-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
7 changed files
with
450 additions
and
19 deletions.
There are no files selected for viewing
Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.
Oops, something went wrong.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,277 @@ | ||
| // | ||
| // MIT License | ||
| // | ||
| // Copyright (c) 2019 pengxu | ||
| // | ||
| // Permission is hereby granted, free of charge, to any person obtaining a copy | ||
| // of this software and associated documentation files (the "Software"), to deal | ||
| // in the Software without restriction, including without limitation the rights | ||
| // to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | ||
| // copies of the Software, and to permit persons to whom the Software is | ||
| // furnished to do so, subject to the following conditions: | ||
| // | ||
| // The above copyright notice and this permission notice shall be included in all | ||
| // copies or substantial portions of the Software. | ||
| // | ||
| // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | ||
| // IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | ||
| // FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | ||
| // AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | ||
| // LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | ||
| // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE | ||
| // SOFTWARE. | ||
| // | ||
|
|
||
| package fi.helsinki.cs.udbms | ||
|
|
||
| import com.xenomachina.argparser.mainBody | ||
| import fi.helsinki.cs.udbms.struct.* | ||
| import fi.helsinki.cs.udbms.util.* | ||
| import java.util.concurrent.ThreadLocalRandom | ||
| import kotlin.math.pow | ||
| import kotlin.system.measureTimeMillis | ||
|
|
||
| fun main(args: Array<String>): Unit = mainBody { | ||
| val params = EstimationParameters.initialise(args) | ||
| Dispatcher.initialise(params.singleThread) | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| print("Reading string... ") | ||
| val list1 = IO.readSegmentedStrings(params.list1) | ||
| val list2 = IO.readSegmentedStrings(params.list2) | ||
| println("${list1.size} + ${list2.size} strings loaded") | ||
|
|
||
| var syn: SynonymKnowledge? = null | ||
| if (params.synonym.isNotEmpty()) { | ||
| print("Reading synonym... ") | ||
| syn = IO.readSynonym(params.synonym) | ||
| println("${syn.knowledge.size} rules loaded") | ||
| } | ||
| var tax: TaxonomyKnowledge? = null | ||
| if (params.taxonomy.isNotEmpty()) { | ||
| print("Reading taxonomy... ") | ||
| tax = IO.readTaxonomy(params.taxonomy) | ||
| println("${tax.knowledge.size} nodes loaded") | ||
| } | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| val pebbles1 = list1.mapParallel { Pair(it, PebbleGenerator(syn, tax, params.gram).generate(it)) }.toMap() | ||
| val pebbles2 = list2.mapParallel { Pair(it, PebbleGenerator(syn, tax, params.gram).generate(it)) }.toMap() | ||
|
|
||
| val order = GlobalOrder() | ||
| order.addAll(pebbles1.values.flatten()) | ||
| order.addAll(pebbles2.values.flatten()) | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| print("Running test drive... ") | ||
| var verified = false | ||
| val (filterTime, verifyTime) = params.overlapList.map { | ||
| val result = testDrive( | ||
| params, list1.shuffled().take(2000), list2.shuffled().take(2000), | ||
| pebbles1, pebbles2, order, syn, tax, it, verified | ||
| ) | ||
| verified = true | ||
| return@map Pair(result.filterTime, result.verifyTime) | ||
| }.reduceIndexed { i, acc, now -> | ||
| Pair( | ||
| acc.first + (now.first - acc.first) / (i + 1), | ||
| if (now.second == 0.0) acc.second else acc.second + (acc.second - acc.second) / (i + 1) | ||
