When applying K-means clustering we are picking k initial clusters and then iterating through all the points and assigning them to some cluster and also updating the centers of the clusters. Eventually we do not do any other update. Yet I noticed that sometimes we end up with some empty clusters, in the examples I've done usually this is done in the last iteration. Is this always the case? Is it possible for an empty cluster to become non-empty at some later iterations after it becomes empty?
Edit: Different from a relevant question that was previously asked, I want to know if there is a way to prove or disprove that once it becomes empty it cannot become nonempty.
There is no reason why this would happen late. It can happen very early too.
Clusters can become non-empty again, if you leave them as they are.
Related
Can someone please point me to the part in the documentation explaining the difference between nodes and indices? I'm going over code that was written by someone else and it seems to use nodes and indices interchangeably. Also, when I apply NodeToIndex or IndexToNode on a variable, the value does not change.
Please read: https://developers.google.com/optimization/routing
indices are internal object belonging to the solver, nodes are linked to the distance matrix and the user visits.
In the underlying constraint programming model of routing problems, each stop is exactly visited once. Each stop is a index. The routing library allows several vehicles to start and end at a stop. This causes a conflict because a stop may be visited by several vehicles. In ortools this conflict is resolved by creating dummy indices for nodes that are visited by several vehicles. Hence there may be several indicies that are map to the same node. The depot is a typical example.
This page about the auxillary graph helped me: https://acrogenesis.com/or-tools/documentation/user_manual/manual/tsp/model_behind_scenes.html#the-auxiliary-graph
Problem: It appears to me that a fundamental property of a clustering method c() is whether we can combine the results c(A) and c(B) by some function f() of two clusterings in a way that we do not have to apply the full clustering c(A+B) again but instead do f(c(A),c(B)) and still end up with the same result:
c(A+B) == f(c(A),c(B))
I suppose that a necessary condition for some c() to have this property is that it is determistic, that is the order of its internal processing is irrelevant for the result. However, this might not be sufficient.
It would be really nice to have some reference where to look up which cluster methods support this and what a good f() looks like in the respective case.
Example: At the moment I am thinking about DBSCAN which should be deterministic if I allow border points to belong to multiple clusters at the same time (without connecting them):
One point is reachable from another point if it is in its eps-neighborhood
A core point is a point with at least minPts reachable
An edge goes from every core point to all points reachable from it
Every point with incoming edge from a core point is in the same cluster as the latter
If you miss the noise points then assume that each core node reaches itself (reflexivity) and afterwards we define noise points to be clusters of size one. Border points are non-core points. Afterwards if we want a partitioning, we can assign randomly the border points that are in multiple clusters to one of them. I do not consider this relevant for the method itself.
Supposedly the only clustering where this is efficiently possible is single linkage hierarchical clustering, because edges removed from A x A and B x B are not necessary for finding the MST of the joined set.
For DBSCAN precisely, you have the problem that the core point property can change when you add data. So c(A+B) likely has core points that were not core in either A not B. This can cause clusters to merge. f() supposedly needs to re-check all data points, i.e., rerun DBSCAN. While you can exploit that core points of the subset must be core of the entire set, you'll still need to find neighbors and missing core points.
I have a big doubt about the way I should cluster sets using MinHash together with the banding technique.
I assume everyone reading has a good knowledge of MinHash so I won't define most of the terms I'm using.
My goal is to use MinHash to cluster users according to the similarity of their signatures. In a local, non-banded settings this would be trivial: if their signature hash is the same, they go in the same cluster.
If we split signatures in bands and process them indipendently, I can treat a band as I said before and generate a group of clusters for every band. My question is: how should I aggregate these clusters? Just merge them if they have at least an element in common? Or should I do something different?
Thanks
MinHash is not really meant as standalone clustering algorithm. It is meant as a candidate filter for near-duplicate detection.
When looking for similar documents, you compute the minhashes to retrieve candidates. You then still need to check these candidates - they could be false positives!
The more signatures agree, the more likely they really match.
So if you consider the near-duplicate scenario again: if a is a near duplicate of b and b is a near duplicate of c, then a should also be a near duplicate of c. If this holds, you can throw all these matches (after verification) together. If it doesn't consider a hierarchical clustering like strategy to merge (or not merge) candidates.
I experience a strange situation when running Mahout K-means:
Using the a pre-selected set of initial centroids, I run K-means on a SequenceFile generated by lucene.vector. The run is for testing purposes, so the file is small (around 10MB~10000 vectors).
When K-means is executed with a single mapper (the default considering the Hadoop split size which in my cluster is 128MB), it reaches a given clustering result in 2 iterations (Case A).
However, I wanted to test if there would be any improvement/deterioration in the algorithm's execution speed by firing more mapping tasks (the Hadoop cluster has in total 6 nodes).
