Jokeren / gBolt

gBolt--very fast implementation for gSpan algorithm in data mining
BSD 2-Clause "Simplified" License
52 stars 14 forks source link

Split thread error #15

Closed dionman closed 7 years ago

dionman commented 7 years ago

Running the code on gspan.txt I am receiving split thread error! size 1, thread_size 4

What is the problem in this?

Kind regards, Dionysis

Jokeren commented 7 years ago

I have not updated this repository for long time. Let me have a look.

Jokeren commented 7 years ago

Hi, @dionman , after examining my codes, I notice this error happens because of the ProjectionMap size. That is, the original projection map only contains one start point greater than the support value, so it cannot be split by four threads. I think it will be fine after adjusting the thread number (THREAD_NUM) to 1.

Jokeren commented 7 years ago

@dionman , Hi, have you solved your data?

I have tried it, but it appears that as your edge labels and vertex labels are all 1, gspan cannot efficiently solve the problem.

dionman commented 7 years ago

Hi Keren,

Thanks for your interest. Can I use the node degree as the node label? Moreover, can I apply the algorithm on directed networks?

On Saturday, 13 May 2017, Keren Zhou notifications@github.com wrote:

@dionman https://github.com/dionman , Hi, have you solved your data?

I have tried it, but it appears that as your edge labels and vertex labels are all 1, gspan cannot efficiently solve the problem.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Jokeren/DataMining-gSpan/issues/15#issuecomment-301227502, or mute the thread https://github.com/notifications/unsubscribe-auth/AQilqiPdo92Dguf6sk0UuzeTeh02kBpsks5r5UTRgaJpZM4NLlg2 .

Jokeren commented 7 years ago
  1. Yes, you can. Whereas, you have to know that if all labels are the same, gspan may not have the capability to find FSMs. Indeed, as FSM is an NP problem, I think only approximate algorithms could solve it in that case.

  2. Yes, the algorithm could be applied to directed networks. However, I have not implemented the feature yet. If you need it soon, you could use other repositories, which are not so fast. I will put this issue on my schedule, but sorry that I should concentrate on other projects this month.