vector< vector<u32> > kmeans_pp(const vector< vector<f64> >& points, u32 k,
algo::iLCG& lcg,
vector< vector<f64> >& centers)
{
if (points.size() == 0)
throw eInvalidArgument("There must be points to cluster!");
if (points[0].size() == 0)
throw eInvalidArgument("There must be at least one dimension!");
for (size_t i = 1; i < points.size(); i++)
if (points[i].size() != points[0].size())
throw eInvalidArgument("All data points must have the same dimensionality.");
if (k == 0)
throw eInvalidArgument("Clustering into zero clusters makes no sense.");
centers.resize(k);
centers[0] = points[lcg.next() % points.size()];
for (u32 i = 1; i < k; i++)
{
vector<f64> cumDists(points.size());
for (size_t d = 0; d < cumDists.size(); d++)
{
f64 mind = s_distanceSquared(points[d], centers[0]);
for (u32 j = 1; j < i; j++)
mind = std::min(mind, s_distanceSquared(points[d], centers[j]));
cumDists[d] = (d == 0) ? (mind) : (cumDists[d-1]+mind);
}
f64 ran = ((f64)lcg.next()) / ((f64)lcg.randMax()) * cumDists.back();
size_t d;
for (d = 0; ran > cumDists[d]; d++) { }
centers[i] = points[d];
}
return kmeans(points, k, centers);
}
static
vector<u8> s_genMessage(algo::iLCG& lcg)
{
u32 len = (lcg.next() % kMaxMessageLen) + 1;
vector<u8> mess(len);
for (u32 i = 0; i < len; i++)
mess[i] = (lcg.next() % 256);
return mess;
}
static
vector<f64> s_randPoint(algo::iLCG& lcg, f64 rmin, f64 rmax, u32 dimensions)
{
vector<f64> v(dimensions);
for (u32 i = 0; i < dimensions; i++)
{
v[i] = ((f64)lcg.next()) / ((f64)lcg.randMax()); // [0.0, 1.0]
v[i] *= rmax-rmin; // [0.0, rmax-rmin]
v[i] += rmin; // [rmin, rmax]
}
return v;
}