I therefore set the -Dmapred.max.split.size parameter to 5242880 bytes, in order to make mahout fire 2 mapping tasks (Case B).
I indeed succeeded in starting two mappers, but the strange thing was that the job finished after 5 iterations instead of 2, and that even at the first assignment of points to clusters, the mappers made different choices compared to the single-map execution . What I mean is that after close inspection of the clusterDump for the first iteration for both two cases, I found that in case B some points were not assigned to their closest cluster.
Could this behavior be justified by the existing K-means Mahout implementation?
From a quick look at the sources, I see two problems with the Mahout k-means implementation.
First of all, the way the S0, S1, S2 statistics are kept is probably not numerically stable for large data sets. Oh, and since k-means actually does not even use S2, it is also unnecessary slow. I bet a good implementation can beat this version of k-means by a factor of 2-5 at least.
For small data sets split onto multiple machines, there seems to be an error in the way they compute their means. Ouch. This will amplify if the reducer is applied to more than one input, in particular when the partitions are small. To be more verbose, the cluster mean apparently is initialized with the previous mean instead of the 0 vector. Now if you if you reduce 't' copies of it, the resulting vector will be off by 't' times the previous mean.
Initialization of AbstractCluster:
setS1(center.like());
Update of the mean:
getS1().assign(x, Functions.PLUS);
Merge of multiple copies of a cluster:
setS1(getS1().plus(cl.getS1()));
Finalization to new center:
setCenter(getS1().divide(getS0()));
So with this approach, the center will be offset from the proper value by the previous center times t / n where t is the number of splits, and n the number of objects.
To fix the numerical instability (which arises whenever the data set is not centered on the 0 vector), I recommend replacing the S1 statistic by the true mean, not S0*mean. Both S1 and S2 can be incrementally updated at little cost using the incremental mean formula which AFAICT was used in the original "k-means" publication by MacQueen (which actually is an online kmeans, while this is Lloyd style batch iterations). Well, for an incremental k-means you obviously need the updatable mean vector anyway... I believe the formula was also discussed by Knuth in his essential books. I'm surprised that Mahout does not seem to use it. It's fairly cheap (just a few CPU instructions more, no additional data, so it all happens in the CPU cache line) and gives you extra precision when you are dealing with large data sets.
I am clustering a large set of points. Throughout the iterations, I want to avoid re-computing cluster properties if the assigned points are the same as the previous iteration. Each cluster keeps the IDs of its points. I don't want to compare them element wise, comparing the sum of the ID vector is risky (a small ID can be compensated with a large one), may be I should compare the sum of squares? Is there a hashing method in Matlab which I can use with confidence?
Example data:
a=[2,13,14,18,19,21,23,24,25,27]
b=[6,79,82,85,89,111,113,123,127,129]
c=[3,9,59,91,99,101,110,119,120,682]
d=[11,57,74,83,86,90,92,102,103,104]
So the problem is that if I just check the sum, it could be that cluster d for example, looses points 11,103 and gets 9,105. Then I would mistakenly think that there has been no change in the cluster.
This is one of those (very common) situations where the more we know about your data and application the better we are able to help. In the absence of better information than you provide, and in the spirit of exposing the weakness of answers such as this in that absence, here are a couple of suggestions you might reject.
One appropriate data structure for set operations is a bit-set, that is a set of length equal to the cardinality of the underlying universe of things in which each bit is set on or off according to the things membership of the (sub-set). You could implement this in Matlab in at least two ways:
a) (easy, but possibly consuming too much space): define a matrix with as many columns as there are points in your data, and one row for each cluster. Set the (cluster, point) value to true if point is a member of cluster. Set operations are then defined by vector operations. I don't have a clue about the relative (time) efficiency of setdiff versus rowA==rowB.
b) (more difficult): actually represent the clusters by bit sets. You'll have to use Matlab's bit-twiddling capabilities of course, but the pain might be worth the gain. Suppose that your universe comprises 1024 points, then you'll need an array of 16 uint64 values to represent the bit set for each cluster. The presence of, say, point 563 in a cluster requires that you set, for the bit set representing that cluster, bit 563 (which is probably bit 51 in the 9th element of the set) to 1.
And perhaps I should have started by writing that I don't think that this is a hashing sort of a problem, it's a set sort of a problem. Yeah, you could use a hash but then you'll have to program around the limitations of using a screwdriver on a nail (choose your preferred analogy).
If I understand correctly, to hash the ID's I would recommend using the matlab Java interface to use the Java hashing algorithms
http://docs.oracle.com/javase/1.4.2/docs/api/java/security/MessageDigest.html
You'll do something like:
hash = java.security.MessageDigest.getInstance('SHA');
Hope this helps.
I found the function
DataHash on FEX it is quiet fast for vectors and the strcmp on the keys is a lot faster than I expected.