Skip to content

Commit

Permalink
normal sapcing
Browse files Browse the repository at this point in the history
  • Loading branch information
Guy committed Jun 3, 2015
1 parent 3fb1dc8 commit 8317a7e
Showing 1 changed file with 144 additions and 144 deletions.
288 changes: 144 additions & 144 deletions __tests__/testRandomOps.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,181 +6,181 @@ var z = 3.29;


function assertProp(observed_p, expected_p, N) {
var se = z * Math.sqrt(expected_p * (1 - expected_p) / N);
expect(Math.abs(observed_p - expected_p) <= se).toBe(true);
var se = z * Math.sqrt(expected_p * (1 - expected_p) / N);
expect(Math.abs(observed_p - expected_p) <= se).toBe(true);
}

function sum(obj) {
return obj.reduce(function(totalSum, cur) {
return totalSum + cur;
}, 0);
return obj.reduce(function(totalSum, cur) {
return totalSum + cur;
}, 0);
}

function valueMassToDensity(value_mass) {
var values = value_mass.map(function(val) { return Object.keys(val)[0]; });
var ns = value_mass.map(function(val) { return val[Object.keys(val)[0]]});
var ns_sum = parseFloat(sum(ns));
ns = ns.map(function(val) {
return val / ns_sum;
});
var ret = {};
for (var i = 0; i < values.length; i++) {
ret[values[i]] = ns[i];
}
return ret;
var values = value_mass.map(function(val) { return Object.keys(val)[0]; });
var ns = value_mass.map(function(val) { return val[Object.keys(val)[0]]});
var ns_sum = parseFloat(sum(ns));
ns = ns.map(function(val) {
return val / ns_sum;
});
var ret = {};
for (var i = 0; i < values.length; i++) {
ret[values[i]] = ns[i];
}
return ret;
}

function Counter(l) {
ret = {}
l.forEach(function(el) {
if (ret[el]) {
ret[el] += 1;
} else {
ret[el] = 1;
}
});
return ret;
ret = {}
l.forEach(function(el) {
if (ret[el]) {
ret[el] += 1;
} else {
ret[el] = 1;
}
});
return ret;
}

function assertProbs(xs, value_density, N) {
var hist = Counter(xs);
Object.keys(hist).forEach(function(el) {
assertProp(hist[el] / N, value_density[el], N);
});
var hist = Counter(xs);
Object.keys(hist).forEach(function(el) {
assertProp(hist[el] / N, value_density[el], N);
});
}

function distributionTester(xs, value_mass, N) {
if (!N) {
N =10000;
}
value_density = valueMassToDensity(value_mass);
if (!N) {
N =10000;
}
value_density = valueMassToDensity(value_mass);

assertProbs(xs, value_density, parseFloat(N));
assertProbs(xs, value_density, parseFloat(N));
}

describe('Test randomization ops', function() {
it('salts correctly', function() {
var i = 20;
var a = new Assignment("assign_salt_a");
it('salts correctly', function() {
var i = 20;
var a = new Assignment("assign_salt_a");

a.set('x', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i}));
a.set('y', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i}));
expect(a.get('x')).not.toEqual(a.get('y'));
a.set('x', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i}));
a.set('y', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i}));
expect(a.get('x')).not.toEqual(a.get('y'));

a.set('z', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'salt': 'x'}));
expect(a.get('x')).toEqual(a.get('z'));
a.set('z', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'salt': 'x'}));
expect(a.get('x')).toEqual(a.get('z'));

var b = new Assignment('assign_salt_b');
var b = new Assignment('assign_salt_b');
b.set('x', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i}));
expect(a.get('x')).not.toEqual(b.get('x'));

a.set('f', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'full_salt':'fs'}));
b.set('f', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'full_salt':'fs'}));
expect(a.get('f')).toEqual(b.get('f'));

a.set('f', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'full_salt':'fs2'}));
a.set('f', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'full_salt':'fs2'}));
b.set('f', new Random.RandomInteger({ 'min': 0, 'max': 100000, 'unit': i, 'full_salt':'fs2'}));
expect(a.get('f')).toEqual(b.get('f'));

});

it('works for bernoulli trials', function() {
var N = 10000;
function bernoulli(p) {
var xs = [];
for (var i = 0; i < N; i++) {
var a = new Assignment(p);
a.set('x', new Random.BernoulliTrial({ 'p': p, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}
distributionTester(bernoulli(0.0), [{0: 1}, {1: 0}], N);
distributionTester(bernoulli(0.1), [{0: 0.9}, {1: 0.1}], N);
distributionTester(bernoulli(1.0), [{0: 0}, {1: 1}], N);
});
it('works for uniform choice', function() {
var N = 10000;
function uniformChoice(choices) {
var xs = [];
for (var i = 0; i < N; i++) {
var a = new Assignment(choices.join(','));
a.set('x', new Random.UniformChoice({ 'choices': choices, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}
distributionTester(uniformChoice(['a']), [{'a': 1}], N);
distributionTester(uniformChoice(['a', 'b']), [{'a': 1}, {'b': 1}], N);
});

