From 6e5ac29c675220aea94eeaf5358e2853462ee120 Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Fri, 13 Nov 2026 22:11:49 +0000 Subject: [PATCH] Add pre-commit nbstripout for cleaner notebook diffs. Strip outputs and execution counts on commit; document install in README and normalize existing notebooks (including tutorial.ipynb). --- .pre-commit-config.yaml | 7 + README.md | 18 + .../notebooks/01_toy_hierarchy.ipynb | 38 +- .../notebooks/02_toy_warped_regression.ipynb | 40 +- .../notebooks/03_usps_digits.ipynb | 41 ++- .../notebooks/04_highfive_mrd.ipynb | 39 +- examples/tutorial.ipynb | 345 +++--------------- 7 files changed, 176 insertions(+), 352 deletions(-) create mode 100644 .pre-commit-config.yaml diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml new file mode 100644 index 0000000..63cc5ac --- /dev/null +++ b/.pre-commit-config.yaml @@ -0,0 +1,7 @@ +# Notebook cleanup: strip outputs / execution counts before commit. +# Install once: pip install pre-commit && pre-commit install +repos: + - repo: https://github.com/kynan/nbstripout + rev: 0.8.2 + hooks: + - id: nbstripout diff --git a/README.md b/README.md index 52bbaea..7da25d1 100644 --- a/README.md +++ b/README.md @@ -46,6 +46,24 @@ encoder and MRD (see `.coveragerc`). MPI parallel inference and prediction-only SVI helpers remain omitted until those paths are removed or revived (CIP-0001 Phases 2–4). +Pre-commit (notebooks) +---------------------- + +Commits strip Jupyter outputs and execution counts via +[nbstripout](https://github.com/kynan/nbstripout) so notebook diffs stay +readable. Install once after cloning:: + +```bash +python -m pip install pre-commit +pre-commit install +``` + +Hooks run automatically on ``git commit``. To strip existing notebooks:: + +```bash +pre-commit run nbstripout --all-files +``` + Multi-view / MRD (CIP-0002) --------------------------- diff --git a/examples/damianou2013/notebooks/01_toy_hierarchy.ipynb b/examples/damianou2013/notebooks/01_toy_hierarchy.ipynb index a1f4104..656968e 100644 --- a/examples/damianou2013/notebooks/01_toy_hierarchy.ipynb +++ b/examples/damianou2013/notebooks/01_toy_hierarchy.ipynb @@ -1,19 +1,11 @@ { - "nbformat": 4, - "nbformat_minor": 5, - "metadata": { - "kernelspec": { - "display_name": "Python 3", - "language": "python", - "name": "python3" - } - }, "cells": [ { "cell_type": "markdown", + "id": "0", "metadata": {}, "source": [ - "# Toy hierarchy \u2014 \u00a74.1 / fig. 2\n", + "# Toy hierarchy — §4.1 / fig. 2\n", "\n", "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/SheffieldML/PyDeepGP/blob/main/examples/damianou2013/notebooks/01_toy_hierarchy.ipynb)\n", "\n", @@ -26,6 +18,7 @@ }, { "cell_type": "markdown", + "id": "1", "metadata": {}, "source": [ "## Setup\n" @@ -34,6 +27,7 @@ { "cell_type": "code", "execution_count": null, + "id": "2", "metadata": {}, "outputs": [], "source": [ @@ -76,6 +70,7 @@ { "cell_type": "code", "execution_count": null, + "id": "3", "metadata": {}, "outputs": [], "source": [ @@ -92,6 +87,7 @@ { "cell_type": "code", "execution_count": null, + "id": "4", "metadata": {}, "outputs": [], "source": [ @@ -104,6 +100,7 @@ }, { "cell_type": "markdown", + "id": "5", "metadata": {}, "source": [ "## Load Matlab ground-truth hierarchy\n" @@ -112,6 +109,7 @@ { "cell_type": "code", "execution_count": null, + "id": "6", "metadata": {}, "outputs": [], "source": [ @@ -134,14 +132,16 @@ }, { "cell_type": "markdown", + "id": "7", "metadata": {}, "source": [ - "## Fit two-view DeepGP (Q shared \u2192 top)\n" + "## Fit two-view DeepGP (Q shared → top)\n" ] }, { "cell_type": "code", "execution_count": null, + "id": "8", "metadata": {}, "outputs": [], "source": [ @@ -163,6 +163,7 @@ }, { "cell_type": "markdown", + "id": "9", "metadata": {}, "source": [ "## Recovered latents (dominant ARD dims)\n" @@ -171,6 +172,7 @@ { "cell_type": "code", "execution_count": null, + "id": "10", "metadata": {}, "outputs": [], "source": [ @@ -192,6 +194,7 @@ }, { "cell_type": "markdown", + "id": "11", "metadata": {}, "source": [ "## Baselines: stacked PCA and Isomap\n" @@ -200,6 +203,7 @@ { "cell_type": "code", "execution_count": null, + "id": "12", "metadata": {}, "outputs": [], "source": [ @@ -226,6 +230,7 @@ { "cell_type": "code", "execution_count": null, + "id": "13", "metadata": {}, "outputs": [], "source": [ @@ -250,5 +255,14 @@ "plt.show()\n" ] } - ] + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + } + }, + "nbformat": 4, + "nbformat_minor": 5 } diff --git a/examples/damianou2013/notebooks/02_toy_warped_regression.ipynb b/examples/damianou2013/notebooks/02_toy_warped_regression.ipynb index 789c1bc..a0b50af 100644 --- a/examples/damianou2013/notebooks/02_toy_warped_regression.ipynb +++ b/examples/damianou2013/notebooks/02_toy_warped_regression.ipynb @@ -1,19 +1,11 @@ { - "nbformat": 4, - "nbformat_minor": 5, - "metadata": { - "kernelspec": { - "display_name": "Python 3", - "language": "python", - "name": "python3" - } - }, "cells": [ { "cell_type": "markdown", + "id": "0", "metadata": {}, "source": [ - "# Warped regression \u2014 \u00a74.1 / fig. 3\n", + "# Warped regression — §4.1 / fig. 3\n", "\n", "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/SheffieldML/PyDeepGP/blob/main/examples/damianou2013/notebooks/02_toy_warped_regression.ipynb)\n", "\n", @@ -26,6 +18,7 @@ }, { "cell_type": "markdown", + "id": "1", "metadata": {}, "source": [ "## Setup\n" @@ -34,6 +27,7 @@ { "cell_type": "code", "execution_count": null, + "id": "2", "metadata": {}, "outputs": [], "source": [ @@ -76,6 +70,7 @@ { "cell_type": "code", "execution_count": null, + "id": "3", "metadata": {}, "outputs": [], "source": [ @@ -91,11 +86,12 @@ { "cell_type": "code", "execution_count": null, + "id": "4", "metadata": {}, "outputs": [], "source": [ "QUICK = True\n", - "PAPER = False # True \u2192 N=120, D=10\n", + "PAPER = False # True → N=120, D=10\n", "n_toy, d_toy = (120, 10) if PAPER else (150, 15)\n", "n_tr = 25\n", "q = 3 if QUICK else 6\n", @@ -109,6 +105,7 @@ }, { "cell_type": "markdown", + "id": "5", "metadata": {}, "source": [ "## Generative process (`hierGpsNEW`)\n" @@ -117,6 +114,7 @@ { "cell_type": "code", "execution_count": null, + "id": "6", "metadata": {}, "outputs": [], "source": [ @@ -135,6 +133,7 @@ }, { "cell_type": "markdown", + "id": "7", "metadata": {}, "source": [ "## Fit sparse GP and DeepGP across repeats\n" @@ -143,6 +142,7 @@ { "cell_type": "code", "execution_count": null, + "id": "8", "metadata": {}, "outputs": [], "source": [ @@ -171,6 +171,7 @@ { "cell_type": "code", "execution_count": null, + "id": "9", "metadata": {}, "outputs": [], "source": [ @@ -188,13 +189,14 @@ " print(f\"Repeat {r + 1}/{repeats}: MSE GP={gp_mses[-1]:.4f} DeepGP={dgp_mses[-1]:.4f}\")\n", "\n", "print(\n", - " f\"Mean MSE (\u00b1std): GP {np.mean(gp_mses):.4f}\u00b1{np.std(gp_mses):.4f} \"\n", - " f\"DeepGP {np.mean(dgp_mses):.4f}\u00b1{np.std(dgp_mses):.4f}\"\n", + " f\"Mean MSE (±std): GP {np.mean(gp_mses):.4f}±{np.std(gp_mses):.4f} \"\n", + " f\"DeepGP {np.mean(dgp_mses):.4f}±{np.std(dgp_mses):.4f}\"\n", ")\n" ] }, { "cell_type": "markdown", + "id": "10", "metadata": {}, "source": [ "## Fig. 3(b)-style MSE plot\n" @@ -203,6 +205,7 @@ { "cell_type": "code", "execution_count": null, + "id": "11", "metadata": {}, "outputs": [], "source": [ @@ -218,5 +221,14 @@ "plt.show()\n" ] } - ] + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + } + }, + "nbformat": 4, + "nbformat_minor": 5 } diff --git a/examples/damianou2013/notebooks/03_usps_digits.ipynb b/examples/damianou2013/notebooks/03_usps_digits.ipynb