CircleCI update of dev docs (2851).

Author: Circle CI

master/_downloads/0228827bdcc2a6735b1faf26c864145d/plot_OT_1D_smooth.zip

@@ -0,0 +1,212 @@
+# -*- coding: utf-8 -*-
+r"""
+======================================================================================================================================
+Detecting outliers by learning sample marginal distribution with CO-Optimal Transport and by using unbalanced Co-Optimal Transport
+======================================================================================================================================
+
+In this example, we consider two point clouds living in different Euclidean spaces, where the outliers
+are artifically injected into the target data. We illustrate two methods which allow to filter out
+these outliers.
+
+The first method requires learning the sample marginal distribution which minimizes
+the CO-Optimal Transport distance [49] between two input spaces.
+More precisely, given a source data
+:math:`(X, \mu_x^{(s)}, \mu_x^{(f)})` and a target matrix :math:`Y` associated with a fixed
+histogram on features :math:`\mu_y^{(f)}`, we want to solve the following problem
+
+.. math::
+    \min_{\mu_y^{(s)} \in \Delta} \text{COOT}\left( (X, \mu_x^{(s)}, \mu_x^{(f)}), (Y, \mu_y^{(s)}, \mu_y^{(f)}) \right)
+
+where :math:`\Delta` is the probability simplex. This minimization is done with a
+simple projected gradient descent in PyTorch. We use the automatic backend of POT that
+allows us to compute the CO-Optimal Transport distance with :func:`ot.coot.co_optimal_transport2`
+with differentiable losses.
+
+The second method simply requires direct application of unbalanced Co-Optimal Transport [71].
+More precisely, it is enough to use the sample and feature coupling from solving
+
+.. math::
+    \text{UCOOT}\left( (X, \mu_x^{(s)}, \mu_x^{(f)}), (Y, \mu_y^{(s)}, \mu_y^{(f)}) \right)
+
+where all the marginal distributions are uniform.
+
+.. [49] Redko, I., Vayer, T., Flamary, R., and Courty, N. (2020).
+   `CO-Optimal Transport <https://proceedings.neurips.cc/paper/2020/file/cc384c68ad503482fb24e6d1e3b512ae-Paper.pdf>`_.
+   Advances in Neural Information Processing Systems, 33.
+.. [71] H. Tran, H. Janati, N. Courty, R. Flamary, I. Redko, P. Demetci & R. Singh (2023). [Unbalanced Co-Optimal Transport](https://dl.acm.org/doi/10.1609/aaai.v37i8.26193).
+    AAAI Conference on Artificial Intelligence.
+"""
+
+# Author: Remi Flamary <[email protected]>
+#         Quang Huy Tran <[email protected]>
+# License: MIT License
+
+from matplotlib.patches import ConnectionPatch
+import torch
+import numpy as np
+
+import matplotlib.pyplot as pl
+import ot
+
+from ot.coot import co_optimal_transport as coot
+from ot.coot import co_optimal_transport2 as coot2
+from ot.gromov._unbalanced import unbalanced_co_optimal_transport
+
+
+# %%
+# Generate data
+# -------------
+# The source and clean target matrices are generated by
+# :math:`X_{i,j} = \cos(\frac{i}{n_1} \pi) + \cos(\frac{j}{d_1} \pi)` and
+# :math:`Y_{i,j} = \cos(\frac{i}{n_2} \pi) + \cos(\frac{j}{d_2} \pi)`.
+# The target matrix is then contaminated by adding 5 row outliers.
+# Intuitively, we expect that the estimated sample distribution should ignore these outliers,
+# i.e. their weights should be zero.
