Semisupervised Mean Generate from Track data in supervision region.

[EricR86]

Original report (BitBucket issue) by Sakura Tamaki (Bitbucket: Tamaki_Sakura).

---

Currently in Semi-supervised mode the initial Gaussian mean is generated in the same way as unsupervised mode i.e. based on the mean of the whole track data plus a random noise.

It might be useful to change it, or, at least, add a parameter, such that when initialized the Gaussian Mean in the first step of Expectation Maximization, for each track and each supervision label, the mean will come from the mean from the track data inside all the supervision label regions with that supervision label, instead of the whole track mean.

[EricR86]

Original comment by Sakura Tamaki (Bitbucket: Tamaki_Sakura).

---

• Edited issue description

[EricR86]

Original comment by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

Pull request #25 was made for this enhancement

[EricR86]

Original comment by Sakura Tamaki (Bitbucket: Tamaki_Sakura).

---

• changed state from "new" to "on hold"

We are unsure now if this issue is really useful since in our clustering it is still very likely to shift our mean to elsewhere.