `impact` occurred. Note that no events occur after 7.7 seconds, since we enforced a prediction horizon at this value. "
    ]
   },
   {
@@ -144,7 +138,6 @@
    "source": [
     "metrics = mc_results.time_of_event.metrics()\n",
     "print(\"\\nPredicted Time of Event:\")\n",
-    "print(\"\\nSamples where impact occurs before horizon: {:.2f}%\".format(metrics['impact']['number of samples']/100 * 100))\n",
     "pprint(metrics)  # Note this takes some time\n",
     "mc_results.time_of_event.plot_hist(keys = 'impact')\n",
     "mc_results.time_of_event.plot_hist(keys = 'falling')"
@@ -154,20 +147,51 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "As you can see from the results..."
+    "Now let's calculate the percentage of each event that occurred before the prediction horizon was met. "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"\\nSamples where falling occurs before horizon: {:.2f}%\".format(metrics['falling']['number of samples']/NUM_SAMPLES * 100))\n",
+    "print(\"\\nSamples where impact occurs before horizon: {:.2f}%\".format(metrics['impact']['number of samples']/NUM_SAMPLES * 100))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "All samples reach `falling` before the prediction horizon, but only some of the samples reach `impact`. "
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "To conclude..."
+    "To conclude, in this example, we've shown how to implement a prediction `horizon`. Specifying a prediction horizon defines the time value with which to predict to, and can be used anytime a user is only interested in events that occur before a specific point in time.  "
    ]
   }
  ],
  "metadata": {
+  "kernelspec": {
+   "display_name": "ProgPyEnvAll",
+   "language": "python",
+   "name": "python3"
+  },
   "language_info": {
-   "name": "python"
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.6"
   },
   "orig_nbformat": 4
  },