update name

docs/demos/benefits.ipynb

@@ -4,14 +4,14 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Benefits of Contextualized ML"
+    "# Benefits of Contextualized"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Contextualized ML Enables High-Resolution Heterogeneity \n",
+    "## Contextualized Enables High-Resolution Heterogeneity \n",
     "By sharing information between all contexts, contextualized learning is able to estimate heterogeneity at fine-grained resolution.\n",
     "Cluster or cohort-based models treat every partition independently, limiting heterogeneity to\n",
     "coarse-grained resolution where there are large enough cohorts for independent estimation."
@@ -256,7 +256,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Contextualized ML Enables Analysis of Latent Processes\n",
+    "## Contextualized Enables Analysis of Latent Processes\n",
     "Cluster or cohort models are inferred by partitioning data into groups, assumed to be iid (independent identically distributed), and estimating models for each groups. This is only likely to be satisfied when contexts are discrete, low-dimensional, and every context-specific population is well observed. In real life, contexts are often continuous, high dimensional, and sparsely observed. When cluster or cohort approaches are applied in these circumstances, downstream modeling tasks are distorted by mis-specification, where many non-id samples are funneled into a single model. Consequently, there are no theoretical guarantees in many real life circumstances about how well a cluster or cohort model can represent heterogeneous populations. Alternatively, contextualized learning provides a way to estimate latent, non-id models for all samples with minimal assumptions about the grouping or clustering of these samples, or the functional relationship between latent models and contexts. Samples can then be grouped on the basis of model parameters and distributional differences to produce clusters in the latent model space underlying each sample. Contextualized ML intuitively recovers latent structures underlying data generation in a way a priori clustering cannot. Allowing downstream models to determine the grouping of samples rather than upstream contexts replaces traditional cluster analysis with contextualized analysis clusters."
    ]
   },
@@ -523,7 +523,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Contextualized ML Interpolates Between Observed Contexts \n",
+    "## Contextualized Interpolates Between Observed Contexts \n",
     "By learning to translate contextual information into model parameters, contextualized models learn about the meta-distribution of\n",
     "contexts. As a result, contextualized models can adapt to contexts which were never observed\n",
     "in the training data by interpolating between observed contexts or extrapolating to new contexts."