public void init() throws IOException {
super.init();
if (start != null && (coeffBasename != null || order.length > 0))
throw new IllegalArgumentException(
"You cannot choose a preference vector when computing coefficients or derivatives");
// Creates the arrays, if necessary
if (previousRank == null) previousRank = new double[numNodes];
derivative = new double[order.length][subset != null ? subset.length : g.numNodes()];
if (IntArrayList.wrap(order).indexOf(0) != -1)
throw new IllegalArgumentException(
"You cannot compute the derivative of order 0 (use PageRank instead)");
if (coeffBasename != null) BinIO.storeDoubles(rank, coeffBasename + "-" + 0);
logger.info("Completed.");
}
public static void main(final String[] arg)
throws IOException, JSAPException, ConfigurationException, ClassNotFoundException {
SimpleJSAP jsap =
new SimpleJSAP(
WeightedPageRankPowerMethod.class.getName(),
"Computes PageRank of a graph with given graphBasename using the power method."
+ "The resulting doubles are stored in binary form in rankFile."
+ "\n[STOPPING CRITERION] The computation is stopped as soon as two successive iterates have"
+ "an L2-distance smaller than a given threshold (-t option); in any case no more than a fixed"
+ "number of iterations (-i option) is performed.",
new Parameter[] {
new FlaggedOption(
"alpha",
JSAP.DOUBLE_PARSER,
Double.toString(WeightedPageRank.DEFAULT_ALPHA),
JSAP.NOT_REQUIRED,
'a',
"alpha",
"Damping factor."),
new FlaggedOption(
"maxIter",
JSAP.INTEGER_PARSER,
Integer.toString(WeightedPageRank.DEFAULT_MAX_ITER),
JSAP.NOT_REQUIRED,
'i',
"max-iter",
"Maximum number of iterations."),
new FlaggedOption(
"threshold",
JSAP.DOUBLE_PARSER,
Double.toString(WeightedPageRank.DEFAULT_THRESHOLD),
JSAP.NOT_REQUIRED,
't',
"threshold",
"Threshold to determine whether to stop."),
new FlaggedOption(
"coeff",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.NOT_REQUIRED,
'c',
"coeff",
"Save the k-th coefficient of the Taylor polynomial using this basename."),
new FlaggedOption(
"derivative",
JSAP.INTEGER_PARSER,
JSAP.NO_DEFAULT,
JSAP.NOT_REQUIRED,
'd',
"derivative",
"The order(s) of the the derivative(s) to be computed (>0).")
.setAllowMultipleDeclarations(true),
new FlaggedOption(
"preferenceVector",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.NOT_REQUIRED,
'p',
"preference-vector",
"A preference vector stored as a vector of binary doubles."),
new FlaggedOption(
"preferenceObject",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.NOT_REQUIRED,
'P',
"preference-object",
"A preference vector stored as a serialised DoubleList."),
new FlaggedOption(
"startFilename",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.NOT_REQUIRED,
'1',
"start",
"Start vector filename."),
new FlaggedOption(
"buckets",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.NOT_REQUIRED,
'b',
"buckets",
"The buckets of the graph; if supplied, buckets will be treated as dangling nodes."),
new Switch("offline", 'o', "offline", "use loadOffline() to load the graph"),
new Switch(
"strongly",
'S',
"strongly",
"use the preference vector to redistribute the dangling rank."),
new Switch(
"sortedRank",
's',
"sorted-ranks",
"Store the ranks (from highest to lowest) into <rankBasename>-sorted.ranks."),
new FlaggedOption(
"norm",
JSAP.STRING_PARSER,
WeightedPageRank.Norm.INFTY.toString(),
JSAP.NOT_REQUIRED,
'n',
"norm",
"Norm type. Possible values: " + Arrays.toString(WeightedPageRank.Norm.values())),
new UnflaggedOption(
"graphBasename",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.REQUIRED,
JSAP.NOT_GREEDY,
"The basename of the graph."),
new UnflaggedOption(
"rankBasename",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.REQUIRED,
JSAP.NOT_GREEDY,
"The filename where the resulting rank (doubles in binary form) are stored.")
