All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.deeplearning4j.eval.curves.PrecisionRecallCurve Maven / Gradle / Ivy

There is a newer version: 1.0.0-M2.1
Show newest version
package org.deeplearning4j.eval.curves;

import com.google.common.base.Preconditions;
import lombok.AllArgsConstructor;
import lombok.Data;
import lombok.EqualsAndHashCode;
import org.nd4j.shade.jackson.annotation.JsonProperty;

import java.util.Arrays;

/**
 * Precision recall curve: A set of (recall, precision) points and different thresholds
 *
 * @author Alex Black
 */
@Data
@EqualsAndHashCode(exclude = {"area"}, callSuper = false)
public class PrecisionRecallCurve extends BaseCurve {

    private double[] threshold;
    private double[] precision;
    private double[] recall;
    private int[] tpCount;
    private int[] fpCount;
    private int[] fnCount;
    private int totalCount;

    private Double area;

    public PrecisionRecallCurve(@JsonProperty("threshold") double[] threshold,
                    @JsonProperty("precision") double[] precision, @JsonProperty("recall") double[] recall,
                    @JsonProperty("tpCount") int[] tpCount, @JsonProperty("fpCount") int[] fpCount,
                    @JsonProperty("fnCount") int[] fnCount, @JsonProperty("totalCount") int totalCount) {
        this.threshold = threshold;
        this.precision = precision;
        this.recall = recall;
        this.tpCount = tpCount;
        this.fpCount = fpCount;
        this.fnCount = fnCount;
        this.totalCount = totalCount;
    }

    @Override
    public int numPoints() {
        return threshold.length;
    }

    @Override
    public double[] getX() {
        return recall;
    }

    @Override
    public double[] getY() {
        return precision;
    }

    @Override
    public String getTitle() {
        return "Precision-Recall Curve (Area=" + format(calculateAUPRC(), DEFAULT_FORMAT_PREC) + ")";
    }

    /**
     * @param i Point number, 0 to numPoints()-1 inclusive
     * @return Threshold of a given point
     */
    public double getThreshold(int i) {
        Preconditions.checkArgument(i >= 0 && i < threshold.length, "Invalid index: " + i);
        return threshold[i];
    }

    /**
     * @param i Point number, 0 to numPoints()-1 inclusive
     * @return Precision of a given point
     */
    public double getPrecision(int i) {
        Preconditions.checkArgument(i >= 0 && i < precision.length, "Invalid index: " + i);
        return precision[i];
    }

    /**
     * @param i Point number, 0 to numPoints()-1 inclusive
     * @return Recall of a given point
     */
    public double getRecall(int i) {
        Preconditions.checkArgument(i >= 0 && i < recall.length, "Invalid index: " + i);
        return recall[i];
    }

    /**
     * @return The area under the precision recall curve
     */
    public double calculateAUPRC() {
        if (area != null) {
            return area;
        }

        area = calculateArea();
        return area;
    }

    /**
     * Get the point (index, threshold, precision, recall) at the given threshold.
* Note that if the threshold is not found exactly, the next highest threshold exceeding the requested threshold * is returned * * @param threshold Threshold to get the point for * @return point (index, threshold, precision, recall) at the given threshold */ public Point getPointAtThreshold(double threshold) { //Return (closest) point number, precision, recall, whether it's interpolated or not //Binary search to find closest threshold int idx = Arrays.binarySearch(this.threshold, threshold); if (idx < 0) { //Not found (usual case). binarySearch javadoc: /* index of the search key, if it is contained in the array; otherwise, (-(insertion point) - 1). The insertion point is defined as the point at which the key would be inserted into the array: the index of the first element greater than the key, or a.length if all elements in the array are less than the specified key. */ idx = -idx - 1; } //At this point: idx = exact, on the next highest double thr = this.threshold[idx]; double pr = precision[idx]; double rec = recall[idx]; return new Point(idx, thr, pr, rec); } /** * Get the point (index, threshold, precision, recall) at the given precision.
* Specifically, return the points at the lowest threshold that has precision equal to or greater than the * requested precision. * * @param precision Precision to get the point for * @return point (index, threshold, precision, recall) at (or closest exceeding) the given precision */ public Point getPointAtPrecision(double precision) { //Find the LOWEST threshold that gives the specified precision for (int i = 0; i < this.precision.length; i++) { if (this.precision[i] >= precision) { return new Point(i, threshold[i], this.precision[i], recall[i]); } } //Not found, return last point. Should never happen though... int i = threshold.length - 1; return new Point(i, threshold[i], this.precision[i], this.recall[i]); } /** * Get the point (index, threshold, precision, recall) at the given recall.
* Specifically, return the points at the highest threshold that has recall equal to or greater than the * requested recall. * * @param recall Recall to get the point for * @return point (index, threshold, precision, recall) at (or closest exceeding) the given recall */ public Point getPointAtRecall(double recall) { //Find the HIGHEST threshold that gives the specified recall for (int i = this.recall.length - 1; i >= 0; i--) { if (this.recall[i] >= recall) { return new Point(i, threshold[i], precision[i], this.recall[i]); } } //Not found - return first point. Should never happen... return new Point(0, threshold[0], precision[0], this.recall[0]); } /** * Get the binary confusion matrix for the given threshold. As per {@link #getPointAtThreshold(double)}, * if the threshold is not found exactly, the next highest threshold exceeding the requested threshold * is returned * * @param threshold Threshold at which to get the confusion matrix * @return Binary confusion matrix */ public Confusion getConfusionMatrixAtThreshold(double threshold) { Point p = getPointAtThreshold(threshold); int idx = p.idx; int tn = totalCount - (tpCount[idx] + fpCount[idx] + fnCount[idx]); return new Confusion(p, tpCount[idx], fpCount[idx], fnCount[idx], tn); } /** * Get the binary confusion matrix for the given position. As per {@link #getPointAtThreshold(double)}. * * @param point Position at which to get the binary confusion matrix * @return Binary confusion matrix */ public Confusion getConfusionMatrixAtPoint(int point) { return getConfusionMatrixAtThreshold(threshold[point]); } public static PrecisionRecallCurve fromJson(String json) { return fromJson(json, PrecisionRecallCurve.class); } public static PrecisionRecallCurve fromYaml(String yaml) { return fromYaml(yaml, PrecisionRecallCurve.class); } @AllArgsConstructor @Data public static class Point { private final int idx; private final double threshold; private final double precision; private final double recall; } @AllArgsConstructor @Data public static class Confusion { private final Point point; private final int tpCount; private final int fpCount; private final int fnCount; private final int tnCount; } }




© 2015 - 2024 Weber Informatics LLC | Privacy Policy