CircleCI update of dev docs (817).

dev/_downloads/01119b09b904483a657e921d21bc15d1/demo_pca_via_affinity_matcher.ipynb

@@ -15,7 +15,7 @@
       },
       "outputs": [],
       "source": [
-        "# Author: Hugues Van Assel <[email protected]>\n#\n# License: BSD 3-Clause License\n\nimport matplotlib.pyplot as plt\nfrom sklearn.datasets import load_digits\n\nfrom torchdr.spectral import PCA\nfrom torchdr import AffinityMatcher, ScalarProductAffinity"
+        "# Author: Hugues Van Assel <[email protected]>\n#\n# License: BSD 3-Clause License\n\nimport matplotlib.pyplot as plt\nfrom sklearn.datasets import load_digits\n\nfrom torchdr import AffinityMatcher, ScalarProductAffinity\nfrom torchdr.spectral import PCA"
       ]
     },
     {

dev/_downloads/17d429f11ca029172465ca0d7bb1879e/demo_ea_adaptivity.ipynb

@@ -15,7 +15,7 @@
       },
       "outputs": [],
       "source": [
-        "# Author: Hugues Van Assel <[email protected]>\n#\n# License: BSD 3-Clause License\n\nimport torch\nimport matplotlib.pyplot as plt\nfrom matplotlib import cm\n\nfrom torchdr import (\n    NormalizedGaussianAffinity,\n    EntropicAffinity,\n)"
+        "# Author: Hugues Van Assel <[email protected]>\n#\n# License: BSD 3-Clause License\n\nimport matplotlib.pyplot as plt\nimport torch\nfrom matplotlib import cm\n\nfrom torchdr import EntropicAffinity, NormalizedGaussianAffinity"
       ]
     },
     {
@@ -65,7 +65,7 @@
       "cell_type": "markdown",
       "metadata": {},
       "source": [
-        "## Entropic affinity (adaptive bandwidth)\n\nTo remedy this issue, we can use an **entropic affinity**. The entropic affinity\nemploys an **adaptive bandwidth** that depends on the local density of points.\nBy controling the entropy of each row of the affinity matrix, it ensures that\n**each point has the same number of effective neighbors** (given by\nthe ``perplexity`` parameter) regardless of the local density around it.\n\nIn ``TorchDR``, this object is available here :\n:class:`torchdr.EntropicAffinity`.\n\n"
+        "## Entropic affinity (adaptive bandwidth)\n\nTo remedy this issue, we can use an **entropic affinity**. The entropic affinity\nemploys an **adaptive bandwidth** that depends on the local density of points.\nBy controlling the entropy of each row of the affinity matrix, it ensures that\n**each point has the same number of effective neighbors** (given by\nthe ``perplexity`` parameter) regardless of the local density around it.\n\nIn ``TorchDR``, this object is available here :\n:class:`torchdr.EntropicAffinity`.\n\n"
       ]
     },
     {

dev/_downloads/197d4a929674e9f514c5e8f05ea2d125/demo_tsne_swiss_roll.py

@@ -39,7 +39,7 @@
 ax.scatter(X[:, 0], X[:, 1], X[:, 2], c=t, s=50, alpha=0.8)
 ax.set_title("Swiss Roll in Ambient Space")
 ax.view_init(azim=-66, elev=12)
-_ = ax.text2D(0.8, 0.05, s="n_samples={}".format(n_samples), transform=ax.transAxes)
+_ = ax.text2D(0.8, 0.05, s=f"n_samples={n_samples}", transform=ax.transAxes)
 
 # %%
 # Compute the TSNE embedding
@@ -77,7 +77,7 @@
 for i, perplexity in enumerate(perplexity_values):
     plt.subplot(1, 4, i + 1)
     plt.scatter(X_embedded[i][:, 0], X_embedded[i][:, 1], c=t, s=50, alpha=0.8)
-    plt.title("Perplexity = {}".format(perplexity))
+    plt.title(f"Perplexity = {perplexity}")
 