index 492af9a..efda38f 100644 --- a/examples/damianou2013/notebooks/03_usps_digits.ipynb +++ b/examples/damianou2013/notebooks/03_usps_digits.ipynb @@ -1,31 +1,24 @@ { - "nbformat": 4, - "nbformat_minor": 5, - "metadata": { - "kernelspec": { - "display_name": "Python 3", - "language": "python", - "name": "python3" - } - }, "cells": [ { "cell_type": "markdown", + "id": "0", "metadata": {}, "source": [ - "# USPS digits \u2014 \u00a74.3 / figs. 6\u20138\n", + "# USPS digits — §4.3 / figs. 6–8\n", "\n", "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/SheffieldML/PyDeepGP/blob/main/examples/damianou2013/notebooks/03_usps_digits.ipynb)\n", "\n", "Port of Matlab [`demUsps.m`](https://github.com/SheffieldML/deepGP/blob/master/deepGP/matlab/demUsps.m) /\n", "[`demDigitsDemonstration.m`](https://github.com/SheffieldML/deepGP/blob/master/deepGP/matlab/demDigitsDemonstration.m):\n", - "`vercat2` (50\u00d7{0,6,1}), depth-5 widths Q={15,12,10,8,6}, K=80.\n", + "`vercat2` (50×{0,6,1}), depth-5 widths Q={15,12,10,8,6}, K=80.\n", "\n", "Metrics: variational bound and leave-one-out 1-NN error in the top latent space.\n" ] }, { "cell_type": "markdown", + "id": "1", "metadata": {}, "source": [ "## Setup\n" @@ -34,6 +27,7 @@ { "cell_type": "code", "execution_count": null, + "id": "2", "metadata": {}, "outputs": [], "source": [ @@ -76,6 +70,7 @@ { "cell_type": "code", "execution_count": null, + "id": "3", "metadata": {}, "outputs": [], "source": [ @@ -92,10 +87,11 @@ { "cell_type": "code", "execution_count": null, + "id": "4", "metadata": {}, "outputs": [], "source": [ - "QUICK = True # True \u2192 fewer digits / smaller Q / depths 1\u20132\n", + "QUICK = True # True → fewer digits / smaller Q / depths 1–2\n", "DEPTHS = [1, 2] if QUICK else [1, 2, 3, 4, 5]\n", "per_class = 20 if QUICK else 50\n", "num_inducing = 25 if QUICK else 80\n", @@ -105,6 +101,7 @@ }, { "cell_type": "markdown", + "id": "5", "metadata": {}, "source": [ "## Load USPS `vercat2` subset\n" @@ -113,6 +110,7 @@ { "cell_type": "code", "execution_count": null, + "id": "6", "metadata": {}, "outputs": [], "source": [ @@ -124,6 +122,7 @@ }, { "cell_type": "markdown", + "id": "7", "metadata": {}, "source": [ "## Fit DeepGPs at increasing depth\n" @@ -132,6 +131,7 @@ { "cell_type": "code", "execution_count": null, + "id": "8", "metadata": {}, "outputs": [], "source": [ @@ -168,14 +168,16 @@ }, { "cell_type": "markdown", + "id": "9", "metadata": {}, "source": [ - "## Top latent scatter and ARD (figs. 6\u20138 style)\n" + "## Top latent scatter and ARD (figs. 6–8 style)\n" ] }, { "cell_type": "code", "execution_count": null, + "id": "10", "metadata": {}, "outputs": [], "source": [ @@ -202,6 +204,7 @@ { "cell_type": "code", "execution_count": null, + "id": "11", "metadata": {}, "outputs": [], "source": [ @@ -220,6 +223,7 @@ { "cell_type": "code", "execution_count": null, + "id": "12", "metadata": {}, "outputs": [], "source": [ @@ -235,5 +239,14 @@ "plt.show()\n" ] } - ] + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + } + }, + "nbformat": 4, + "nbformat_minor": 5 } diff --git a/examples/damianou2013/notebooks/04_highfive_mrd.ipynb b/examples/damianou2013/notebooks/04_highfive_mrd.ipynb index c8d3002..61988d5 100644 --- a/examples/damianou2013/notebooks/04_highfive_mrd.ipynb +++ b/examples/damianou2013/notebooks/04_highfive_mrd.ipynb @@ -1,29 +1,22 @@ { - "nbformat": 4, - "nbformat_minor": 5, - "metadata": { - "kernelspec": { - "display_name": "Python 3", - "language": "python", - "name": "python3" - } - }, "cells": [ { "cell_type": "markdown", + "id": "0", "metadata": {}, "source": [ - "# High-five MRD \u2014 \u00a74.2 / figs. 4\u20135\n", + "# High-five MRD — §4.2 / figs. 4–5\n", "\n", "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/SheffieldML/PyDeepGP/blob/main/examples/damianou2013/notebooks/04_highfive_mrd.ipynb)\n", "\n", "Port of Matlab [`demHighFiveDemonstration.m`](https://github.com/SheffieldML/deepGP/blob/master/deepGP/matlab/demHighFiveDemonstration.m)\n", - "using `demHighFiveHgplvm1.mat` (subjects A/B, 78\u00d762). Fits a two-view DeepGP and\n", + "using `demHighFiveHgplvm1.mat` (subjects A/B, 78×62). Fits a two-view DeepGP and\n", "compares ARD / latents to the stored hgplvm reference spaces when present.