commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Phil Steitz <phil.ste...@gmail.com>
Subject Re: [math] releasing 3.2?
Date Wed, 13 Mar 2013 02:28:35 GMT
On 3/10/13 12:20 PM, Luc Maisonobe wrote:
> Le 10/03/2013 19:47, Phil Steitz a écrit :
>> On 3/10/13 11:30 AM, Phil Steitz wrote:
>>> On 3/4/13 4:24 AM, Luc Maisonobe wrote:
>>>> Hi all,
>>>>
>>>> I would like to release version 3.2 of [math] soon.
>>>> Any thoughts about it?
>>>>
>>>> Luc
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
>>>> For additional commands, e-mail: dev-help@commons.apache.org
>>>>
>>>>
>>> Once again, sorry to be late on reviewing stuff.  I think
>>> "DiscreteDistribution" added in MATH-942 is misnamed.  It should be
>>> called "FiniteDistribution" as the standard definition (linked in
>>> the javadoc) applies to distributions such as Poisson and others
>>> (some of which we already have) that have infinitely many values.  
>>> The javadoc should be clarified to indicate clearly what the class
>>> actually is.  I am +1 for adding the class, as it is useful to be
>>> able to construct pmfs from parallel arrays; but the name should be
>>> changed.  Also, I would drop "DiscreteEmpiricalDistribution" as it
>>> basically duplicates EmpiricalDistribution. 
>> Sorry - the last sentence is not correct.  What I should have said
>> is that EmpiricalDistribution should support this functionality.  I
>> had actually started implementing this, but not proposed / committed
>> it yet.  You can force this in EmpiricalDistribution now using
>> singleton bins, but it should be supported directly there.
>>
>>>  I will make the
>>> necessary changes if others are OK with them.
> Sure. You are much more fluent with this stuff than me. I'll let you
> take care of this.

I am working on this.  Hope to finish tonight or tomorrow.  I am
also improving the javadoc, unit tests and exceptions.  Finally, I
think the names should actually be "EnumeratedXxxDistribution" since
in fact we have other finite distributions and what is special about
these is that the pmf is enumerated.

One interesting thing about the implementations is that the
enumerations don't actually have to be functions - i.e. repeats are
allowed and in the case of EnumeratedDistribution, nulls can be
provided and are treated as equal.  So for example, <"dog", 0.2>,
<"pig", 0.1>, <null, 0.1>, <"dog", 0.1>, <null, 0.5> provided to the
constructor results in a pmf that puts 0.1 on "pig", 0.3 on "dog"
and 0.6 on null.  I have clarified the javadoc to make this
explicit.  I think this is fine as long as it is documented.

Phil
>
> Luc
>
>> Including enhancement to EmpiricalDistribution, most likely > 3.2
>>> Phil
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
>> For additional commands, e-mail: dev-help@commons.apache.org
>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
>


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


Mime
View raw message