Merge pull request #91 from aimalz/issue/86/desc_revision

Issue/86/desc revision, closes #86

docs/desc-0000-qp-photo-z_approximation/acknowledgments.tex

@@ -1,3 +1,3 @@
-% 
-This is the text imported from \code{acknowledgments.tex}, and will be replaced by some standard LSST DESC boilerplate at some point.
-% 
+%
+%This is the text imported from \code{acknowledgments.tex}, and will be replaced by some standard LSST DESC boilerplate at some point.
+%

docs/desc-0000-qp-photo-z_approximation/authors.csv

@@ -31,7 +31,7 @@
 Lastname,Firstname,Authorname,AuthorType,Affiliation,Contribution,Email
 Malz,Alex,A.I.~Malz,Contact,"Center for Cosmology and Particle Physics, New York University, 726 Broadway, New York, 10003","Initiated project, led development work.",[email protected]
 Marshall,Phil,P.J.~Marshall,Contributor,"SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA","Advised on statistics, and project design and management.",[email protected]
-Schmidt,Samuel,S.J.~Schmidt,Contributor,"Dept. of Physics, University of California, One Shields Ave., Davis, CA, 95616","Provided the lower quality data mock catalog.",[email protected]
-Graham,Melissa,M.~Graham,Contributor,"University of Washington","Provided the higher quality data mock catalog.",[email protected]
-DeRose,Joe,J.~DeRose,Contributor,"Stanford University","Contributed to the production of the lower quality data mock catalog.",[email protected]
-Wechsler,Risa,R.~Wechsler,Contributor,"Stanford University","Contributed to the production of the lower quality data mock catalog.",[email protected]
+Schmidt,Samuel,S.J.~Schmidt,Contributor,"Dept. of Physics, University of California, One Shields Ave., Davis, CA, 95616","Produced the PDFs for the fainter mock catalog.",[email protected]
+Graham,Melissa,M.L.~Graham,Contributor,"Department of Astronomy, University of Washington, Box 351580 U.W., Seattle, WA 98195-1580","Produced the photometry and PDFs for the brighter mock catalog.",[email protected]
+DeRose,Joe,J.~DeRose,Contributor,"Stanford University","Produced the photometry for the fainter mock catalog.",[email protected]
+Wechsler,Risa,R.~Wechsler,Contributor,"Stanford University","Produced the photometry for the fainter mock catalog.",[email protected]

docs/desc-0000-qp-photo-z_approximation/authors.tex

@@ -7,8 +7,8 @@
 \affiliation{SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA}
 \author{S.J.~Schmidt}
 \affiliation{Dept. of Physics, University of California, One Shields Ave., Davis, CA, 95616}
-\author{M.~Graham}
-\affiliation{University of Washington}
+\author{M.L.~Graham}
+\affiliation{Department of Astronomy, University of Washington, Box 351580 U.W., Seattle, WA 98195-1580}
 \author{J.~DeRose}
 \affiliation{Stanford University}
 \author{R.~Wechsler}

docs/desc-0000-qp-photo-z_approximation/contributions.tex

@@ -1,7 +1,7 @@
 Author contributions are listed below. \\
 A.I.~Malz: Initiated project, led development work. \\
 P.J.~Marshall: Advised on statistics, and project design and management. \\
-S.J.~Schmidt: Provided the Optical dataset. \\
-M.~Graham: Provided the Optical+IR dataset. \\
-J.~DeRose: Contributed to the production of the Optical dataset. \\
-R.~Wechsler: Contributed to the production of the Optical dataset. \\
+S.J.~Schmidt: Produced the PDFs for the fainter mock catalog. \\
+M.L.~Graham: Produced the photometry and PDFs for the brighter mock catalog. \\
+J.~DeRose: Produced the photometry for the fainter mock catalog. \\
+R.~Wechsler: Produced the photometry for the fainter mock catalog. \\