\n" ] }, { "cell_type": "markdown", + "id": "1", "metadata": {}, "source": [ "## Setup\n" @@ -32,6 +25,7 @@ { "cell_type": "code", "execution_count": null, + "id": "2", "metadata": {}, "outputs": [], "source": [ @@ -74,6 +68,7 @@ { "cell_type": "code", "execution_count": null, + "id": "3", "metadata": {}, "outputs": [], "source": [ @@ -90,6 +85,7 @@ { "cell_type": "code", "execution_count": null, + "id": "4", "metadata": {}, "outputs": [], "source": [ @@ -102,6 +98,7 @@ }, { "cell_type": "markdown", + "id": "5", "metadata": {}, "source": [ "## Load high-five views\n" @@ -110,6 +107,7 @@ { "cell_type": "code", "execution_count": null, + "id": "6", "metadata": {}, "outputs": [], "source": [ @@ -123,6 +121,7 @@ }, { "cell_type": "markdown", + "id": "7", "metadata": {}, "source": [ "## Fit two-view MRD DeepGP\n" @@ -131,6 +130,7 @@ { "cell_type": "code", "execution_count": null, + "id": "8", "metadata": {}, "outputs": [], "source": [ @@ -154,6 +154,7 @@ }, { "cell_type": "markdown", + "id": "9", "metadata": {}, "source": [ "## ARD bars (fig. 4 style)\n" @@ -162,6 +163,7 @@ { "cell_type": "code", "execution_count": null, + "id": "10", "metadata": {}, "outputs": [], "source": [ @@ -181,6 +183,7 @@ }, { "cell_type": "markdown", + "id": "11", "metadata": {}, "source": [ "## Latent projections (fig. 5 style)\n" @@ -189,6 +192,7 @@ { "cell_type": "code", "execution_count": null, + "id": "12", "metadata": {}, "outputs": [], "source": [ @@ -209,6 +213,7 @@ }, { "cell_type": "markdown", + "id": "13", "metadata": {}, "source": [ "## Reference hgplvm latents from the Matlab mat (if present)\n" @@ -217,6 +222,7 @@ { "cell_type": "code", "execution_count": null, + "id": "14", "metadata": {}, "outputs": [], "source": [ @@ -235,5 +241,14 @@ " print(\"No Xparent/Xchild* fields in mat.\")\n" ] } - ] + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + } + }, + "nbformat": 4, + "nbformat_minor": 5 } diff --git a/examples/tutorial.ipynb b/examples/tutorial.ipynb index c319ba2..8a3ecad 100644 --- a/examples/tutorial.ipynb +++ b/examples/tutorial.ipynb @@ -9,10 +9,8 @@ }, { "cell_type": "code", - "execution_count": 1, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "import numpy as np\n", @@ -26,10 +24,8 @@ }, { "cell_type": "code", - "execution_count": 2, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Add parent directory to path if deepgp is not there already\n", @@ -40,10 +36,8 @@ }, { "cell_type": "code", - "execution_count": 3, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "import deepgp" @@ -51,10 +45,8 @@ }, { "cell_type": "code", - "execution_count": 4, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Utility to load sample data. It can be installed with pip. Otherwise just load some other data.