| ) | ||
| } | ||
| println("filtering time ${filterTime.format(4)} ms/pair, verification time ${verifyTime.format(4)} ms/candidate") | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| print("\t") | ||
| print("i\t") | ||
| params.overlapList.forEach { print("mean$it\t") } | ||
| params.overlapList.forEach { print("error$it\t") } | ||
| println() | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| val p1 = params.sampleSize.toDouble() / list1.size | ||
| val p2 = params.sampleSize.toDouble() / list2.size | ||
|
|
||
| var lastEstimations = emptyMap<Int, Estimation>() | ||
|
|
||
| var iteration = 0 | ||
| while (++iteration <= params.iteration) { | ||
| val sample1 = getBernoulliSample(list1, p1) | ||
| val sample2 = getBernoulliSample(list2, p2) | ||
|
|
||
| val estimations = params.overlapList.map { overlap -> | ||
| val result = testDrive(params, sample1, sample2, pebbles1, pebbles2, order, syn, tax, overlap, true) | ||
|
|
||
| val scaledCost = | ||
| result.numberOfPairs / (p1 * p2) * filterTime + result.numberOfCandidates / (p1 * p2) * verifyTime | ||
| val mean = movingMean(scaledCost, iteration, lastEstimations[overlap]) | ||
| val variance = movingVariance(scaledCost, iteration, lastEstimations[overlap]) | ||
| val error = params.quantile * (variance / iteration).pow(0.5) | ||
|
|
||
| return@map Estimation( | ||
| iteration, overlap, result.numberOfPairs, result.numberOfCandidates, | ||
| scaledCost, mean, variance, error | ||
| ) | ||
| }.associateBy { it.overlap } | ||
|
|
||
| print("\t") | ||
| print("$iteration\t") | ||
| estimations.forEach { print("${it.value.meanOfScaledCost.format(0)}\t") } | ||
| estimations.forEach { print("${it.value.errorOfScaledCost.format(0)}\t") } | ||
| println() | ||
|
|
||
| lastEstimations = estimations | ||
|
|
||
| // should we stop here? | ||
| if (shouldStop(iteration, estimations)) | ||
| break | ||
| } | ||
|
|
||
| print("Overlap parameters from the best to the worst: ") | ||
| print(lastEstimations.values.sortedBy { it.scaledCost }.map { it.overlap }.joinToString()) | ||
| println() | ||
|
|
||
| Dispatcher.shutdown() | ||
| return@mainBody | ||
| } | ||
|
|
||
| private fun shouldStop(iteration: Int, estimations: Map<Int, Estimation>): Boolean { | ||
| if (estimations.size == 1) return true | ||
| if (iteration < 5) return false | ||
|
|
||
| val estimationsSorted = estimations.values.sortedBy { it.scaledCost } | ||
|
|
||
| return estimationsSorted.first().getMaxScaledCost() < estimationsSorted.drop(1).first().getMinScaledCost() | ||
| } | ||
|
|
||
| private fun <T> getBernoulliSample(data: Iterable<T>, p: Double): List<T> { | ||
| val rand = ThreadLocalRandom.current() | ||
| @Suppress("UNCHECKED_CAST") | ||
| return data.mapParallel { if (rand.nextDouble(1.0) < p) it else Unit }.filterNot { it == Unit } as List<T> | ||
| } | ||
|
|
||
| // region Estimation | ||
|
|
||
| private fun movingMean(estimate: Double, iteration: Int, last: Estimation?): Double { | ||
| val lastSafe = last ?: Estimation() | ||
|
|
||
| if (iteration == 1) | ||
| return estimate | ||
|
|
||
| return lastSafe.meanOfScaledCost + (estimate - lastSafe.meanOfScaledCost) / iteration | ||
| } | ||
|
|
||
| private fun movingVariance(estimate: Double, iteration: Int, last: Estimation?): Double { | ||
| val lastSafe = last ?: Estimation() | ||
|
|
||
| if (iteration == 1) | ||
| return 0.0 | ||
|
|
||
| return (iteration - 2).toDouble() / (iteration - 1).toDouble() * lastSafe.varianceOfScaledCost + | ||
| ((estimate - lastSafe.meanOfScaledCost).pow(2) / iteration) | ||
| } | ||
|
|
||
| private data class Estimation( | ||
| val iteration: Int = 0, | ||
