mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dan Filimon <>
Subject Re: MultiNormal distribution radius
Date Wed, 14 Nov 2012 11:18:53 GMT
On Wed, Nov 14, 2012 at 12:33 PM, Sean Owen <> wrote:
> In the univariate case, there is no max/min possible value. We just have
> the variance to say how unlikely a value is that is far from the
> distribution mean, though any value is possible. Same in multivariate, so I
> don't think you could say the distribution fits strictly inside a sphere.

Right, I probably want a modified version in my case where I normalize
the distances somehow.

> The distribution will only be symmetrical and not 'elongated' if the
> variances are the same, which is the case I think you're talking about.
> Ted I am also confused by the naming in this class. What I'd imagine is the
> vector of means is called "offset". The variances come in to the picture
> via a matrix called "mean". (That's not the covariance matrix right? might
> expect that from an API perspective but I don't think that's how it is
> used.) And the parameter for the case where all variances are the same is
> "radius".

So "radius" in this case is the variance of the ith component of the
vector, since the covariance matrix is diagonal?

> On Wed, Nov 14, 2012 at 8:32 AM, Dan Filimon <>wrote:
>> Hi,
>> I'm familiar with the basic univariate normal distribution but am
>> having trouble understanding how the Mahout multivariate normal
>> distribution works.
>> Specifically, what does the radius of the distribution stand for?
>> What I'm imagining (at lest for 3 dimensions) is that all points would
>> fit into a sphere centered in the mean with the given radius and that
>> they would be normally distributed inside.
>> This however doesn't seem to be the case (unless my tests are broken).
>> What am I missing?
>> Thanks!

View raw message