\n", @@ -70,10 +62,8 @@ }, { "cell_type": "code", - "execution_count": 5, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Load data\n", @@ -100,10 +90,8 @@ }, { "cell_type": "code", - "execution_count": 7, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "m = deepgp.DeepGP([Y.shape[1],Q1,Q2],\n", @@ -115,10 +103,8 @@ }, { "cell_type": "code", - "execution_count": 8, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Make sure initial noise variance gives a reasonable signal to noise ratio\n", @@ -129,82 +115,18 @@ }, { "cell_type": "code", - "execution_count": 9, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Running L-BFGS-B (Scipy implementation) Code:\n", - " runtime i f |g| \n", - " 00s12 0001 7.755559e+04 8.005322e+09 \n", - " 01s12 0037 1.108158e+02 7.873395e+02 \n", - " 03s13 0121 -5.507231e+01 1.949809e+02 \n", - " 11s21 0447 -1.728118e+02 2.852418e+01 \n", - " 35s51 1340 -2.465651e+02 1.361611e+02 \n", - " 49s69 1848 -2.566525e+02 1.575437e+01 \n", - " 02m26s44 5002 -2.615853e+02 1.870644e+00 \n", - "Runtime: 02m26s44\n", - "Optimization status: Maximum number of f evaluations reached\n", - "\n" - ] - } - ], + "outputs": [], "source": [ "m.optimize(max_iters=5000, messages=True)" ] }, { "cell_type": "code", - "execution_count": 10, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "name": "stderr", - "output_type": "stream", - "text": [ - " /Users/neil/anaconda3/lib/python3.6/site-packages/matplotlib/__init__.py:1710: MatplotlibDeprecationWarning:The *left* kwarg to `bar` is deprecated use *x* instead. Support for *left* will be removed in Matplotlib 3.0\n" - ] - }, - { - "data": { - "text/plain": [ - "" - ] - }, - "execution_count": 10, - "metadata": {}, - "output_type": "execute_result" - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - " /Users/neil/anaconda3/lib/python3.6/site-packages/matplotlib/figure.py:1999: UserWarning:This figure includes Axes that are not compatible with tight_layout, so results might be incorrect.\n" - ] - }, - { - "data": { - "image/png": "iVBORw0KGgoAAAANSUhEUgAAAagAAAEYCAYAAAAJeGK1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAFC9JREFUeJzt3X+UZ3V93/Hni9+CyI8wROuii0ox\nxirUgaokpmLI2RqLiTGJRAm14KaJSdAkoBzPqfbHObXRmnhiKtkiRSMlx6A2mihCUKBpCbD8FFk0\nlqLZaNgxKhAiPxbe/eN7B8dxZ/fuzt7v/czO83HOnO/9Nffznssyr/ncH5+bqkKSpNbsNXYBkiRt\niwElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJatI+YxfQxxFHHFFr164du4wV\nb9NXvjl2CVP1Q08/fOwSJG3DjTfe+I2qmtnRdisioNauXcvGjRvHLmPFO2H9xWOXMFU3bHjt2CVI\n2oYkX+mznaf4JElNMqAkSU0yoCRJTTKgJElNMqAkSU0aLKCSXJhkS5Lbt7Hut5JUkiOGal+StLIN\n2YO6CFi3eGGSo4BTgK8O2LYkaYUbLKCq6hpgW0+G/g5wLuC75iVJS5rqNagkpwJ/U1W39th2fZKN\nSTbOzc1NoTpJUkumFlBJDgTeBvzbPttX1Yaqmq2q2ZmZHY6IIUnaw0yzB/VM4Gjg1iR3A2uAm5I8\neYo1SJJWiKmNxVdVnweOnJ/vQmq2qr4xrRokSSvHkLeZXwJcCxybZHOSM4dqS5K05xmsB1VVp+1g\n/dqh2pYkrXyOJCFJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJ\napIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqSASVJapIBJUlqkgElSWqS\nASVJatJgAZXkwiRbkty+YNm7ktyZ5LYkH09y6FDtS5JWtiF7UBcB6xYtuwJ4blU9D/gScN6A7UuS\nVrDBAqqqrgG+uWjZ5VW1tZv9S2DNUO1Lkla2Ma9B/Wvg0yO2L0lq2CgBleRtwFbg4u1ssz7JxiQb\n5+bmplecJKkJUw+oJGcArwBeW1W11HZVtaGqZqtqdmZmZnoFSpKasM80G0uyDngL8GNV9Q/TbFuS\ntLIMeZv5JcC1wLFJNic5E3gfcDBwRZJbkpw/VPuSpJVtsB5UVZ22jcUfGKo9SdKexZEkJElNMqAk\nSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElN\nMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0aLKCSXJhk\nS5LbFyw7PMkVSf6q+zxsqPYlSSvbkD2oi4B1i5a9Fbiyqo4BruzmJUn6PoMFVFVdA3xz0eJXAh/s\npj8I/NRQ7UuSVrZpX4P6war6OkD3eeRSGyZZn2Rjko1zc3NTK1CS1IZmb5Koqg1VNVtVszMzM2OX\nI0masmkH1D1JngLQfW6ZcvuSpBVi2gH1CeCMbvoM4E+m3L4kaYUY8jbzS4BrgWOTbE5yJvBO4JQk\nfwWc0s1LkvR99umzUZIXA2sXbl9VH9re91TVaUuselnf4iRJq9cOAyrJHwLPBG4BHu0WF7DdgJIk\naTn69KBmgedUVQ1djCRJ8/pcg7odePLQhUiStFCfHtQRwB1Jrgceml9YVacOVpUkadXrE1DvGLoI\nSZIW22FAVdXVSX4QOKFbdH1V+YCtJGlQO7wGleTngOuBnwV+DrguyauHLkyStLr1OcX3NuCE+V5T\nkhngz4FLhyxMkrS69bmLb69Fp/T+ruf3SZK0y/r0oC5L8hngkm7+54FPDVeSJEn9bpI4J8nPACcB\nATZU1ccHr0yStKr1Gouvqj4KfHTgWiRJetySAZXkL6rqR5Lcz2TsvcdXAVVVTxq8OknSqrVkQFXV\nj3SfB0+vHEmSJvo8B/WHfZZJkrQ79bld/IcXziTZB3jBMOVIkjSxZEAlOa+7/vS8JPd1X/cD9+Cr\n2iVJA1syoKrqP3XXn95VVU/qvg6uqh+oqvOmWKMkaRXqc5v5p5O8ZPHCqrpmgHokSQL6BdQ5C6YP\nAE4EbgROHqQiSZLoN5LEv1w4n+Qo4LcHq0iSJHZt0NfNwHN3dyGSJC20wx5Ukt/juyNJ7AUcB9w6\nZFGSJPW5BrVxwfRW4JKq+t/LaTTJm4GzmATf54HXV9WDy9mnJGnP0uca1AeT7Ac8m0mgfHE5DSZ5\nKvDrwHOq6jtJPgK8BrhoOfuVJO1Z+pzieznwB8D/ZTJQ7NFJfqmqPr3Mdp+Q5BHgQOBry9iXJGkP\n1OcU33uAl1bVlwGSPBP4M2CXAqqq/ibJu4GvAt8BLq+qyxdvl2Q9sB7gaU972q40JUlawfrcxbdl\nPpw6dwFbltp4R5IcBrwSOBr4R8BBSV63eLuq2lBVs1U1OzMzs6vNSZJWqO29D+pV3eQXknwK+AiT\na1A/C9ywjDZ/HPh/VTXXtfMx4MXAh5exT0nSHmZ7p/gWPqB7D/Bj3fQccNgy2vwq8MIkBzI5xfcy\nvvdOQUmStvvCwtcP0WBVXZfkUuAmJret3wxsGKItSdLKtb1TfOdW1W8velD3cVX167vaaFW9HXj7\nrn6/JGnPt71TfJu6T0+/SZKmbnun+D6ZZG/guVV1zlLbSZI0hO3eZl5Vj+Lr3SVJI+jzoO7NST4B\n/DHwwPzCqvrYYFVJkla9PgF1OPB3fO8LCgswoCRJg+kTUBcsHr08yUkD1SNJEtBvqKPf67lMkqTd\nZnvPQb2IyRBEM0l+Y8GqJwF7D12YJGl1294pvv2AJ3bbHLxg+X3Aq4csSpKk7T0HdTVwdZKLquor\nU6xJkqReN0nsn2QDsHbh9lV18pLfIUnSMvUJqD8GzgcuAB4dthxJkib6BNTWqnr/4JVIkrRAn9vM\nP5nkV5I8Jcnh81+DVyZJWtX69KDO6D4XDhhbwDN2fzmSJE3sMKCq6uhpFCJJ0kI7DKgk+wK/DLyk\nW3QV8AdV9ciAdUmSVrk+p/jeD+wL/Ndu/vRu2VlDFSVJUp+AOqGqnr9g/rNJbh2qIEmSoN9dfI8m\neeb8TJJn4PNQkqSB9elBnQN8LsldQICnA68ftCpJ0qrX5y6+K5McAxzLJKDurKqHBq9MkrSq7fAU\nX5I3Ak+oqtuq6lbgwCS/spxGkxya5NIkdybZ1L3aQ5Kkx/W5BvWGqvr2/ExVfQt4wzLbfS9wWVU9\nG3g+sGmZ+5Mk7WH6XIPaK0mqqgCS7M3kXVG7JMmTmDxT9a8Aquph4OFd3Z8kac/Upwf1GeAjSV6W\n5GTgEuCyZbT5DGAO+O9Jbk5yQZKDlrE/SdIeqE9AvQW4ksloEm/sps9dRpv7AP8UeH9VHQ88ALx1\n8UZJ1ifZmGTj3NzcMpqTJK1Efe7ie4zJ+6DO301tbgY2V9V13fylbCOgqmoDsAFgdna2dlPbkqQV\nok8Pareqqr8F/jrJsd2ilwF3TLsOSVLb+twkMYRfAy5Osh9wFz74K0laZJSAqqpbgNkx2pYkrQxL\nBlSSTzJ5MeE2VdWpg1QkSRLb70G9u/t8FfBk4MPd/GnA3QPWJEnS0gFVVVcDJPkPVfWSBas+meSa\nwSuTJK1qfe7im+lesQFAkqOBmeFKkiSp300Sbwau6l63AbAWWD9YRZIksYOASrIXcB9wDPDsbrGv\n25AkDW67AVVVjyX5L1X1IsDXvEuSpqbPNajLk/xMkgxejSRJnT7XoH4DOAjYmuRBJm/Vrap60qCV\nSZJWtT6DxR48jUIkSVqo11BHSQ5jcqPEAfPLqspnoSRJg9lhQCU5CzgbWAPcArwQuBY4edjSJEmr\nWZ+bJM4GTgC+UlUvBY5n8kZcSZIG0yegHqyqBwGS7F9VdwLH7uB7JElalj7XoDYnORT4n8AVSb4F\nfG3YsiRJq12fu/h+upt8R5LPAYcAlw1alSRp1dupFxbOj3AuSdLQ+lyDkiRp6gwoSVKTDChJUpMM\nKElSkwwoSVKTDChJUpNGC6gkeye5OcmfjlWDJKldY/agzgY2jdi+JKlhowRUkjXATwIXjNG+JKl9\nY/Wgfhc4F3hsqQ2SrE+yMcnGuTkHT5ek1WbqAZXkFcCWqrpxe9tV1Yaqmq2q2ZmZmSlVJ0lqxRg9\nqJOAU5PcDfwRcHKSD49QhySpYVMPqKo6r6rWVNVa4DXAZ6vqddOuQ5LUNp+DkiQ1aadet7G7VdVV\nwFVj1iBJapM9KElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElS\nkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMMKElSkwwoSVKTDChJUpMM\nKElSkwwoSVKT9pl2g0mOAj4EPBl4DNhQVe+ddh3S9pyw/uKxS5iaGza8duwSpG2aekABW4HfrKqb\nkhwM3Jjkiqq6Y4RaJEmNmvopvqr6elXd1E3fD2wCnjrtOiRJbRv1GlSStcDxwHXbWLc+ycYkG+fm\n5qZdmiRpZKMFVJInAh8F3lRV9y1eX1Ubqmq2qmZnZmamX6AkaVSjBFSSfZmE08VV9bExapAktW3q\nAZUkwAeATVX1nmm3L0laGcboQZ0EnA6cnOSW7uvlI9QhSWrY1G8zr6q/ADLtdiVJK4sjSUiSmmRA\nSZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmS\nmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkppkQEmSmmRASZKaZEBJkpq0z9gFLNfmzZvZsmXL2GVM\nzZFHHsmaNWvGLkOSBreiA+ree+/lWc86hoceenDsUqZm//0P4J57/pZDDjlk7FIkaVCjBFSSdcB7\ngb2BC6rqnbuynwceeICHHnqQZ607h30P2PN/YT/y4L18+bJ38cADDxhQkvZ4Uw+oJHsDvw+cAmwG\nbkjyiaq6Y1f3eeDhT2O/gw7bXSU26+EHvjV2CZI0NWP0oE4EvlxVdwEk+SPglcAuB9Qj3/n2biqt\nbavl55QkgFTVdBtMXg2sq6qzuvnTgX9WVb+6aLv1wPpu9ljgi1MtdMeOAL4xdhErhMeqP49VPx6n\n/lo8Vk+vqpkdbTRGDyrbWPZ9KVlVG4ANw5eza5JsrKrZsetYCTxW/Xms+vE49beSj9UYz0FtBo5a\nML8G+NoIdUiSGjZGQN0AHJPk6CT7Aa8BPjFCHZKkhk39FF9VbU3yq8BnmNxmfmFVfWHadewGzZ5+\nbJDHqj+PVT8ep/5W7LGa+k0SkiT14Vh8kqQmGVCSpCYZULsgybokX0zy5SRvHbueViW5MMmWJLeP\nXUvLkhyV5HNJNiX5QpKzx66pVUkOSHJ9klu7Y/Xvxq6pZUn2TnJzkj8du5ZdYUDtpAVDNf0L4DnA\naUmeM25VzboIWDd2ESvAVuA3q+qHgBcCb/Tf1JIeAk6uqucDxwHrkrxw5JpadjawaewidpUBtfMe\nH6qpqh4G5odq0iJVdQ3wzbHraF1Vfb2qbuqm72fyC+Wp41bVppr4+2523+7LO722Icka4CeBC8au\nZVcZUDvvqcBfL5jfjL9MtJskWQscD1w3biXt6k5b3QJsAa6oKo/Vtv0ucC7w2NiF7CoDauf1GqpJ\n2llJngh8FHhTVd03dj2tqqpHq+o4JqPQnJjkuWPX1JokrwC2VNWNY9eyHAbUznOoJu12SfZlEk4X\nV9XHxq5nJaiqbwNX4XXObTkJODXJ3UwuQ5yc5MPjlrTzDKid51BN2q2SBPgAsKmq3jN2PS1LMpPk\n0G76CcCPA3eOW1V7quq8qlpTVWuZ/I76bFW9buSydpoBtZOqaiswP1TTJuAjK3SopsEluQS4Fjg2\nyeYkZ45dU6NOAk5n8lfuLd3Xy8cuqlFPAT6X5DYmfyxeUVUr8hZq7ZhDHUmSmmQPSpLUJANKktQk\nA0qS1CQDSpLUJANKktQkA0qrSpL/M8A+1yb5hZ7bXpTk1d30BWMMCpvk3yT5xWm3K+2sqb/yXRpT\nVb14gN2uBX4B+B87WctZA9TSp93zx2hX2ln2oLSqJPn77vOfJ7kqyaVJ7kxycTeiA0nuTvKfu/cO\nXZ/kWd3yx3s/C/cFvBP40e4B2zcvai9J3pfkjiR/Bhy5YN1VSWbn99W1eWOSP09yYrf+riSndtvs\nneRdSW5IcluSX+rxs7yza/u2JO/ulr0jyW9108cl+ctu/ceTHLagtvlj8KUkP7rb/2NIO2BAaTU7\nHngTk/d6PYPJiA7z7quqE4H3MRkVenveCvyvqjquqn5n0bqfBo4F/gnwBmCpHtxBwFVV9QLgfuA/\nAqd03//vu23OBO6tqhOAE4A3JDl6qZ8lyeHd9/9wVT2v2+diHwLe0q3/PPD2Bev26Y7BmxYtl6bC\ngNJqdn1Vba6qx4BbmJyqm3fJgs8XLaONlwCXdCNwfw347BLbPQxc1k1/Hri6qh7ppufr+gngF7tX\nTVwH/ABwzHZ+lvuAB4ELkrwK+IeFDSY5BDi0qq7uFn2wq3fe/KC1N/K9x0aaCgNKq9lDC6Yf5Xuv\nydY2prfS/T/TnULbr2c7fcYTe6S+O+7YY/O1dYEzX1eAX+t6asdV1dFVdflSP0s3buSJTEZJ/ym+\nG4B9ze9z8bGRpsKAkrbt5xd8XttN3w28oJt+JZO3ucLklNzBS+znGuA13fWjpwAvXUZNnwF+uXs1\nB0n+cZKDltq4e7/UIVX1KSan6Y5buL6q7gW+teD60unA1UiN8K8iadv2T3Idkz/iTuuW/TfgT5Jc\nD1wJPNAtvw3YmuRW4KJF16E+DpzM5FTdl1heAFzA5FTbTV0Pbo5Jz2gpB3f1HsCk9/XmbWxzBnB+\nkgOBu4DXL6M+abdyNHNpke4lb7NV9Y2xa5FWM0/xSZKaZA9KktQke1CSpCYZUJKkJhlQkqQmGVCS\npCYZUJKkJv1/dgccXeK756IAAAAASUVORK5CYII=\n", - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "image/png": "iVBORw0KGgoAAAANSUhEUgAAAagAAAEYCAYAAAAJeGK1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ndN4AAAAASUVORK5CYII=\n", - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], + "outputs": [], "source": [ "# Plot ARD scales per layer\n", "m.obslayer.kern.plot_ARD()\n", @@ -213,40 +135,9 @@ }, { "cell_type": "code", - "execution_count": 11, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "Text(0.5,1,'Layer 1')" - ] - }, - "execution_count": 11, - "metadata": {}, - "output_type": "execute_result" - }, - { - "data": { - "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAE/CAYAAADG70EzAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLAeHp5lZDv8f0iDu2XLgT6kAAAAASUVORK5CYII=\n", - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT4AAAE/CAYAAADfZK+VAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n", - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], + "outputs": [], "source": [ "# From the plots above, we see which ones are the dominant dimensions for each layer. \n", "# So we use these dimensions in the visualization of the latent space below.\n", @@ -265,10 +156,8 @@ }, { "cell_type": "code", - "execution_count": 12, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Load some mocap data.\n", @@ -299,10 +188,8 @@ }, { "cell_type": "code", - "execution_count": 13, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Number of latent dimensions (single hidden layer, since the top layer is observed)\n", @@ -324,10 +211,8 @@ }, { "cell_type": "code", - "execution_count": 14, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "m = deepgp.DeepGP([Y_tr.shape[1],Q,X_tr.shape[1]],\n", @@ -337,10 +222,8 @@ }, { "cell_type": "code", - "execution_count": 15, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Make sure initial noise variance gives a reasonable signal to noise ratio.\n", @@ -353,99 +236,26 @@ }, { "cell_type": "code", - "execution_count": 16, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "\n", - "\n", - "

\n", - "Model: deepgp
\n", - "Objective: 27729.285582716577
\n", - "Number of Parameters: 14492
\n", - "Number of Optimization Parameters: 14490
\n", - "Updates: True
\n", - "

\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
deepgp. valueconstraintspriors
obslayer.inducing inputs (40, 5)
obslayer.sum.rbf.variance 1.0 +ve
obslayer.sum.rbf.lengthscale (5,) +ve
obslayer.sum.bias.variance 1.0 +ve
obslayer.Gaussian_noise.variance 2.13179847772 +ve fixed
obslayer.Kuu_var (40,) +ve
obslayer.latent space.mean (100, 5)
obslayer.latent space.variance (100, 5) +ve