+
+np.random.seed(182)
+
+n1, d1 = 20, 16
+n2, d2 = 10, 8
+n = 15
+
+X = (
+    torch.cos(torch.arange(n1) * torch.pi / n1)[:, None] +
+    torch.cos(torch.arange(d1) * torch.pi / d1)[None, :]
+)
+
+# Generate clean target data mixed with outliers
+Y_noisy = torch.randn((n, d2)) * 10.0
+Y_noisy[:n2, :] = (
+    torch.cos(torch.arange(n2) * torch.pi / n2)[:, None] +
+    torch.cos(torch.arange(d2) * torch.pi / d2)[None, :]
+)
+Y = Y_noisy[:n2, :]
+
+X, Y_noisy, Y = X.double(), Y_noisy.double(), Y.double()
+
+fig, axes = pl.subplots(nrows=1, ncols=3, figsize=(12, 5))
+axes[0].imshow(X, vmin=-2, vmax=2)
+axes[0].set_title('$X$')
+
+axes[1].imshow(Y, vmin=-2, vmax=2)
+axes[1].set_title('Clean $Y$')
+
+axes[2].imshow(Y_noisy, vmin=-2, vmax=2)
+axes[2].set_title('Noisy $Y$')
+
+pl.tight_layout()
+
+# %%
+# Optimize the COOT distance with respect to the sample marginal distribution
+# ---------------------------------------------------------------------------
+
+losses = []
+lr = 1e-3
+niter = 1000
+
+b = torch.tensor(ot.unif(n), requires_grad=True)
+
+for i in range(niter):
+
+    loss = coot2(X, Y_noisy, wy_samp=b, log=False, verbose=False)
+    losses.append(float(loss))
+
+    loss.backward()
+
+    with torch.no_grad():
+        b -= lr * b.grad  # gradient step
+        b[:] = ot.utils.proj_simplex(b)  # projection on the simplex
+
+    b.grad.zero_()
+
+# Estimated sample marginal distribution and training loss curve
+pl.plot(losses[10:])
+pl.title('CO-Optimal Transport distance')
+
+print(f"Marginal distribution = {b.detach().numpy()}")
+
+# %%
+# Visualizing the row and column alignments with the estimated sample marginal distribution
+# -----------------------------------------------------------------------------------------
+#
+# Clearly, the learned marginal distribution completely and successfully ignores the 5 outliers.
+
+X, Y_noisy = X.numpy(), Y_noisy.numpy()
+b = b.detach().numpy()
+
+pi_sample, pi_feature = coot(X, Y_noisy, wy_samp=b, log=False, verbose=True)
+
+fig = pl.figure(4, (9, 7))
+pl.clf()
+
+ax1 = pl.subplot(2, 2, 3)
+pl.imshow(X, vmin=-2, vmax=2)
+pl.xlabel('$X$')
+
+ax2 = pl.subplot(2, 2, 2)
+ax2.yaxis.tick_right()
+pl.imshow(np.transpose(Y_noisy), vmin=-2, vmax=2)
+pl.title("Transpose(Noisy $Y$)")
+ax2.xaxis.tick_top()
+
+for i in range(n1):
+    j = np.argmax(pi_sample[i, :])
+    xyA = (d1 - .5, i)
+    xyB = (j, d2 - .5)
+    con = ConnectionPatch(xyA=xyA, xyB=xyB, coordsA=ax1.transData,
+                          coordsB=ax2.transData, color="black")
+    fig.add_artist(con)
+
+for i in range(d1):
+    j = np.argmax(pi_feature[i, :])
+    xyA = (i, -.5)
+    xyB = (-.5, j)
+    con = ConnectionPatch(
+        xyA=xyA, xyB=xyB, coordsA=ax1.transData, coordsB=ax2.transData, color="blue")
+    fig.add_artist(con)
+
+# %%
+# Now, let see if we can use unbalanced Co-Optimal Transport to recover the clean OT plans,
+# without the need of learning the marginal distribution as in Co-Optimal Transport.
+# -----------------------------------------------------------------------------------------
+
+pi_sample, pi_feature = unbalanced_co_optimal_transport(
+    X=X, Y=Y_noisy, reg_marginals=(10, 10), epsilon=0, divergence="kl",
+    unbalanced_solver="mm", max_iter=1000, tol=1e-6,
+    max_iter_ot=1000, tol_ot=1e-6, log=False, verbose=False
+)
+
+# %%
+# Visualizing the row and column alignments learned by unbalanced Co-Optimal Transport.
+# -----------------------------------------------------------------------------------------
+#
+# Similar to Co-Optimal Transport, we are also be able to fully recover the clean OT plans.
+
+fig = pl.figure(4, (9, 7))
+pl.clf()
+
+ax1 = pl.subplot(2, 2, 3)
+pl.imshow(X, vmin=-2, vmax=2)
+pl.xlabel('$X$')
+
+ax2 = pl.subplot(2, 2, 2)
+ax2.yaxis.tick_right()
+pl.imshow(np.transpose(Y_noisy), vmin=-2, vmax=2)
+pl.title("Transpose(Noisy $Y$)")
+ax2.xaxis.tick_top()
+
+for i in range(n1):
+    j = np.argmax(pi_sample[i, :])
+    xyA = (d1 - .5, i)
+    xyB = (j, d2 - .5)
+    con = ConnectionPatch(xyA=xyA, xyB=xyB, coordsA=ax1.transData,
+                          coordsB=ax2.transData, color="black")
+    fig.add_artist(con)
+
+for i in range(d1):
+    j = np.argmax(pi_feature[i, :])
+    xyA = (i, -.5)
+    xyB = (-.5, j)
+    con = ConnectionPatch(
+        xyA=xyA, xyB=xyB, coordsA=ax1.transData, coordsB=ax2.transData, color="blue")
+    fig.add_artist(con)

master/_downloads/06f3dd7c3258690ab730be0a658b5e36/plot_ssw_unif_torch.zip

@@ -0,0 +1,133 @@
+{
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "\n# Detecting outliers by learning sample marginal distribution with CO-Optimal Transport and by using unbalanced Co-Optimal Transport\n\nIn this example, we consider two point clouds living in different Euclidean spaces, where the outliers\nare artifically injected into the target data. We illustrate two methods which allow to filter out\nthese outliers.\n\nThe first method requires learning the sample marginal distribution which minimizes\nthe CO-Optimal Transport distance [49] between two input spaces.\nMore precisely, given a source data\n$(X, \\mu_x^{(s)}, \\mu_x^{(f)})$ and a target matrix $Y$ associated with a fixed\nhistogram on features $\\mu_y^{(f)}$, we want to solve the following problem\n\n\\begin{align}\\min_{\\mu_y^{(s)} \\in \\Delta} \\text{COOT}\\left( (X, \\mu_x^{(s)}, \\mu_x^{(f)}), (Y, \\mu_y^{(s)}, \\mu_y^{(f)}) \\right)\\end{align}\n\nwhere $\\Delta$ is the probability simplex. This minimization is done with a\nsimple projected gradient descent in PyTorch. We use the automatic backend of POT that\nallows us to compute the CO-Optimal Transport distance with :func:`ot.coot.co_optimal_transport2`\nwith differentiable losses.\n\nThe second method simply requires direct application of unbalanced Co-Optimal Transport [71].\nMore precisely, it is enough to use the sample and feature coupling from solving\n\n\\begin{align}\\text{UCOOT}\\left( (X, \\mu_x^{(s)}, \\mu_x^{(f)}), (Y, \\mu_y^{(s)}, \\mu_y^{(f)}) \\right)\\end{align}\n\nwhere all the marginal distributions are uniform.\n\n.. [49] Redko, I., Vayer, T., Flamary, R., and Courty, N. (2020).\n   [CO-Optimal Transport](https://proceedings.neurips.cc/paper/2020/file/cc384c68ad503482fb24e6d1e3b512ae-Paper.pdf).\n   Advances in Neural Information Processing Systems, 33.\n.. [71] H. Tran, H. Janati, N. Courty, R. Flamary, I. Redko, P. Demetci & R. Singh (2023). [Unbalanced Co-Optimal Transport](https://dl.acm.org/doi/10.1609/aaai.v37i8.26193).\n    AAAI Conference on Artificial Intelligence.\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "# Author: Remi Flamary <[email protected]>\n#         Quang Huy Tran <[email protected]>\n# License: MIT License\n\nfrom matplotlib.patches import ConnectionPatch\nimport torch\nimport numpy as np\n\nimport matplotlib.pyplot as pl\nimport ot\n\nfrom ot.coot import co_optimal_transport as coot\nfrom ot.coot import co_optimal_transport2 as coot2\nfrom ot.gromov._unbalanced import unbalanced_co_optimal_transport"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Generate data\nThe source and clean target matrices are generated by\n$X_{i,j} = \\cos(\\frac{i}{n_1} \\pi) + \\cos(\\frac{j}{d_1} \\pi)$ and\n$Y_{i,j} = \\cos(\\frac{i}{n_2} \\pi) + \\cos(\\frac{j}{d_2} \\pi)$.\nThe target matrix is then contaminated by adding 5 row outliers.\nIntuitively, we expect that the estimated sample distribution should ignore these outliers,\ni.e. their weights should be zero.