 # %%
 # We can observe that the perplexity parameter significantly influences the embedding.

dev/_downloads/286dcc8a82b9a5553a5d809cc0f6fa61/demo_ne_methods_affinity_matcher.ipynb

@@ -15,7 +15,7 @@
       },
       "outputs": [],
       "source": [
-        "# Author: Titouan Vayer <[email protected]>\n#         Hugues Van Assel <[email protected]>\n#\n# License: BSD 3-Clause License"
+        "# Author: Titouan Vayer <[email protected]>\n#         Hugues Van Assel <[email protected]>\n#\n# License: BSD 3-Clause License\n\nimport urllib.request"
       ]
     },
     {
@@ -26,7 +26,7 @@
       },
       "outputs": [],
       "source": [
-        "import matplotlib.pyplot as plt\n\nfrom torchdr import (\n    AffinityMatcher,\n    SNE,\n    UMAP,\n    TSNE,\n    EntropicAffinity,\n    NormalizedGaussianAffinity,\n)\nimport numpy as np\nimport urllib.request"
+        "import matplotlib.pyplot as plt\nimport numpy as np\n\nfrom torchdr import (\n    SNE,\n    TSNE,\n    UMAP,\n    AffinityMatcher,\n    EntropicAffinity,\n    NormalizedGaussianAffinity,\n)"
       ]
     },
     {
@@ -62,7 +62,7 @@
       },
       "outputs": [],
       "source": [
-        "params = {\n    \"optimizer\": \"Adam\",\n    \"optimizer_kwargs\": None,\n    \"max_iter\": 100,\n    \"lr\": 1e0,\n}\n\nsne = SNE(early_exaggeration=1, **params)\n\numap = UMAP(early_exaggeration=1, **params)\n\ntsne = TSNE(early_exaggeration=1, **params)\n\nall_methods = {\n    \"TSNE\": tsne,\n    \"SNE\": sne,\n    \"UMAP\": umap,\n}\n\nfor method_name, method in all_methods.items():\n    print(\"--- Computing {} ---\".format(method_name))\n    method.fit(X)"
+        "params = {\n    \"optimizer\": \"Adam\",\n    \"optimizer_kwargs\": None,\n    \"max_iter\": 100,\n    \"lr\": 1e0,\n}\n\nsne = SNE(early_exaggeration=1, **params)\n\numap = UMAP(early_exaggeration=1, **params)\n\ntsne = TSNE(early_exaggeration=1, **params)\n\nall_methods = {\n    \"TSNE\": tsne,\n    \"SNE\": sne,\n    \"UMAP\": umap,\n}\n\nfor method_name, method in all_methods.items():\n    print(f\"--- Computing {method_name} ---\")\n    method.fit(X)"
       ]
     },
     {
@@ -80,7 +80,7 @@
       },
       "outputs": [],
       "source": [
-        "fig = plt.figure(figsize=(12, 4))\n\nfor i, (method_name, method) in enumerate(all_methods.items()):\n    ax = fig.add_subplot(1, 3, i + 1)\n    emb = method.embedding_.detach().numpy()  # get the embedding\n    ax.scatter(emb[:, 0], emb[:, 1], c=Y, s=10)\n    ax.set_title(\"{0}\".format(method_name), fontsize=24)\n    ax.set_xticks([])\n    ax.set_yticks([])\nplt.tight_layout()"
+        "fig = plt.figure(figsize=(12, 4))\n\nfor i, (method_name, method) in enumerate(all_methods.items()):\n    ax = fig.add_subplot(1, 3, i + 1)\n    emb = method.embedding_.detach().numpy()  # get the embedding\n    ax.scatter(emb[:, 0], emb[:, 1], c=Y, s=10)\n    ax.set_title(f\"{method_name}\", fontsize=24)\n    ax.set_xticks([])\n    ax.set_yticks([])\nplt.tight_layout()"
       ]
     },
     {
@@ -98,7 +98,7 @@
       },
       "outputs": [],
       "source": [
-        "sne_affinity_matcher = AffinityMatcher(\n    n_components=2,\n    # SNE matches an EntropicAffinity\n    affinity_in=EntropicAffinity(sparsity=False),\n    # with a Gaussian kernel normalized by row\n    affinity_out=NormalizedGaussianAffinity(normalization_dim=1),\n    loss_fn=\"cross_entropy_loss\",  # and the cross_entropy loss\n    **params,\n)\nsne_affinity_matcher.fit(X)\n\nfig = plt.figure(figsize=(8, 4))\ntwo_sne_dict = {\"SNE\": sne, \"SNE (with affinity matcher)\": sne_affinity_matcher}\nfor i, (method_name, method) in enumerate(two_sne_dict.items()):\n    ax = fig.add_subplot(1, 2, i + 1)\n    emb = method.embedding_.detach().numpy()  # get the embedding\n    ax.scatter(emb[:, 0], emb[:, 1], c=Y, s=10)\n    ax.set_title(\"{0}\".format(method_name), fontsize=15)\n    ax.set_xticks([])\n    ax.set_yticks([])\nplt.tight_layout()"
+        "sne_affinity_matcher = AffinityMatcher(\n    n_components=2,\n    # SNE matches an EntropicAffinity\n    affinity_in=EntropicAffinity(sparsity=False),\n    # with a Gaussian kernel normalized by row\n    affinity_out=NormalizedGaussianAffinity(normalization_dim=1),\n    loss_fn=\"cross_entropy_loss\",  # and the cross_entropy loss\n    **params,\n)\nsne_affinity_matcher.fit(X)\n\nfig = plt.figure(figsize=(8, 4))\ntwo_sne_dict = {\"SNE\": sne, \"SNE (with affinity matcher)\": sne_affinity_matcher}\nfor i, (method_name, method) in enumerate(two_sne_dict.items()):\n    ax = fig.add_subplot(1, 2, i + 1)\n    emb = method.embedding_.detach().numpy()  # get the embedding\n    ax.scatter(emb[:, 0], emb[:, 1], c=Y, s=10)\n    ax.set_title(f\"{method_name}\", fontsize=15)\n    ax.set_xticks([])\n    ax.set_yticks([])\nplt.tight_layout()"
       ]
     },
     {