layer_1.inducing inputs (40, 55)
layer_1.sum.rbf.variance 1.0 +ve
layer_1.sum.rbf.lengthscale 1.0 +ve
layer_1.sum.bias.variance 1.0 +ve
layer_1.Gaussian_noise.variance 0.0100039017522 +ve fixed
layer_1.Kuu_var (40,) +ve
layer_1.latent space.mean (100, 55)
layer_1.latent space.variance (100, 55) +ve
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], + "outputs": [], "source": [ "display(m)" ] }, { "cell_type": "code", - "execution_count": 17, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Running L-BFGS-B (Scipy implementation) Code:\n", - " runtime i f |g| \n", - " 00s18 001 2.772929e+04 2.660103e+08 \n", - " 01s20 040 5.172149e+03 2.273552e+05 \n", - " 05s22 196 1.541046e+03 1.879665e+02 \n", - " 06s24 236 1.530905e+03 7.159914e+02 \n", - " 19s41 743 1.429718e+03 1.042016e+02 \n", - " 20s97 802 1.425838e+03 7.004558e+01 \n", - "Runtime: 20s97\n", - "Optimization status: Maximum number of f evaluations reached\n", - "\n" - ] - } - ], + "outputs": [], "source": [ "m.optimize(max_iters=800, messages=True)" ] }, { "cell_type": "code", - "execution_count": 18, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "# Unfix noise variance now that we have initialized the model\n", @@ -455,26 +265,9 @@ }, { "cell_type": "code", - "execution_count": 19, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Running L-BFGS-B (Scipy implementation) Code:\n", - " runtime i f |g| \n", - " 00s14 0001 1.425838e+03 3.782786e+03 \n", - " 03s18 0101 1.307856e+03 8.537983e+02 \n", - " 10s24 0339 1.293169e+03 2.541807e+01 \n", - " 32s56 1027 1.292130e+03 1.878501e+01 \n", - " 48s12 1502 1.291413e+03 1.317024e+01 \n", - "Runtime: 48s12\n", - "Optimization status: Maximum number of f evaluations reached\n", - "\n" - ] - } - ], + "outputs": [], "source": [ "m.optimize(max_iters=1500, messages=True)" ] @@ -488,10 +281,8 @@ }, { "cell_type": "code", - "execution_count": 20, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "m_GP = GPy.models.SparseGPRegression(X=X_tr, Y=Y_tr, kernel=GPy.kern.RBF(X_tr.shape[1])+GPy.kern.Bias(X_tr.shape[1]), num_inducing=num_inducing)" @@ -499,41 +290,9 @@ }, { "cell_type": "code", - "execution_count": 21, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Running L-BFGS-B (Scipy implementation) Code:\n", - " runtime i f |g| \n", - " 00s19 006 2.463751e+03 4.986610e+02 \n", - " 00s33 027 2.339956e+03 5.774437e-11 \n", - "Runtime: 00s33\n", - "Optimization status: Converged\n", - "\n", - "Running L-BFGS-B (Scipy implementation) Code:\n", - " runtime i f |g| \n", - " 00s07 004 1.855557e+03 2.678079e+03 \n", - " 00s18 011 1.778417e+03 5.465010e-02 \n", - " 00s32 029 1.778190e+03 1.885596e-09 \n", - "Runtime: 00s32\n", - "Optimization status: Converged\n", - "\n" - ] - }, - { - "data": { - "text/plain": [ - "" - ] - }, - "execution_count": 21, - "metadata": {}, - "output_type": "execute_result" - } - ], + "outputs": [], "source": [ "m_GP.Gaussian_noise.variance = m_GP.Y.var()*0.01\n", "m_GP.Gaussian_noise.variance.fix()\n", @@ -544,10 +303,8 @@ }, { "cell_type": "code", - "execution_count": 22, - "metadata": { - "collapsed": true - }, + "execution_count": null, + "metadata": {}, "outputs": [], "source": [ "def rmse(predictions, targets):\n", @@ -556,19 +313,9 @@ }, { "cell_type": "code", - "execution_count": 23, + "execution_count": null, "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "# RMSE DGP : 3.03000726496\n", - "# RMSE DGP (with samples): 3.03436330576\n", - "# RMSE GP : 3.02347121967\n" - ] - } - ], + "outputs": [], "source": [ "Y_pred = m.predict(X_ts)[0]\n", "Y_pred_s = m.predict_withSamples(X_ts, nSamples=500)[0]\n", @@ -582,9 +329,7 @@ { "cell_type": "code", "execution_count": null, - "metadata": { - "collapsed": true - }, + "metadata": {}, "outputs": [], "source": [] }