[GSoC] HMM formats

[Sergey Bartunov <sb...@gmail.com>]

Hi all. I'd like to discuss several things about Dhruv's and mine
projects which are related to parallel HMM functionality in Mahout.
Since we're working on the different parts of the same thing there are
some shared questions. By this mail I just want to initiate the
communication within us and keep it clear to the community.

1) It's reasonable to use the existing HmmModel class, right?
2) How would you format the input sequence of observed variables? I
saw the discussions here about the avro usage and so on and I wonder
if you could share your considerations.

Re: [GSoC] HMM formats

[Sergey Bartunov <sb...@gmail.com>]

I'd written the ObservedSequenceWritable for myself as the start point
and placed it at my fork on github
https://github.com/sbos/mahout/tree/input

Dhruv, feel free to criticize and/or modify the code. I will keep all
shared parts in the "input" branch. It would be much better to be
compatible with each other from the start of summer)

On 31 May 2011 23:42, Sergey Bartunov <sb...@gmail.com> wrote:
> Hi all. I'd like to discuss several things about Dhruv's and mine
> projects which are related to parallel HMM functionality in Mahout.
> Since we're working on the different parts of the same thing there are
> some shared questions. By this mail I just want to initiate the
> communication within us and keep it clear to the community.
>
> 1) It's reasonable to use the existing HmmModel class, right?
> 2) How would you format the input sequence of observed variables? I
> saw the discussions here about the avro usage and so on and I wonder
> if you could share your considerations.
>