dev/_downloads/3782b8d6c3b9610739919e42de4d08a1/demo_pca_via_affinity_matcher.py

@@ -14,8 +14,8 @@
 import matplotlib.pyplot as plt
 from sklearn.datasets import load_digits
 
-from torchdr.spectral import PCA
 from torchdr import AffinityMatcher, ScalarProductAffinity
+from torchdr.spectral import PCA
 
 # %%
 # Load toy images

dev/_downloads/3e483e2c5442faf121372695ae6355bc/demo_ne_methods_affinity_matcher.py

@@ -12,20 +12,20 @@
 #
 # License: BSD 3-Clause License
 
+import urllib.request
+
 # %%
 import matplotlib.pyplot as plt
+import numpy as np
 
 from torchdr import (
-    AffinityMatcher,
     SNE,
-    UMAP,
     TSNE,
+    UMAP,
+    AffinityMatcher,
     EntropicAffinity,
     NormalizedGaussianAffinity,
 )
-import numpy as np
-import urllib.request
-
 
 # %%
 # Load the SNARE-seq dataset (gene expression) with cell type labels
@@ -71,7 +71,7 @@ def load_numpy_from_url(url, delimiter="\t"):
 }
 
 for method_name, method in all_methods.items():
-    print("--- Computing {} ---".format(method_name))
+    print(f"--- Computing {method_name} ---")
     method.fit(X)
 