docs/desc-0000-qp-photo-z_approximation/research/data_exploration.ipynb

@@ -85,8 +85,8 @@
    "outputs": [],
    "source": [
     "# choose one of these:\n",
-    "dataset_key = 'Euclid'# Melissa Graham's data\n",
-    "# dataset_key = 'LSST'# Sam Schmidt's data\n",
+    "# dataset_key = 'Euclid'# Melissa Graham's data\n",
+    "dataset_key = 'LSST'# Sam Schmidt's data\n",
     "dataname = dataset_key\n",
     "\n",
     "dataset_info = {}\n",
@@ -192,24 +192,24 @@
    },
    "outputs": [],
    "source": [
-    "colors = ['red','green','blue','cyan','magenta','yellow']\n",
-    "n_plot = len(colors)\n",
-    "\n",
-    "# if dataset_key == 'mg':\n",
-    "#     indices = [1, 3, 14, 16, 19, 21]\n",
-    "# elif dataset_key == 'ss':\n",
-    "n_gals_tot = len(pdfs)\n",
-    "full_gal_range = range(n_gals_tot)\n",
-    "indices = np.random.choice(full_gal_range, n_plot)\n",
-    "\n",
-    "for i in range(n_plot):\n",
-    "    plt.plot(dataset_info[dataset_key]['z_grid'], pdfs[indices[i]], \n",
-    "             color=colors[i], label=dataset_key+'#'+str(indices[i]))\n",
-    "plt.xlabel(r'$z$', fontsize=16)\n",
-    "plt.ylabel(r'$p(z)$', fontsize=16)\n",
-    "plt.title(dataset_key+' mock catalog')\n",
-    "plt.legend()\n",
-    "plt.savefig('pz_placeholder_'+dataset_key+'.png', dpi=250)"
+    "# colors = ['red','green','blue','cyan','magenta','yellow']\n",
+    "# n_plot = len(colors)\n",
+    "\n",
+    "# # if dataset_key == 'mg':\n",
+    "# #     indices = [1, 3, 14, 16, 19, 21]\n",
+    "# # elif dataset_key == 'ss':\n",
+    "# n_gals_tot = len(pdfs)\n",
+    "# full_gal_range = range(n_gals_tot)\n",
+    "# indices = np.random.choice(full_gal_range, n_plot)\n",
+    "\n",
+    "# for i in range(n_plot):\n",
+    "#     plt.plot(dataset_info[dataset_key]['z_grid'], pdfs[indices[i]], \n",
+    "#              color=colors[i], label=dataset_key+' #'+str(indices[i]))\n",
+    "# plt.xlabel(r'$z$', fontsize=16)\n",
+    "# plt.ylabel(r'$p(z)$', fontsize=16)\n",
+    "# plt.title(dataset_key+' mock catalog')\n",
+    "# plt.legend()\n",
+    "# plt.savefig('pz_placeholder_'+dataset_key+'.pdf', dpi=250)"
    ]
   },
   {
@@ -246,8 +246,10 @@
     "if dataset_key == 'Euclid':\n",
     "    chosen = 5390\n",
     "elif dataset_key == 'LSST':\n",
-    "    chosen = 108019\n",
-    "\n",
+    "#     chosen = 108019 \n",
+    "    indices = [ 12543,  52661,  46216,  53296,  95524,  84574 ,  2607  ,56017 , 64794, 7600]\n",
+    "    chosen = indices[9]\n",
+    "    \n",
     "start_time = timeit.default_timer()\n",
     "G = qp.PDF(gridded=(dataset_info[dataset_key]['z_grid'], pdfs[chosen]))\n",
     "print(timeit.default_timer() - start_time)\n",
@@ -265,13 +267,13 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": true
+    "collapsed": false
    },
    "outputs": [],
    "source": [
     "if dataset_key == 'Euclid':\n",
     "    nc_needed = 3\n",
-    "elif datanset_key == 'LSST':\n",
+    "elif dataset_key == 'LSST':\n",
     "    nc_needed = 5\n",
     "    \n",
     "dataset_info[dataset_key]['N_GMM'] = nc_needed"
@@ -312,7 +314,8 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": false
+    "collapsed": false,
+    "scrolled": true
    },
    "outputs": [],
    "source": [
@@ -332,24 +335,24 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "The `qp.composite` object can be used as the `qp.PDF.truth` to initialize a new `qp.PDF` object that doesn't have any information about the gridded or sample approximations but has a qualitatively similar shape and is thus \"realistically complex\" enough to draw conclusions about real data.  Now we can approximate it any way we like!  Consider this example for $N_f=20$ parameters."
+    "The `qp.composite` object can be used as the `qp.PDF.truth` to initialize a new `qp.PDF` object that doesn't have any information about the gridded or sample approximations but has a qualitatively similar shape and is thus \"realistically complex\" enough to draw conclusions about real data.  Now we can approximate it any way we like!  Consider this example for $N_f=7$ parameters."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
     "collapsed": false,
-    "scrolled": false
+    "scrolled": true
    },
    "outputs": [],
    "source": [
-    "N_f = 20\n",
+    "N_f = 7\n",
     "M = qp.PDF(truth=G.mix_mod, limits=dataset_info[dataset_key]['z_lim'])\n",
     "M.quantize(N=N_f, vb=False)\n",
     "M.histogramize(N=N_f, binrange=dataset_info[dataset_key]['z_lim'], vb=False)\n",
     "M.sample(N=N_f, using='truth', vb=False)\n",
-    "M.plot(loc=dataset_key+'_example_pz.png', vb=True)"
+    "M.plot(loc=dataset_key+'_example_pz.pdf', vb=True)"
    ]
   },
   {
@@ -407,7 +410,7 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": true
+    "collapsed": false
    },
    "outputs": [],
    "source": [
@@ -424,7 +427,7 @@
     "\n",
     "if dataset_key == 'Euclid':\n",
     "    dataset_info[dataset_key]['N_GMM'] = 3\n",
-    "elif datanset_key == 'LSST':\n",
+    "elif dataset_key == 'LSST':\n",
     "    dataset_info[dataset_key]['N_GMM'] = 5\n",
     "fit_components = dataset_info[dataset_key]['N_GMM']\n",
     "\n",
@@ -485,7 +488,7 @@
     "    N_pdfs =len(pdfs)\n",
     "    randos = np.random.choice(range(N_pdfs), n_plot)\n",
     "    for i in range(n_plot):\n",
-    "        plt.plot(z_grid, pdfs[randos[i]], label=dataset_key+'#'+str(randos[i]))\n",
+    "        plt.plot(z_grid, pdfs[randos[i]], label=dataset_key+r'\\#'+str(randos[i]))\n",
     "    plt.xlabel(r'$z$', fontsize=16)\n",
     "    plt.ylabel(r'$p(z)$', fontsize=16)\n",
     "    plt.title(dataset_key+' mock catalog')\n",
@@ -1657,14 +1660,14 @@
   "language_info": {
    "codemirror_mode": {
     "name": "ipython",
-    "version": 2
+    "version": 3
    },
    "file_extension": ".py",
    "mimetype": "text/x-python",
    "name": "python",
    "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython2",
-   "version": "2.7.13"
+   "pygments_lexer": "ipython3",
+   "version": "3.6.3"
   }
  },
  "nbformat": 4,