| val overlap: Int = 0, | ||
| val numberOfPairs: Int = 0, | ||
| val numberOfCandidates: Int = 0, | ||
| val scaledCost: Double = 0.0, | ||
| val meanOfScaledCost: Double = 0.0, | ||
| val varianceOfScaledCost: Double = 0.0, | ||
| val errorOfScaledCost: Double = 0.0 | ||
| ) { | ||
| fun getMinScaledCost() = scaledCost - errorOfScaledCost | ||
| fun getMaxScaledCost() = scaledCost + errorOfScaledCost | ||
| } | ||
|
|
||
| // endregion | ||
|
|
||
| // region TestDrive | ||
|
|
||
| private fun testDrive( | ||
| params: EstimationParameters, | ||
| list1: List<SegmentedString>, | ||
| list2: List<SegmentedString>, | ||
| pebbles1: Map<SegmentedString, List<Pebble>>, | ||
| pebbles2: Map<SegmentedString, List<Pebble>>, | ||
| order: GlobalOrder, | ||
| syn: SynonymKnowledge?, | ||
| tax: TaxonomyKnowledge?, | ||
| overlap: Int, | ||
| skipVerify: Boolean | ||
| ): TestDriveResult { | ||
| System.gc(); System.runFinalization() | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| var signatures1: Map<SegmentedString, List<Pebble>> = emptyMap() | ||
| var signatures2: Map<SegmentedString, List<Pebble>> = emptyMap() | ||
|
|
||
| run { | ||
| val reducer = when (params.filter) { | ||
| "Fast" -> FastPebbleReducer(params.threshold, overlap, order) | ||
| "DP" -> DynamicProgrammingPebbleReducer(params.threshold, overlap, order) | ||
| else -> throw Exception("Invalid filtering method: ${params.filter}") | ||
| } | ||
|
|
||
| signatures1 = list1.mapParallel { Pair(it, reducer.reduce(it, pebbles1[it] ?: emptyList())) }.toMap() | ||
| signatures2 = list2.mapParallel { Pair(it, reducer.reduce(it, pebbles2[it] ?: emptyList())) }.toMap() | ||
| }.run { System.gc(); System.runFinalization(); } | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| val index2 = InvertedIndex() | ||
| signatures2.map { str -> str.value.map { p -> index2.add(p, p.segment) } } | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| var candidates: List<SegmentedStringPair> = emptyList() | ||
| var numberOfPairs = 0 | ||
| val filterTimeTotal = measureTimeMillis { | ||
| val result = AdaptivePrefixFilter(params.threshold, overlap).getCandidates(signatures1, index2) | ||
| candidates = result.first | ||
| numberOfPairs = result.second | ||
| } | ||
|
|
||
| /*=================================================================*/ | ||
|
|
||
| var verifyTimeTotal = 0L | ||
| if (!skipVerify) { | ||
| val verifier = when (params.verify) { | ||
| "Greedy" -> GreedySimilarityVerifier(params.threshold, syn, tax, params.gram) | ||
| "SquareImp" -> SquareImpSimilarityVerifier(params.threshold, syn, tax, params.gram) | ||
| "SquareImp-Improved" -> SquareImpSimilarityVerifier(params.threshold, syn, tax, params.gram, true) | ||
| else -> throw Exception("Invalid verification method: ${params.verify}") | ||
| } | ||
|
|
||
| verifyTimeTotal = measureTimeMillis { | ||
| candidates.mapParallelOrSequential { Pair(it, verifier.getSimilarity(it.first, it.second)) } | ||
| .filter { it.second.start >= params.threshold } | ||
| .toList() | ||
| } | ||
| } | ||
| /*=================================================================*/ | ||
|
|
||
| System.gc(); System.runFinalization() | ||
|
|
||
| return TestDriveResult( | ||
| filterTimeTotal.toDouble() / numberOfPairs, | ||
| verifyTimeTotal.toDouble() / candidates.size, | ||
| numberOfPairs, | ||
| candidates.size | ||
| ) | ||
| } | ||
|
|
||
| private data class TestDriveResult( | ||
| val filterTime: Double, | ||
| val verifyTime: Double, | ||
| val numberOfPairs: Int, | ||
| val numberOfCandidates: Int | ||
| ) | ||
|
|
||
| // endregion |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.