 # %%
@@ -84,7 +84,7 @@ def load_numpy_from_url(url, delimiter="\t"):
     ax = fig.add_subplot(1, 3, i + 1)
     emb = method.embedding_.detach().numpy()  # get the embedding
     ax.scatter(emb[:, 0], emb[:, 1], c=Y, s=10)
-    ax.set_title("{0}".format(method_name), fontsize=24)
+    ax.set_title(f"{method_name}", fontsize=24)
     ax.set_xticks([])
     ax.set_yticks([])
 plt.tight_layout()
@@ -119,7 +119,7 @@ def load_numpy_from_url(url, delimiter="\t"):
     ax = fig.add_subplot(1, 2, i + 1)
     emb = method.embedding_.detach().numpy()  # get the embedding
     ax.scatter(emb[:, 0], emb[:, 1], c=Y, s=10)
-    ax.set_title("{0}".format(method_name), fontsize=15)
+    ax.set_title(f"{method_name}", fontsize=15)
     ax.set_xticks([])
     ax.set_yticks([])
 plt.tight_layout()

dev/_downloads/d07a5a618e87545c420604821fa60541/demo_tsne_swiss_roll.ipynb

@@ -62,7 +62,7 @@
       },
       "outputs": [],
       "source": [
-        "fig = plt.figure(figsize=(8, 6))\nax = fig.add_subplot(111, projection=\"3d\")\nfig.add_axes(ax)\nax.scatter(X[:, 0], X[:, 1], X[:, 2], c=t, s=50, alpha=0.8)\nax.set_title(\"Swiss Roll in Ambient Space\")\nax.view_init(azim=-66, elev=12)\n_ = ax.text2D(0.8, 0.05, s=\"n_samples={}\".format(n_samples), transform=ax.transAxes)"
+        "fig = plt.figure(figsize=(8, 6))\nax = fig.add_subplot(111, projection=\"3d\")\nfig.add_axes(ax)\nax.scatter(X[:, 0], X[:, 1], X[:, 2], c=t, s=50, alpha=0.8)\nax.set_title(\"Swiss Roll in Ambient Space\")\nax.view_init(azim=-66, elev=12)\n_ = ax.text2D(0.8, 0.05, s=f\"n_samples={n_samples}\", transform=ax.transAxes)"
       ]
     },
     {
@@ -116,7 +116,7 @@
       },
       "outputs": [],
       "source": [
-        "perplexity_values = [2, 5, 10, 20]\nX_embedded = []\nfor perplexity in perplexity_values:\n    if len(X_embedded) == 0:\n        init = \"pca\"\n    else:\n        init = X_embedded[-1]\n    tsne = TSNE(n_components=2, perplexity=perplexity, max_iter=200, init=init)\n    X_embedded.append(tsne.fit_transform(X))\n\nplt.figure(figsize=(12, 4))\n\nfor i, perplexity in enumerate(perplexity_values):\n    plt.subplot(1, 4, i + 1)\n    plt.scatter(X_embedded[i][:, 0], X_embedded[i][:, 1], c=t, s=50, alpha=0.8)\n    plt.title(\"Perplexity = {}\".format(perplexity))"
+        "perplexity_values = [2, 5, 10, 20]\nX_embedded = []\nfor perplexity in perplexity_values:\n    if len(X_embedded) == 0:\n        init = \"pca\"\n    else:\n        init = X_embedded[-1]\n    tsne = TSNE(n_components=2, perplexity=perplexity, max_iter=200, init=init)\n    X_embedded.append(tsne.fit_transform(X))\n\nplt.figure(figsize=(12, 4))\n\nfor i, perplexity in enumerate(perplexity_values):\n    plt.subplot(1, 4, i + 1)\n    plt.scatter(X_embedded[i][:, 0], X_embedded[i][:, 1], c=t, s=50, alpha=0.8)\n    plt.title(f\"Perplexity = {perplexity}\")"
       ]
     },
     {

dev/_downloads/e08ce8ce7e2733d588920ebecfa7ca60/demo_ea_adaptivity.py

@@ -11,14 +11,11 @@
 #
 # License: BSD 3-Clause License
 
-import torch
 import matplotlib.pyplot as plt
+import torch
 from matplotlib import cm
 
-from torchdr import (
-    NormalizedGaussianAffinity,
-    EntropicAffinity,
-)
+from torchdr import EntropicAffinity, NormalizedGaussianAffinity
 
