Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
{
"metadata": {
"name": "",
"signature": "sha256:0c550503013c5dffd9e999dc3324315c8672ffb9bcda1ffc453aaa1c1a17c078"
},
"nbformat": 3,
"nbformat_minor": 0,
"worksheets": [
{
"cells": [
{
"cell_type": "code",
"collapsed": false,
"input": [
"from __future__ import division\n",
"%matplotlib inline\n",
"import numpy as np\n",
"import pylab as pl"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 1
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"from sklearn.metrics.pairwise import chi2_kernel\n",
"from sklearn.svm import NuSVR\n",
"from sklearn.cross_validation import KFold, cross_val_score"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 2
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"from sklearn.externals.joblib import Parallel, delayed"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 3
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# make some fake histogram data with easy labels\n",
"n_train = 500\n",
"y_train = np.random.uniform(0.1, 10, size=n_train)\n",
"X_train = np.random.poisson(lam=y_train.reshape(-1, 1), size=(n_train, 50))\n",
"\n",
"n_test = 50\n",
"y_test = np.random.uniform(0.1, 10, size=n_test)\n",
"X_test = np.random.poisson(lam=y_test.reshape(-1, 1), size=(n_test, 50))"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 4
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# do parameter tuning. unfortunately, have to do it\n",
"# pretty manually here, because the sklearn tuning\n",
"# system is not all that powerful.\n",
"\n",
"# choose the parameter values we'll try\n",
"gammas = 10. ** np.arange(-5, 3) # not entirely sure what a good range is\n",
"Cs = 10. ** np.arange(-3, 7)\n",
"nus = np.array([.1, .2, .3, .5, .7])\n",
"n_tune_folds = 3 # how many folds to tune with\n",
"\n",
"# big array of all the scores we get out of tuning\n",
"tuning_scores = np.empty((gammas.size, Cs.size, nus.size, n_tune_folds))\n",
"tuning_scores.fill(np.nan)\n",
"\n",
"# precompute the kernel matrices for all the gamma values\n",
"kernels = [chi2_kernel(X_train, gamma=g) for g in gammas]\n",
"\n",
"# function to try out a paremeter setting; so we can parallelize\n",
"def do_tune(gamma_i, C_i, nu_i): # uses nus, Cs, kernels, y_train from the environment\n",
" model = NuSVR(nu=nus[nu_i], C=Cs[C_i])\n",
" tuning_scores[gamma_i, C_i, nu_i, :] = cross_val_score(model, kernels[gamma_i], y_train)\n",
"\n",
"# run do_tune for each of the gamma, C, nu values in parallel\n",
"Parallel(n_jobs=-1, backend=\"threading\", verbose=True)(\n",
" delayed(do_tune)(gamma_i, C_i, nu_i)\n",
" for gamma_i in range(gammas.size)\n",
" for C_i in range(Cs.size)\n",
" for nu_i in range(nus.size))\n",
"\n",
"# this'll take a while: 4 minutes using all of my (beefy-ish) laptop. sorry. :|\n",
"None"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "stream",
"stream": "stderr",
"text": [
"[Parallel(n_jobs=-1)]: Done 1 out of 111 | elapsed: 0.1s remaining: 15.3s\n",
"[Parallel(n_jobs=-1)]: Done 400 out of 400 | elapsed: 4.4min finished\n"
]
}
],
"prompt_number": 5
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# check out what the tuning scores look like\n",
"# these scores are R^2: 1 means perfect\n",
"means = np.mean(tuning_scores, axis=-1)\n",
"\n",
"for nu_i, nu in enumerate(nus):\n",
" pl.matshow(means[:, :, nu_i])\n",
" \n",
" pl.yticks(np.arange(gammas.size), gammas)\n",
" pl.ylabel(r'$\\gamma$');\n",
" \n",
" pl.gca().xaxis.tick_bottom()\n",
" pl.xticks(np.arange(Cs.size), Cs, rotation=45)\n",
" pl.xlabel('C')\n",
" #; pl.gca().xaxis.set_label_position('top')\n",
" \n",
" pl.colorbar()\n",
" pl.title(r'$\\nu$ = {}'.format(nu))"
],
"language": "python",
"metadata": {},
"outputs": [
{
"metadata": {},
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAEXCAYAAAAtEMZoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
"text": [
"<matplotlib.figure.Figure at 0x10c9a7a90>"
]
},
{
"metadata": {},
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAEXCAYAAAAtEMZoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+D9OXAF9siOpvwAAAABJRU5ErkJggg==\n",
"text": [
"<matplotlib.figure.Figure at 0x10f340ed0>"
]
},
{
"metadata": {},
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAEXCAYAAAAtEMZoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nbIb2hQAAAABJRU5ErkJggg==\n",
"text": [
"<matplotlib.figure.Figure at 0x10f48c950>"
]
},
{
"metadata": {},
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAEXCAYAAAAtEMZoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nsxL+P9fKB0HLhUkfAAAAAElFTkSuQmCC\n",
"text": [
"<matplotlib.figure.Figure at 0x10f6f3b50>"
]
},
{
"metadata": {},
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAEXCAYAAAAtEMZoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
"text": [
"<matplotlib.figure.Figure at 0x10f83c750>"
]
}
],
"prompt_number": 6
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"So it looks like, in this simple case, $\\nu$ doesn't really matter, $C$ should be highish, $\\gamma$ should be smallish."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"At this point, you can pick one set of parameters:\n",
"* by eye\n",
"* by picking the max value of `means`\n",
"* by fitting a Gaussian Process to the values and choosing its peak"
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"gamma_i, C_i, nu_i = np.unravel_index(means.argmax(), means.shape)\n",
"gamma = gammas[gamma_i]\n",
"C = Cs[C_i]\n",
"nu = nus[nu_i]"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 11
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"gamma, C, nu"
],
"language": "python",
"metadata": {},
"outputs": [
{
"metadata": {},
"output_type": "pyout",
"prompt_number": 12,
"text": [
"(1.0000000000000001e-05, 1000000.0, 0.20000000000000001)"
]
}
],
"prompt_number": 12
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Now, let's do the final prediction on test data:"
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"model = NuSVR(nu=nu, C=C, kernel='precomputed')\n",
"model.fit(kernels[gamma_i], y_train)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"metadata": {},
"output_type": "pyout",
"prompt_number": 14,
"text": [
"NuSVR(C=1000000.0, cache_size=200, coef0=0.0, degree=3, gamma=0.0,\n",
" kernel='precomputed', max_iter=-1, nu=0.20000000000000001,\n",
" probability=False, random_state=None, shrinking=True, tol=0.001,\n",
" verbose=False)"
]
}
],
"prompt_number": 14
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"y_preds = model.predict(chi2_kernel(X_test, X_train, gamma=gamma))"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 15
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"from sklearn import metrics"
],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 16
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"metrics.r2_score(y_test, y_preds)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"metadata": {},
"output_type": "pyout",
"prompt_number": 19,
"text": [
"0.9565823934452764"
]
}
],
"prompt_number": 19
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"metrics.mean_squared_error(y_test, y_preds)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"metadata": {},
"output_type": "pyout",
"prompt_number": 20,
"text": [
"0.32837950870236776"
]
}
],
"prompt_number": 20
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Hey, we learned it pretty well! (Of course the problem was stupid easy, but still.)"
]
}
],
"metadata": {}
}
]
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.