\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "np.random.seed(182)\n\nn1, d1 = 20, 16\nn2, d2 = 10, 8\nn = 15\n\nX = (\n    torch.cos(torch.arange(n1) * torch.pi / n1)[:, None] +\n    torch.cos(torch.arange(d1) * torch.pi / d1)[None, :]\n)\n\n# Generate clean target data mixed with outliers\nY_noisy = torch.randn((n, d2)) * 10.0\nY_noisy[:n2, :] = (\n    torch.cos(torch.arange(n2) * torch.pi / n2)[:, None] +\n    torch.cos(torch.arange(d2) * torch.pi / d2)[None, :]\n)\nY = Y_noisy[:n2, :]\n\nX, Y_noisy, Y = X.double(), Y_noisy.double(), Y.double()\n\nfig, axes = pl.subplots(nrows=1, ncols=3, figsize=(12, 5))\naxes[0].imshow(X, vmin=-2, vmax=2)\naxes[0].set_title('$X$')\n\naxes[1].imshow(Y, vmin=-2, vmax=2)\naxes[1].set_title('Clean $Y$')\n\naxes[2].imshow(Y_noisy, vmin=-2, vmax=2)\naxes[2].set_title('Noisy $Y$')\n\npl.tight_layout()"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Optimize the COOT distance with respect to the sample marginal distribution\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "losses = []\nlr = 1e-3\nniter = 1000\n\nb = torch.tensor(ot.unif(n), requires_grad=True)\n\nfor i in range(niter):\n\n    loss = coot2(X, Y_noisy, wy_samp=b, log=False, verbose=False)\n    losses.append(float(loss))\n\n    loss.backward()\n\n    with torch.no_grad():\n        b -= lr * b.grad  # gradient step\n        b[:] = ot.utils.proj_simplex(b)  # projection on the simplex\n\n    b.grad.zero_()\n\n# Estimated sample marginal distribution and training loss curve\npl.plot(losses[10:])\npl.title('CO-Optimal Transport distance')\n\nprint(f\"Marginal distribution = {b.detach().numpy()}\")"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Visualizing the row and column alignments with the estimated sample marginal distribution\n\nClearly, the learned marginal distribution completely and successfully ignores the 5 outliers.\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "X, Y_noisy = X.numpy(), Y_noisy.numpy()\nb = b.detach().numpy()\n\npi_sample, pi_feature = coot(X, Y_noisy, wy_samp=b, log=False, verbose=True)\n\nfig = pl.figure(4, (9, 7))\npl.clf()\n\nax1 = pl.subplot(2, 2, 3)\npl.imshow(X, vmin=-2, vmax=2)\npl.xlabel('$X$')\n\nax2 = pl.subplot(2, 2, 2)\nax2.yaxis.tick_right()\npl.imshow(np.transpose(Y_noisy), vmin=-2, vmax=2)\npl.title(\"Transpose(Noisy $Y$)\")\nax2.xaxis.tick_top()\n\nfor i in range(n1):\n    j = np.argmax(pi_sample[i, :])\n    xyA = (d1 - .5, i)\n    xyB = (j, d2 - .5)\n    con = ConnectionPatch(xyA=xyA, xyB=xyB, coordsA=ax1.transData,\n                          coordsB=ax2.transData, color=\"black\")\n    fig.add_artist(con)\n\nfor i in range(d1):\n    j = np.argmax(pi_feature[i, :])\n    xyA = (i, -.5)\n    xyB = (-.5, j)\n    con = ConnectionPatch(\n        xyA=xyA, xyB=xyB, coordsA=ax1.transData, coordsB=ax2.transData, color=\"blue\")\n    fig.add_artist(con)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "Now, let see if we can use unbalanced Co-Optimal Transport to recover the clean OT plans,\nwithout the need of learning the marginal distribution as in Co-Optimal Transport.\n-----------------------------------------------------------------------------------------\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "pi_sample, pi_feature = unbalanced_co_optimal_transport(\n    X=X, Y=Y_noisy, reg_marginals=(10, 10), epsilon=0, divergence=\"kl\",\n    unbalanced_solver=\"mm\", max_iter=1000, tol=1e-6,\n    max_iter_ot=1000, tol_ot=1e-6, log=False, verbose=False\n)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Visualizing the row and column alignments learned by unbalanced Co-Optimal Transport.\n\nSimilar to Co-Optimal Transport, we are also be able to fully recover the clean OT plans.\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "fig = pl.figure(4, (9, 7))\npl.clf()\n\nax1 = pl.subplot(2, 2, 3)\npl.imshow(X, vmin=-2, vmax=2)\npl.xlabel('$X$')\n\nax2 = pl.subplot(2, 2, 2)\nax2.yaxis.tick_right()\npl.imshow(np.transpose(Y_noisy), vmin=-2, vmax=2)\npl.title(\"Transpose(Noisy $Y$)\")\nax2.xaxis.tick_top()\n\nfor i in range(n1):\n    j = np.argmax(pi_sample[i, :])\n    xyA = (d1 - .5, i)\n    xyB = (j, d2 - .5)\n    con = ConnectionPatch(xyA=xyA, xyB=xyB, coordsA=ax1.transData,\n                          coordsB=ax2.transData, color=\"black\")\n    fig.add_artist(con)\n\nfor i in range(d1):\n    j = np.argmax(pi_feature[i, :])\n    xyA = (i, -.5)\n    xyB = (-.5, j)\n    con = ConnectionPatch(\n        xyA=xyA, xyB=xyB, coordsA=ax1.transData, coordsB=ax2.transData, color=\"blue\")\n    fig.add_artist(con)"
+      ]
+    }
+  ],
+  "metadata": {
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    },
+    "language_info": {
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "file_extension": ".py",
+      "mimetype": "text/x-python",
+      "name": "python",
+      "nbconvert_exporter": "python",
+      "pygments_lexer": "ipython3",
+      "version": "3.10.15"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 0
+}