});
JSAPResult jsapResult = jsap.parse(arg);
if (jsap.messagePrinted()) return;
final boolean offline = jsapResult.getBoolean("offline", false);
final boolean strongly = jsapResult.getBoolean("strongly", false);
final boolean sorted = jsapResult.getBoolean("sortedRank", false);
final int[] order = jsapResult.getIntArray("derivative");
final String graphBasename = jsapResult.getString("graphBasename");
final String rankBasename = jsapResult.getString("rankBasename");
final String buckets = jsapResult.getString("buckets");
final String startFilename = jsapResult.getString("startFilename", null);
final String coeffBasename = jsapResult.getString("coeff");
final String norm = jsapResult.getString("norm");
final ProgressLogger progressLogger = new ProgressLogger(LOGGER, "nodes");
ArcLabelledImmutableGraph graph =
offline
? ArcLabelledImmutableGraph.loadOffline(graphBasename, progressLogger)
: ArcLabelledImmutableGraph.loadSequential(graphBasename, progressLogger);
DoubleList preference = null;
String preferenceFilename = null;
if (jsapResult.userSpecified("preferenceVector"))
preference =
DoubleArrayList.wrap(
BinIO.loadDoubles(preferenceFilename = jsapResult.getString("preferenceVector")));
if (jsapResult.userSpecified("preferenceObject")) {
if (jsapResult.userSpecified("preferenceVector"))
throw new IllegalArgumentException("You cannot specify twice the preference vector");
preference =
(DoubleList)
BinIO.loadObject(preferenceFilename = jsapResult.getString("preferenceObject"));
}
if (strongly && preference == null)
throw new IllegalArgumentException("The 'strongly' option requires a preference vector");
DoubleList start = null;
if (startFilename != null) {
LOGGER.debug("Loading start vector \"" + startFilename + "\"...");
start = DoubleArrayList.wrap(BinIO.loadDoubles(startFilename));
LOGGER.debug("done.");
}
WeightedPageRankPowerMethod pr = new WeightedPageRankPowerMethod(graph);
pr.alpha = jsapResult.getDouble("alpha");
pr.preference = preference;
pr.buckets = (BitSet) (buckets == null ? null : BinIO.loadObject(buckets));
pr.stronglyPreferential = strongly;
pr.start = start;
pr.norm = WeightedPageRank.Norm.valueOf(norm);
pr.order = order != null ? order : null;
pr.coeffBasename = coeffBasename;
// cycle until we reach maxIter interations or the norm is less than the given threshold
// (whichever comes first)
pr.stepUntil(
or(
new WeightedPageRank.NormDeltaStoppingCriterion(jsapResult.getDouble("threshold")),
new WeightedPageRank.IterationNumberStoppingCriterion(jsapResult.getInt("maxIter"))));
System.err.print("Saving ranks...");
BinIO.storeDoubles(pr.rank, rankBasename + ".ranks");
if (pr.numNodes < 100) {
for (int i = 0; i < pr.rank.length; i++) {
System.err.println("PageRank[" + i + "]=" + pr.rank[i]);
}
}
Properties prop = pr.buildProperties(graphBasename, preferenceFilename, startFilename);
prop.save(rankBasename + ".properties");
if (order != null) {
System.err.print("Saving derivatives...");
for (int i = 0; i < order.length; i++)
BinIO.storeDoubles(pr.derivative[i], rankBasename + ".der-" + order[i]);
}
final double[] rank = pr.rank;
pr = null; // Let us free some memory...
graph = null;
if (sorted) {
System.err.print("Sorting ranks...");
Arrays.sort(rank);
final int n = rank.length;
int i = n / 2;
double t;
// Since we need the ranks from highest to lowest, we invert their order.
while (i-- != 0) {
t = rank[i];
rank[i] = rank[n - i - 1];
rank[n - i - 1] = t;
}
System.err.print(" saving sorted ranks...");
BinIO.storeDoubles(rank, rankBasename + "-sorted.ranks");
System.err.println(" done.");
}
}
public static void main(final String[] arg)
throws IOException, JSAPException, NoSuchMethodException {
final SimpleJSAP jsap =
new SimpleJSAP(
BloomFilter.class.getName(),
"Creates a Bloom filter reading from standard input a newline-separated list of terms.",
new Parameter[] {
new FlaggedOption(
"bufferSize",
IntSizeStringParser.getParser(),
"64Ki",
JSAP.NOT_REQUIRED,
'b',
"buffer-size",
"The size of the I/O buffer used to read terms."),
new FlaggedOption(
"encoding",
ForNameStringParser.getParser(Charset.class),
"UTF-8",
JSAP.NOT_REQUIRED,
'e',
"encoding",
"The term file encoding."),
new UnflaggedOption(
"bloomFilter",
JSAP.STRING_PARSER,
JSAP.NO_DEFAULT,
JSAP.REQUIRED,
JSAP.NOT_GREEDY,
"The filename for the serialised front-coded list."),
new UnflaggedOption(
"size",
JSAP.INTSIZE_PARSER,
JSAP.NO_DEFAULT,
JSAP.REQUIRED,
JSAP.NOT_GREEDY,
"The size of the filter (i.e., the expected number of elements in the filter; usually, the number of terms)."),
new UnflaggedOption(
"precision",
JSAP.INTEGER_PARSER,
JSAP.NO_DEFAULT,
JSAP.REQUIRED,
JSAP.NOT_GREEDY,
"The precision of the filter.")
});
JSAPResult jsapResult = jsap.parse(arg);
if (jsap.messagePrinted()) return;
final int bufferSize = jsapResult.getInt("bufferSize");
final String filterName = jsapResult.getString("bloomFilter");
final Charset encoding = (Charset) jsapResult.getObject("encoding");
BloomFilter filter = new BloomFilter(jsapResult.getInt("size"), jsapResult.getInt("precision"));
final ProgressLogger pl = new ProgressLogger();
pl.itemsName = "terms";
pl.start("Reading terms...");
MutableString s = new MutableString();
FastBufferedReader reader =
new FastBufferedReader(new InputStreamReader(System.in, encoding), bufferSize);
while (reader.readLine(s) != null) {
filter.add(s);
pl.lightUpdate();
}
pl.done();
BinIO.storeObject(filter, filterName);
}
