Fix typos and update documentation for ReLax

methods
BirkhoffG · Nov 9, 2023 · fa8a532 · fa8a532
1 parent c152360
commit fa8a532
Showing 1 changed file with 12 additions and 10 deletions.
diff --git a/nbs/tutorials/methods.ipynb b/nbs/tutorials/methods.ipynb
@@ -27,16 +27,16 @@
    "source": [
     "## Types of Recourse Methods\n",
     "\n",
-    "1. Non-parametric methods: These methods do not rely on any learned parameters. They generate counterfactuals solely based on the model's predictions and gradients. Examples in ReLax include `VanillaCF`, `DiverseCF` and `GrowingSpheres` . These methods inherit from NonParametricCFModule.\n",
+    "1. Non-parametric methods: These methods do not rely on any learned parameters. They generate counterfactuals solely based on the model's predictions and gradients. Examples in ReLax include `VanillaCF`, `DiverseCF` and `GrowingSphere` . These methods inherit from `CFModule`.\n",
     "\n",
-    "2. Semi-parametric methods: These methods learn some parameters to aid in counterfactual generation, but do not learn a full counterfactual generation model. Examples in ReLax include `ProtoCF`, `CCHVAE` and `CLUE`. These methods inherit from SemiParametricCFModule.\n",
+    "2. Semi-parametric methods: These methods learn some parameters to aid in counterfactual generation, but do not learn a full counterfactual generation model. Examples in ReLax include `ProtoCF`, `CCHVAE` and `CLUE`. These methods inherit from `ParametricCFModule `.\n",
     "\n",
-    "3. Parametric methods: These methods learn a full parametric model for counterfactual generation. The model is trained to generate counterfactuals that fool the model. Examples in ReLax include `CounterNet` and `VAECF`. These methods inherit from ParametricCFModule.\n",
+    "3. Parametric methods: These methods learn a full parametric model for counterfactual generation. The model is trained to generate counterfactuals that fool the model. Examples in ReLax include `CounterNet` and `VAECF`. These methods inherit from `ParametricCFModule`.\n",
     "\n",
     "\n",
     "|Method Type | Learned Parameters | Training Required | Example Methods | \n",
     "|-----|:-----|:---:|:-----:|\n",
-    "|Non-parametric | None  |No   |`VanillaCF`, `DiverseCF`, `GrowingSpheres` |\n",
+    "|Non-parametric | None  |No   |`VanillaCF`, `DiverseCF`, `GrowingSphere` |\n",
     "|Semi-parametric| Some (θ) |Modest amount   |`ProtoCF`, `CCHVAE`, `CLUE` |\n",
     "|Parametric|Full generator model (φ)|Substantial amount|`CounterNet`, `VAECF` |"
    ]
@@ -45,7 +45,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Basis \n",
+    "## Basic Usages\n",
     "\n",
     "At a high level, you can use the implemented methods in `ReLax` to generate *one* recourse explanation via three lines of code:\n",
     "\n",
@@ -73,8 +73,9 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Example of using `ReLax` for parametric methods (using `CCHVAE`)\n",
-    "\n"
+    "To use parametric and semi-parametric methods, you can first train the model\n",
+    "by calling `ParametricCF.train`, and then generate recourse explanations.\n",
+    "Here is an example of using `ReLax` for `CCHVAE`.\n"
    ]
   },
   {
@@ -85,7 +86,8 @@
     "from relax.methods import CCHVAE\n",
     "\n",
     "cchvae = CCHVAE()\n",
-    "cchvae.train(train_data) # Train counterfactual generator \n",
+    "cchvae.train(train_data) # Train CVAE before generation\n",
+    "cf = cchvae.generate_cf(x, pred_fn=pred_fn) \n",
     "```\n",
     "\n",
     "Or generate a batch of recourse explanation via the `jax.vmap` primitive:\n",
@@ -123,9 +125,9 @@
     "vcf = VanillaCF(config)\n",
     "\n",
     "```\n",
-    "Each Config class inherits from a BaseConfig that defines common options like n_steps. Method-specific parameters are defined on the individual Config classes.\n",
+    "Each Config class inherits from a `BaseConfig` that defines common options like n_steps. Method-specific parameters are defined on the individual Config classes.\n",
     "\n",
-    "See the documentation for each recourse method for details on its supported configuration parameters. The Config class for a method can be imported from relax.methods.[method_name]."
+    "See the documentation for each recourse method for details on its supported configuration parameters. The Config class for a method can be imported from `relax.methods.[method_name]`."
    ]
   },
   {