 # %%
 # We generate three Gaussian clusters of points with different standard deviations
@@ -95,7 +92,7 @@ def plot_affinity_graph(G):
 #
 # To remedy this issue, we can use an **entropic affinity**. The entropic affinity
 # employs an **adaptive bandwidth** that depends on the local density of points.
-# By controling the entropy of each row of the affinity matrix, it ensures that
+# By controlling the entropy of each row of the affinity matrix, it ensures that
 # **each point has the same number of effective neighbors** (given by
 # the ``perplexity`` parameter) regardless of the local density around it.
 #

dev/_modules/torchdr/affinity/base.html

@@ -82,26 +82,25 @@
            <div itemprop="articleBody">
 
   <h1>Source code for torchdr.affinity.base</h1><div class="highlight"><pre>
-<span></span><span class="c1"># -*- coding: utf-8 -*-</span>
-<span class="sd">&quot;&quot;&quot;Base classes for affinity matrices.&quot;&quot;&quot;</span>
+<span></span><span class="sd">&quot;&quot;&quot;Base classes for affinity matrices.&quot;&quot;&quot;</span>
 
 <span class="c1"># Author: Hugues Van Assel &lt;[email protected]&gt;</span>
 <span class="c1">#</span>
 <span class="c1"># License: BSD 3-Clause License</span>
 
 <span class="kn">from</span> <span class="nn">abc</span> <span class="kn">import</span> <span class="n">ABC</span>
 
+<span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
 <span class="kn">import</span> <span class="nn">torch</span>
 
-<span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
 <span class="kn">from</span> <span class="nn">torchdr.utils</span> <span class="kn">import</span> <span class="p">(</span>
+    <span class="n">LazyTensorType</span><span class="p">,</span>
+    <span class="n">handle_keops</span><span class="p">,</span>
+    <span class="n">pairwise_distances</span><span class="p">,</span>
+    <span class="n">pykeops</span><span class="p">,</span>
     <span class="n">symmetric_pairwise_distances</span><span class="p">,</span>
     <span class="n">symmetric_pairwise_distances_indices</span><span class="p">,</span>
-    <span class="n">pairwise_distances</span><span class="p">,</span>
     <span class="n">to_torch</span><span class="p">,</span>
-    <span class="n">LazyTensorType</span><span class="p">,</span>
-    <span class="n">pykeops</span><span class="p">,</span>
-    <span class="n">handle_keops</span><span class="p">,</span>
 <span class="p">)</span>
 
 

dev/_modules/torchdr/affinity/entropic.html

@@ -82,36 +82,36 @@
            <div itemprop="articleBody">
 
   <h1>Source code for torchdr.affinity.entropic</h1><div class="highlight"><pre>
-<span></span><span class="c1"># -*- coding: utf-8 -*-</span>
-<span class="sd">&quot;&quot;&quot;Affinity matrices with entropic constraints.&quot;&quot;&quot;</span>
+<span></span><span class="sd">&quot;&quot;&quot;Affinity matrices with entropic constraints.&quot;&quot;&quot;</span>
 
 <span class="c1"># Author: Hugues Van Assel &lt;[email protected]&gt;</span>
 <span class="c1">#         Titouan Vayer &lt;[email protected]&gt;</span>
 <span class="c1">#         Rémi Flamary &lt;[email protected]&gt;</span>
 <span class="c1">#</span>
 <span class="c1"># License: BSD 3-Clause License</span>
 
-<span class="kn">import</span> <span class="nn">torch</span>
-<span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
-<span class="kn">from</span> <span class="nn">tqdm</span> <span class="kn">import</span> <span class="n">tqdm</span>
-<span class="kn">import</span> <span class="nn">warnings</span>
 <span class="kn">import</span> <span class="nn">contextlib</span>
 <span class="kn">import</span> <span class="nn">math</span>
+<span class="kn">import</span> <span class="nn">warnings</span>
 <span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Tuple</span>
 