it('works for weighted choice', function() {
var N = 10000;
function weightedChoice(choices) {
var xs = [];
var weights = choices.map(function(choice) { return choice[Object.keys(choice)[0]]});
var choices = choices.map(function(choice) { return Object.keys(choice)[0]; });
for (var i = 0; i < N; i++) {
var a = new Assignment(weights.join(', '));
a.set('x', new Random.WeightedChoice({ 'choices': choices, 'weights': weights, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}

var d = [{'a': 1}];
distributionTester(weightedChoice(d), d, N);
d = [{'a': 1}, {'b': 2}];
distributionTester(weightedChoice(d), d, N);
d = [{'a': 0}, {'b': 2}, {'c': 0}];
distributionTester(weightedChoice(d), d, N);

var da = [{'a': 1}, {'b': 2}, {'c': 0}, {'a': 2}];
var db = [{'a': 3}, {'b': 2}, {'c': 0}];
});

it('works for bernoulli trials', function() {
var N = 10000;
function bernoulli(p) {
var xs = [];
for (var i = 0; i < N; i++) {
var a = new Assignment(p);
a.set('x', new Random.BernoulliTrial({ 'p': p, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}
distributionTester(bernoulli(0.0), [{0: 1}, {1: 0}], N);
distributionTester(bernoulli(0.1), [{0: 0.9}, {1: 0.1}], N);
distributionTester(bernoulli(1.0), [{0: 0}, {1: 1}], N);
});
it('works for uniform choice', function() {
var N = 10000;
function uniformChoice(choices) {
var xs = [];
for (var i = 0; i < N; i++) {
var a = new Assignment(choices.join(','));
a.set('x', new Random.UniformChoice({ 'choices': choices, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}
distributionTester(uniformChoice(['a']), [{'a': 1}], N);
distributionTester(uniformChoice(['a', 'b']), [{'a': 1}, {'b': 1}], N);
});

it('works for weighted choice', function() {
var N = 10000;
function weightedChoice(choices) {
var xs = [];
var weights = choices.map(function(choice) { return choice[Object.keys(choice)[0]]});
var choices = choices.map(function(choice) { return Object.keys(choice)[0]; });
for (var i = 0; i < N; i++) {
var a = new Assignment(weights.join(', '));
a.set('x', new Random.WeightedChoice({ 'choices': choices, 'weights': weights, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}

var d = [{'a': 1}];
distributionTester(weightedChoice(d), d, N);
d = [{'a': 1}, {'b': 2}];
distributionTester(weightedChoice(d), d, N);
d = [{'a': 0}, {'b': 2}, {'c': 0}];
distributionTester(weightedChoice(d), d, N);

var da = [{'a': 1}, {'b': 2}, {'c': 0}, {'a': 2}];
var db = [{'a': 3}, {'b': 2}, {'c': 0}];
distributionTester(weightedChoice(da), db, N);
});

it('works for sample', function() {
var N = 100;
function sample(choices, draws) {
var xs = [];
for (var i = 0; i < N; i++) {
var a = new Assignment(choices.join(', '));
a.set('x', new Random.Sample({ 'choices': choices, 'draws': draws, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}

function listDistributionTester(xs_list, value_mass, N) {
var value_density = valueMassToDensity(value_mass);
var l = [];

/* bad equivalent to zip() from python */
xs_list.forEach(function(xs, i){
xs.forEach(function(x, j) {
if (!l[j]) {
l[j] = [x];
} else {
l[j].push(x);
}
});
if (i === xs_list.length-1) {
l.forEach(function(el) {
assertProbs(el, value_density, N);
});
}
});
}

var a = [1, 2, 3];
var ret = [{1: 1}, {2: 1}, {3: 1}];
listDistributionTester(sample(a, 3), ret, N);
listDistributionTester(sample(a, 2), ret, N);
a = ['a', 'a', 'b'];
ret = [{'a': 2}, {'b': 1}];
listDistributionTester(sample(a, 3), ret, N);
});
});

it('works for sample', function() {
var N = 100;
function sample(choices, draws) {
var xs = [];
for (var i = 0; i < N; i++) {
var a = new Assignment(choices.join(', '));
a.set('x', new Random.Sample({ 'choices': choices, 'draws': draws, 'unit': i }));
xs[i] = a.get('x');
}
return xs;
}

function listDistributionTester(xs_list, value_mass, N) {
var value_density = valueMassToDensity(value_mass);
var l = [];

/* bad equivalent to zip() from python */
xs_list.forEach(function(xs, i){
xs.forEach(function(x, j) {
if (!l[j]) {
l[j] = [x];
} else {
l[j].push(x);
}
});
if (i === xs_list.length-1) {
l.forEach(function(el) {
assertProbs(el, value_density, N);
});
}
});
}

var a = [1, 2, 3];
var ret = [{1: 1}, {2: 1}, {3: 1}];
listDistributionTester(sample(a, 3), ret, N);
listDistributionTester(sample(a, 2), ret, N);
a = ['a', 'a', 'b'];
ret = [{'a': 2}, {'b': 1}];
listDistributionTester(sample(a, 3), ret, N);
});
});

0 comments on commit 8317a7e

Please sign in to comment.