Naive Bayesian Classifier for Golang
Jake
https://github.com/jbrukh/bayesian
Thanks,
Jake.
Dmitry Chestnykh
Nice!
Style suggestion: please don't use "this" to name struct references. This makes code less readable (and unidiomatic). For example, I first looked at Score function, where you have:
this.classes
then I looked at getWordProb function where you have:
this.freq
So, naturally, I assumed that "this" here and there are the same object. It turned out that was not the case:
func (this *classData) getWordProb(...
func (this *Classifier) Score(...
Suggested names:
func (d *classData) getWordProb(...
func (cls *Classifier) Score(...
Or something like this.
On the code: I'm not totally sure, but I think there's a float underflow hiding somewhere.
-Dmitry
John Asmuth
Style suggestion: please don't use "this" to name struct references. This makes code less readable (and unidiomatic).
roger peppe
you might use that, but it's certainly not idiomatic go.
i like the fact that if one is using a normal identifier, refactoring
between function and method is trivial.
func X(me *Foo)
doesn't look so good.
i wish it was possible to use gofmt to rename single letter variables.
Jan Mercl
Rick
I'm not sure if short variable names (shrtVarNms) are considered
idiomatic Go, but for the sake of a few characters, I much prefer
func (data *classData) getWordProb(...
func (classifier *Classifier) Score(...
- Rick (not Rck ;-)
John Asmuth
On Wednesday, November 23, 2011 8:52:45 AM UTC-5, rog wrote:
you might use that, but it's certainly not idiomatic go.
i like the fact that if one is using a normal identifier, refactoring
between function and method is trivial.func X(me *Foo)
doesn't look so good.
John Asmuth
Jake
sides of the argument -- to me, using "this" or using short
abbreviated names like "d" each are confusing in their own way. :) In
terms of functionality, do you recommend going to float64, then?
Thanks,
Jake
John Asmuth
terms of functionality, do you recommend going to float64, then?
roger peppe
> On Wednesday, November 23, 2011 8:52:45 AM UTC-5, rog wrote:
>>
>> you might use that, but it's certainly not idiomatic go.
>
> Where can I find this description of idiomatic go, anyway?
> I'd say that methodology rather than choice of identifiers is what should or
> should not fall under "idiomatic go".
the Go libraries are the best expression of idiomatic Go that i know.
you won't find a single instance of "me" or "this" as a receiver
identifier there. that's a pretty good hint IMHO. of course, it *is*
a fairly minor point, but consistency of style is useful.
in Go, the receiver is just a normal variable like any other - why
should it be named differently because it's an implicit
argument?
>> i like the fact that if one is using a normal identifier, refactoring
>> between function and method is trivial.
>>
>> func X(me *Foo)
>>
>> doesn't look so good.
>
> This seems irrelevant.
it's relevant when refactoring
func (f *Foo) X()
to
func X(f *Foo)
or
func (f *Foo) DoWithBar(b *Bar)
to
func (b *Bar) DoWithFoo(f *Foo)
Jonathan Amsterdam
_, present := data.freqs[word]
if !present {
data.freqs[word] = 1
} else {
data.freqs[word]++
}
as this:
data.freqs[word]++
Jake
Dmitry Chestnykh
In terms of functionality, do you recommend going to float64, then?
This would help a bit until you underflow float64 :-)
Here are some resources:
1. https://en.wikipedia.org/wiki/Bayesian_spam_filtering#Other_expression_of_the_formula_for_combining_individual_probabilities
2. http://nlp.stanford.edu/IR-book/html/htmledition/naive-bayes-text-classification-1.html
3. http://stackoverflow.com/questions/2691021/problem-with-precision-floating-point-operation-in-c
-Dmitry
Jake
of these methods.
On Nov 24, 6:50 am, Dmitry Chestnykh <dch...@gmail.com> wrote:
> On Wednesday, November 23, 2011 5:41:07 PM UTC+1, Jake wrote:
>
> > In terms of functionality, do you recommend going to float64, then?
>
> This would help a bit until you underflow float64 :-)
>
> Here are some resources:
>
> 1.https://en.wikipedia.org/wiki/Bayesian_spam_filtering#Other_expressio...
> 2.http://nlp.stanford.edu/IR-book/html/htmledition/naive-bayes-text-cla...
> 3.http://stackoverflow.com/questions/2691021/problem-with-precision-flo...
>
> -Dmitry
Johan Liebert
Naive Bayes classifiers are a collection of classification algorithms based on Bayes' Theorem. It is not a single algorithm but a family of algorithms where all of them share a common principle, i.e. every pair of features being classified is independent of each other.