Merge branch 'main' of github.com:msluszniak/scholar into main

CHANGELOG.md

@@ -1,37 +1,44 @@
 # Changelog
 
-## v0.2.1-dev
+## v0.2.2-dev
 
-## v0.2.0 (2023-08-16)
+## v0.2.1 (2023-08-30)
+
+### Enhancements
+
+  * Remove `VegaLite.Data` in favour of future use of `Tucan`
+  * Do not use EXLA at compile time in `Metrics`
+
+## v0.2.0 (2023-08-29)
 
 This version requires Elixir v1.14+.
 
 ### Enhancements
 
   * Update notebooks
-  * Add support for :f16 and :bf16 types in SVD
-  * Add Affinity Propagation
-  * Add t-SNE
-  * Add Polynomial Regression
-  * Replace seeds with Random.key
+  * Add support for `:f16` and `:bf16` types in `SVD`
+  * Add `Affinity Propagation`
+  * Add `t-SNE`
+  * Add `Polynomial Regression`
+  * Replace seeds with `Random.key`
   * Add 'unrolling loops' option 
-  * Add support for custom optimizers in Logistic Regression
-  * Add Trapezoidal Integration
-  * Add AUC ROC, AUC, and ROC Curve
-  * Add Simpson rule Integration
+  * Add support for custom optimizers in `Logistic Regression`
+  * Add `Trapezoidal Integration`
+  * Add `AUC-ROC`, `AUC`, and `ROC Curve`
+  * Add `Simpson rule integration`
   * Unify tests
-  * Add Radius Nearest Neighbors
-  * Add DBSCAN
-  * Add classification metrics: Average Precision Score, Balanced Accuracy Score,
-  Cohen Kappa Score, Brier Score Loss, Zero-One Loss, Top-k Accuracy Score
-  * Add regression metrics: R2 Score, MSLE, MAPE, Maximum Residual Error
-  * Add support for axes in Confusion Matrix
-  * Add support for broadcasting in Metrics.Distances
+  * Add `Radius Nearest Neighbors`
+  * Add `DBSCAN`
+  * Add classification metrics: `Average Precision Score`, `Balanced Accuracy Score`,
+  `Cohen Kappa Score`, `Brier Score Loss`, `Zero-One Loss`, `Top-k Accuracy Score`
+  * Add regression metrics: `R2 Score`, `MSLE`, `MAPE`, `Maximum Residual Error`
+  * Add support for axes in `Confusion Matrix`
+  * Add support for broadcasting in `Metrics.Distances`
   * Update CI
-  * Add Gaussian Mixtures
-  * Add Model selection functionalities: K-fold, K-fold Cross Validation, Grid Search
-  * Change structure of metrics in Scholar
-  * Add a guide with Cross-Validation and Grid Search
+  * Add `Gaussian Mixtures`
+  * Add Model selection functionalities: `K-fold`, `K-fold Cross Validation`, `Grid Search`
+  * Change structure of metrics in `Scholar`
+  * Add a guide with `Cross-Validation` and `Grid Search`
 
 ## v0.1.0 (2023-03-29)
 

README.md

@@ -19,7 +19,7 @@ Add to your `mix.exs`:
 ```elixir
 def deps do
   [
-    {:scholar, "~> 0.1"}
+    {:scholar, "~> 0.2.1"}
   ]
 end
 ```
@@ -30,7 +30,7 @@ such as EXLA:
 ```elixir
 def deps do
   [
-    {:scholar, "~> 0.1"},
+    {:scholar, "~> 0.2.1"},
     {:exla, ">= 0.0.0"}
   ]
 end
@@ -51,7 +51,7 @@ To use Scholar inside code notebooks, run:
 
 ```elixir
 Mix.install([
-  {:scholar, "~> 0.1"},
+  {:scholar, "~> 0.2.1"},
   {:exla, ">= 0.0.0"}
 ])
 

lib/scholar/cluster/affinity_propagation.ex

@@ -116,6 +116,7 @@ defmodule Scholar.Cluster.AffinityPropagation do
     iterations = opts[:iterations]
     damping_factor = opts[:damping_factor]
     self_preference = opts[:self_preference]
+    data = to_float(data)
 
     {initial_a, initial_r, s, affinity_matrix} =
       initialize_matrices(data, self_preference: self_preference)
@@ -307,14 +308,12 @@ defmodule Scholar.Cluster.AffinityPropagation do
   end
 
   defnp initialize_similarities(data, opts \\ []) do
-    {n, dims} = Nx.shape(data)
+    n = Nx.axis_size(data, 0)
     self_preference = opts[:self_preference]
-    t1 = Nx.reshape(data, {1, n, dims}) |> Nx.broadcast({n, n, dims})
-    t2 = Nx.reshape(data, {n, 1, dims}) |> Nx.broadcast({n, n, dims})
 
-    dist =
-      (-1 * Scholar.Metrics.Distance.squared_euclidean(t1, t2, axes: [-1]))
-      |> Nx.as_type(to_float_type(data))
+    norm1 = Nx.sum(data ** 2, axes: [1], keep_axes: true)
+    norm2 = Nx.transpose(norm1)
+    dist = -1 * (norm1 + norm2 - 2 * Nx.dot(data, [1], data, [1]))
 
     fill_in =
       cond do

lib/scholar/integrate/integrate.ex

@@ -17,7 +17,7 @@ defmodule Scholar.Integrate do
     keep_axis: [
       type: :boolean,
       default: false,
-      doc: "If set to true, the axis which is reduced are kept."
+      doc: "If set to true, the axis which is reduced is kept."
     ]
   ]