/// <summary>
/// Build the index
/// </summary>
public virtual void Build(MetricDB db, int bsize, Random rand)
{
this.DB = db;
var n = this.DB.Count;
// randomized has very good performance, even compared with more "intelligent" strategies
var dseq = new DynamicSequentialOrdered ();
dseq.Build (db, rand);
this.NODES = new List<Node> (n / bsize + 1);
var L = new List<ItemPair> (n);
while (dseq.Count > 0) {
if (this.NODES.Count % 100 == 0) {
Console.WriteLine ("XXX {0}, bucketSize: {1}, remain {2}/{3}, db: {4}, date-time: {5}",
this, bsize, dseq.Count, db.Count, Path.GetFileName(db.Name), DateTime.Now);
}
var refID = dseq.GetAnyItem ();
dseq.Remove (refID);
L.Clear ();
dseq.ComputeDistances (this.DB[refID], L);
var near = new Result(bsize);
var far = new Result (1);
dseq.AppendKExtremes (near, far, L);
var node = new Node (refID);
this.NODES.Add (node);
dseq.Remove (near);
foreach (var p in near) {
node.Add(p.ObjID, p.Dist);
}
}
}
protected virtual void ComputeDistRow(DynamicSequentialOrdered idxseq, Random rand, HashSet<int> already_pivot, List<EPivot> pivs, List<ItemPair> _items)
{
_items.Clear ();
int n = idxseq.DB.Count;
int piv;
do {
piv = rand.Next(0, n);
} while (already_pivot.Contains(piv));
already_pivot.Add (piv);
var stats = new DynamicSequential.Stats ();
idxseq.ComputeDistances (idxseq.DB [piv], _items, out stats);
int pivID = pivs.Count;
pivs.Add(new EPivot(piv, stats.stddev, stats.mean, stats.min, stats.max, 0, 0));
if (this.Items == null) {
this.Items = new ItemPair[n];
for (int objID = 0; objID < n; ++objID) {
this.Items[objID] = new ItemPair(0, _items[objID].Dist);
}
} else {
for (int objID = 0; objID < n; ++objID) {
var new_piv = pivs[pivID];
var new_dist = _items[objID].Dist;
var old_piv = pivs[ this.Items[objID].ObjID ];
var old_dist = this.Items[objID].Dist;
if (Math.Abs(old_dist - old_piv.mean) < Math.Abs (new_dist - new_piv.mean)) {
this.Items[objID] = new ItemPair(pivID, _items[objID].Dist);
}
}
}
}
protected virtual DynamicSequential.Stats ComputeDistRow(int piv, DynamicSequentialOrdered idxseq, Random rand, List<EPivot> pivs, List<ItemPair> _items)
{
_items.Clear ();
int n = idxseq.DB.Count;
var stats = new DynamicSequential.Stats ();
idxseq.ComputeDistances (idxseq.DB [piv], _items, out stats);
int pivID = pivs.Count;
pivs.Add(new EPivot(piv, stats.stddev, stats.mean, stats.min, stats.max, 0, 0));
if (this.Items == null) {
this.Items = new ItemPair[n];
for (int objID = 0; objID < n; ++objID) {
this.Items[objID] = new ItemPair(0, _items[objID].Dist);
}
} else {
for (int objID = 0; objID < n; ++objID) {
var new_piv = pivs[pivID];
var new_dist = _items[objID].Dist;
var old_piv = pivs[ this.Items[objID].ObjID ];
var old_dist = this.Items[objID].Dist;
if (Math.Abs(old_dist - old_piv.mean) < Math.Abs (new_dist - new_piv.mean)) {
this.Items[objID] = new ItemPair(pivID, _items[objID].Dist);
}
}
}
return stats;
}
public int[] GetMetricShell(object q)
{
var seq = new List<int> ();
var idx = new DynamicSequentialOrdered ();
// optimize the following:
idx.Build (this.R, RandomSets.GetIdentity (this.R.Count));
List<ItemPair> cache = new List<ItemPair>(this.R.Count);
// Console.WriteLine ("START GetMetricShell");
while (idx.Count > 0) {
cache.Clear();
DynamicSequential.Stats stats;
int min_objID, max_objID;
idx.ComputeDistances(q, cache, out stats, out min_objID, out max_objID);
for (int i = 0; i < cache.Count; ++i) {
var obj_min = this.DB [min_objID];
var obj_cur = this.DB [cache[i].ObjID];
if (cache[i].Dist >= this.DB.Dist(obj_min, obj_cur)) {
idx.Remove (cache[i].ObjID);
}
}
//Console.WriteLine ("min: {0}, min_dist: {1}, refs_size: {2}", min_objID, stats.min, idx.Count);
seq.Add (min_objID);
}
return seq.ToArray ();
}