docs/notebooks/demo.ipynb

@@ -170,7 +170,7 @@
    },
    "outputs": [],
    "source": [
-    "print P.approximate([0.314], using='quantiles')"
+    "print P.approximate(np.array([0.314]), using='quantiles')"
    ]
   },
   {
@@ -227,7 +227,7 @@
    "outputs": [],
    "source": [
     "print P.scheme\n",
-    "print P.approximate([0.314], using='quantiles', scheme='nearest')\n",
+    "print P.approximate(np.array([0.314]), using='quantiles', scheme='nearest')\n",
     "print P.scheme"
    ]
   },
@@ -373,23 +373,29 @@
    },
    "outputs": [],
    "source": [
-    "qD1 = qp.utils.calculate_kl_divergence(P, Q, limits=(-5.,5.), vb=True)\n",
+    "qD1 = qp.utils.calculate_kl_divergence(P, Q, limits=(-1.,1.), vb=True)\n",
     "qD2 = qp.utils.calculate_kl_divergence(P, Q, limits=(-2.,2.), vb=True)\n",
     "qD3 = qp.utils.calculate_kl_divergence(P, Q, limits=(-3.,3.), vb=True)\n",
-    "\n",
-    "hD1 = qp.utils.calculate_kl_divergence(P, H, limits=(-5.,5.), vb=True)\n",
+    "print 'Quantile approximation: KLD over 1,2,3 sigma ranges = ', qD1, qD2, qD3\n",
+    "hD1 = qp.utils.calculate_kl_divergence(P, H, limits=(-1.,1.), vb=True)\n",
     "hD2 = qp.utils.calculate_kl_divergence(P, H, limits=(-2.,2.), vb=True)\n",
     "hD3 = qp.utils.calculate_kl_divergence(P, H, limits=(-3.,3.), vb=True)\n",
-    "\n",
-    "sD1 = qp.utils.calculate_kl_divergence(P, S, limits=(-5.,5.), vb=True)\n",
+    "print 'Histogram approximation: KLD over 1,2,3 sigma ranges = ', hD1, hD2, hD3\n",
+    "sD1 = qp.utils.calculate_kl_divergence(P, S, limits=(-1.,1.), vb=True)\n",
     "sD2 = qp.utils.calculate_kl_divergence(P, S, limits=(-2.,2.), vb=True)\n",
     "sD3 = qp.utils.calculate_kl_divergence(P, S, limits=(-3.,3.), vb=True)\n",
-    "\n",
-    "print 'Quantile approximation: KLD over 1,2,3 sigma ranges = ', qD1, qD2, qD3\n",
-    "print 'Histogram approximation: KLD over 1,2,3 sigma ranges = ', hD1, hD2, hD3\n",
     "print 'Sampled approximation: KLD over 1,2,3 sigma ranges = ', sD1, sD2, sD3"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "collapsed": true
+   },
+   "source": [
+    "Holy smokes, does the quantile approximation blow everything else out of the water, thanks to using spline interpolation."
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -563,7 +569,7 @@
    "outputs": [],
    "source": [
     "Cs = qp.PDF(truth=C_dist, limits = composite_lims)\n",
-    "Cs.sample(N=20, vb=False)\n",
+    "Cs.sample(N=20, using='truth', vb=False)\n",
     "Cs.plot()"
    ]
   },
@@ -609,14 +615,14 @@
   "language_info": {
    "codemirror_mode": {
     "name": "ipython",
-    "version": 2
+    "version": 3
    },
    "file_extension": ".py",
    "mimetype": "text/x-python",
    "name": "python",
    "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython2",
-   "version": "2.7.13"
+   "pygments_lexer": "ipython3",
+   "version": "3.6.3"
   }
  },
  "nbformat": 4,