+<span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
+<span class="kn">import</span> <span class="nn">torch</span>
+<span class="kn">from</span> <span class="nn">tqdm</span> <span class="kn">import</span> <span class="n">tqdm</span>
+
+<span class="kn">from</span> <span class="nn">torchdr.affinity.base</span> <span class="kn">import</span> <span class="n">LogAffinity</span><span class="p">,</span> <span class="n">SparseLogAffinity</span>
 <span class="kn">from</span> <span class="nn">torchdr.utils</span> <span class="kn">import</span> <span class="p">(</span>
+    <span class="n">OPTIMIZERS</span><span class="p">,</span>
+    <span class="n">batch_transpose</span><span class="p">,</span>
+    <span class="n">check_NaNs</span><span class="p">,</span>
     <span class="n">entropy</span><span class="p">,</span>
     <span class="n">false_position</span><span class="p">,</span>
-    <span class="n">wrap_vectors</span><span class="p">,</span>
-    <span class="n">sum_matrix_vector</span><span class="p">,</span>
-    <span class="n">kmin</span><span class="p">,</span>
     <span class="n">kmax</span><span class="p">,</span>
-    <span class="n">check_NaNs</span><span class="p">,</span>
+    <span class="n">kmin</span><span class="p">,</span>
     <span class="n">logsumexp_red</span><span class="p">,</span>
-    <span class="n">batch_transpose</span><span class="p">,</span>
-    <span class="n">OPTIMIZERS</span><span class="p">,</span>
+    <span class="n">sum_matrix_vector</span><span class="p">,</span>
+    <span class="n">wrap_vectors</span><span class="p">,</span>
 <span class="p">)</span>
-<span class="kn">from</span> <span class="nn">torchdr.affinity.base</span> <span class="kn">import</span> <span class="n">LogAffinity</span><span class="p">,</span> <span class="n">SparseLogAffinity</span>
 
 
 <span class="nd">@wrap_vectors</span>

dev/_modules/torchdr/affinity/knn_normalized.html

@@ -82,25 +82,19 @@
            <div itemprop="articleBody">
 
   <h1>Source code for torchdr.affinity.knn_normalized</h1><div class="highlight"><pre>
-<span></span><span class="c1"># -*- coding: utf-8 -*-</span>
-<span class="sd">&quot;&quot;&quot;Affinity matrices with normalizations using nearest neighbor distances.&quot;&quot;&quot;</span>
+<span></span><span class="sd">&quot;&quot;&quot;Affinity matrices with normalizations using nearest neighbor distances.&quot;&quot;&quot;</span>
 
 <span class="c1"># Author: Hugues Van Assel &lt;[email protected]&gt;</span>
 <span class="c1">#         Cédric Vincent-Cuaz &lt;[email protected]&gt;</span>
 <span class="c1">#</span>
 <span class="c1"># License: BSD 3-Clause License</span>
 
-<span class="kn">import</span> <span class="nn">torch</span>
 <span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Tuple</span>
 
+<span class="kn">import</span> <span class="nn">torch</span>
+
 <span class="kn">from</span> <span class="nn">torchdr.affinity.base</span> <span class="kn">import</span> <span class="n">Affinity</span><span class="p">,</span> <span class="n">LogAffinity</span>
-<span class="kn">from</span> <span class="nn">torchdr.utils</span> <span class="kn">import</span> <span class="p">(</span>
-    <span class="n">kmin</span><span class="p">,</span>
-    <span class="n">wrap_vectors</span><span class="p">,</span>
-    <span class="n">batch_transpose</span><span class="p">,</span>
-    <span class="n">logsumexp_red</span><span class="p">,</span>
-    <span class="n">sum_red</span><span class="p">,</span>
-<span class="p">)</span>
+<span class="kn">from</span> <span class="nn">torchdr.utils</span> <span class="kn">import</span> <span class="n">batch_transpose</span><span class="p">,</span> <span class="n">kmin</span><span class="p">,</span> <span class="n">logsumexp_red</span><span class="p">,</span> <span class="n">sum_red</span><span class="p">,</span> <span class="n">wrap_vectors</span>
 
 
 <span class="nd">@wrap_vectors</span>