info.debatty.java.graphs.build.PartitioningGraphBuilder Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of java-graphs Show documentation
Algorithms that build k-nearest neighbors graph (k-nn graph): Brute-force, NN-Descent,...
There is a newer version: 0.41
/*
 * The MIT License
 *
 * Copyright 2015 Thibault Debatty.
 *
 * 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 info.debatty.java.graphs.build;

import info.debatty.java.graphs.Graph;
import info.debatty.java.graphs.NeighborList;
import info.debatty.java.graphs.Node;
import java.util.HashMap;
import java.util.List;
import java.util.Map.Entry;

/**
 * Abstract class for graph building algorithms that split the dataset into 
 * partitions (for example using LSH).
 * 
 * The number of stages (n_stages) and the number of partitions (n_partitions)
 * allow you to control the speedup and precision.
 * 
 * If brute-force method is used inside the partitions:
 * number of similarities to compute ≃ n² / 2 x n_stages / n_partitions
 * where n is the size of the dataset
 * 
 * Thus speedup with respect to pure brute force ≃ n_partitions / n_stages
 * 
 * At the same time:
 * - increasing n_stages will increase the precision
 * - increasing n_partitions will decrease the precision
 * The exact relation between precision and these 2 parameters depends on 
 * the algorithm used...
 * 
 * @param  The type of nodes value
 */
public abstract class PartitioningGraphBuilder extends GraphBuilder {

    protected int oversampling = 2;
    protected int n_partitions = 4;
    protected GraphBuilder internal_builder = new Brute();

    public int getOversampling() {
        return oversampling;
    }

    /**
     * Default = 2
     * @param oversampling 
     */
    public void setOversampling(int oversampling) {
        this.oversampling = oversampling;
    }

    /**
     * Default = 4
     * @return 
     */
    public int getNPartitions() {
        return n_partitions;
    }

    /**
     * Set the number of partitions to build for each stage.
     * Attention: the number of strings per partition should be at least 100 to get 
     * relevant results!
     * Default = 4
     * @param n_partitions 
     */
    public void setNPartitions(int n_partitions) {
        this.n_partitions = n_partitions;
    }

    public GraphBuilder getInternalBuilder() {
        return internal_builder;
    }

    /**
     * Default = Brute force
     * @param internal_builder 
     */
    public void setInternalBuilder(GraphBuilder internal_builder) {
        this.internal_builder = internal_builder;
    }
    
    @Override
    protected Graph _computeGraph(List> nodes) {
        // Create $n_stages$ x $n_partitions$ partitions
        List>[] partitioning = _partition(nodes);
        HashMap feedback_data = new HashMap();
        
        // Initialize the graph
        Graph neighborlists = new Graph(nodes.size());
        for (Node node : nodes) {
            neighborlists.put(node, new NeighborList(k));
        }
        
        internal_builder.setK(k);
        internal_builder.setSimilarity(similarity);
        
        // Loop over all partitions to compute the subgraphs
        // Could be executed in parallel...
        for (int p = 0; p < n_partitions; p++) {

            if (partitioning[p] != null && !partitioning[p].isEmpty()) {

                HashMap subgraph = internal_builder.computeGraph(partitioning[p]);
                computed_similarities += internal_builder.getComputedSimilarities();

                // Add to current neighborlists
                for (Entry e : subgraph.entrySet()) {
                    neighborlists.get(e.getKey()).addAll(e.getValue());
                }
            }

            if (callback != null) {
                feedback_data.put("step", "Building graph inside partition");
                feedback_data.put("partition", p);
                feedback_data.put("computed-similarities", computed_similarities);
                callback.call(feedback_data);
            }
        }
        
        return neighborlists;
        
    }
    
    @Override
    public double estimatedSpeedup() {
        return (double) n_partitions / (oversampling * oversampling);
    }
    
    abstract protected List>[] _partition(List> nodes);
}