You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@commons.apache.org by Daniel Winterstein <da...@sodash.com> on 2016/05/16 13:40:06 UTC

[Math] Proposal: Add smoothing and fitting to the Kalman Filter

Hello,

I was doing some work with Kalman Filters recently, and I noticed the
implementation in commons.math is missing a couple of features, namely
smoothing and model-fitting.

Would it be good to add these?

A little bit more on the model and the proposed features:
The Kalman Filter is a model for time-series data.
The version in commons.maths can do prediction (i.e. given the state
now, predict the next state).
The algorithm can also do smoothing (given a sequence of data, perhaps
with gaps, estimate the most likely state at each step) -- but this
version doesn't have an implementation.
Also, given a time-series dataset, you can fit a Kalman Filter. There
isn't a precise solution, but using expectation maximisation works
well.

I am new to the Apache Commons mailing list -- please forgive me if
this is the wrong way to go about things.

Kind regards,
 - Daniel

-- 
--------------------------------------
Dr Daniel Winterstein
Director
Edinburgh            +44 (0)772 5172 612
http://sodash.com   http://sogrow.co.uk

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Re: [Math] Proposal: Add smoothing and fitting to the Kalman Filter

Posted by Gilles <gi...@harfang.homelinux.org>.
Hi.

On Tue, 17 May 2016 12:47:59 +0100, Daniel Winterstein wrote:
> Thanks Gilles
>
> This is a new feature, but extends an existing class. Should I target
> version 3.7 or version 4.0 of commons.math?

4.0

> By the way, what are the key differences in version 4 of 
> commons.math?

For several months, new features and bug fixes have not been backported
to the 3.x line.


Regards,
Gilles


> Thank you,
>  - Daniel
>
> On 16 May 2016 at 17:48, Gilles <gi...@harfang.homelinux.org> wrote:
>> Hi.
>>
>> On Mon, 16 May 2016 14:40:06 +0100, Daniel Winterstein wrote:
>>>
>>> Hello,
>>>
>>> I was doing some work with Kalman Filters recently, and I noticed 
>>> the
>>> implementation in commons.math is missing a couple of features, 
>>> namely
>>> smoothing and model-fitting.
>>>
>>> Would it be good to add these?
>>
>>
>> Sure.
>>
>>> A little bit more on the model and the proposed features:
>>> The Kalman Filter is a model for time-series data.
>>> The version in commons.maths can do prediction (i.e. given the 
>>> state
>>> now, predict the next state).
>>> The algorithm can also do smoothing (given a sequence of data, 
>>> perhaps
>>> with gaps, estimate the most likely state at each step) -- but this
>>> version doesn't have an implementation.
>>> Also, given a time-series dataset, you can fit a Kalman Filter. 
>>> There
>>> isn't a precise solution, but using expectation maximisation works
>>> well.
>>>
>>> I am new to the Apache Commons mailing list -- please forgive me if
>>> this is the wrong way to go about things.
>>
>>
>> This is the right way.
>> All significant changes/additions to the code must be decided on 
>> here.
>>
>> Then the implementation details can be discussed within a dedicated 
>> issue
>> on the bug-tracking system:
>>   https://issues.apache.org/jira/browse/MATH
>>
>>
>> Thanks in advance for your contributions,
>> Gilles
>>
>>
>>>
>>> Kind regards,
>>>  - Daniel


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Re: [Math] Proposal: Add smoothing and fitting to the Kalman Filter

Posted by Daniel Winterstein <da...@sodash.com>.
Thanks Gilles

This is a new feature, but extends an existing class. Should I target
version 3.7 or version 4.0 of commons.math?

By the way, what are the key differences in version 4 of commons.math?

Thank you,
 - Daniel

On 16 May 2016 at 17:48, Gilles <gi...@harfang.homelinux.org> wrote:
> Hi.
>
> On Mon, 16 May 2016 14:40:06 +0100, Daniel Winterstein wrote:
>>
>> Hello,
>>
>> I was doing some work with Kalman Filters recently, and I noticed the
>> implementation in commons.math is missing a couple of features, namely
>> smoothing and model-fitting.
>>
>> Would it be good to add these?
>
>
> Sure.
>
>> A little bit more on the model and the proposed features:
>> The Kalman Filter is a model for time-series data.
>> The version in commons.maths can do prediction (i.e. given the state
>> now, predict the next state).
>> The algorithm can also do smoothing (given a sequence of data, perhaps
>> with gaps, estimate the most likely state at each step) -- but this
>> version doesn't have an implementation.
>> Also, given a time-series dataset, you can fit a Kalman Filter. There
>> isn't a precise solution, but using expectation maximisation works
>> well.
>>
>> I am new to the Apache Commons mailing list -- please forgive me if
>> this is the wrong way to go about things.
>
>
> This is the right way.
> All significant changes/additions to the code must be decided on here.
>
> Then the implementation details can be discussed within a dedicated issue
> on the bug-tracking system:
>   https://issues.apache.org/jira/browse/MATH
>
>
> Thanks in advance for your contributions,
> Gilles
>
>
>>
>> Kind regards,
>>  - Daniel
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>



-- 
--------------------------------------
Dr Daniel Winterstein
Director
Edinburgh            +44 (0)772 5172 612
http://sodash.com   http://sogrow.co.uk

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Re: [Math] Proposal: Add smoothing and fitting to the Kalman Filter

Posted by Gilles <gi...@harfang.homelinux.org>.
Hi.

On Mon, 16 May 2016 14:40:06 +0100, Daniel Winterstein wrote:
> Hello,
>
> I was doing some work with Kalman Filters recently, and I noticed the
> implementation in commons.math is missing a couple of features, 
> namely
> smoothing and model-fitting.
>
> Would it be good to add these?

Sure.

> A little bit more on the model and the proposed features:
> The Kalman Filter is a model for time-series data.
> The version in commons.maths can do prediction (i.e. given the state
> now, predict the next state).
> The algorithm can also do smoothing (given a sequence of data, 
> perhaps
> with gaps, estimate the most likely state at each step) -- but this
> version doesn't have an implementation.
> Also, given a time-series dataset, you can fit a Kalman Filter. There
> isn't a precise solution, but using expectation maximisation works
> well.
>
> I am new to the Apache Commons mailing list -- please forgive me if
> this is the wrong way to go about things.

This is the right way.
All significant changes/additions to the code must be decided on here.

Then the implementation details can be discussed within a dedicated 
issue
on the bug-tracking system:
   https://issues.apache.org/jira/browse/MATH


Thanks in advance for your contributions,
Gilles


>
> Kind regards,
>  - Daniel


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org