[ml] My NBML Course Wiki Edits-Additions

Brian Morris cymraegish at gmail.com
Thu Jan 13 21:01:33 PST 2011


Hi

I added a bunch of stuff. I just was going to add a few things but I got on
a roll.

A couple notes:

1) the category Other Perspectives is preliminary (the title and the topics)
but a start.

2)  the stuff I added to Clustering I worked on a few years back

3) the math stuff I added, some was response to what I got impression some
people short on math at the meeting, and some is to support the stuff I
added in the main outline

4) notice I added Optimization in more detail to the math appendix --- I was
tempted to edit connecting in the main outline the same category though
under "Lagrange
Optimization<https://www.noisebridge.net/wiki/Machine_Learning/NBML/Machine_Learning/Optimization/Lagrange_Optimization>"
change (somewhere) to "Constrained Optimization" ? But I left it as is.

References are on my other computer,  I will add later.

Brian


On Fri, Jan 7, 2011 at 3:39 PM, Mike Schachter <mike at mindmech.com> wrote:

> Hey Brian,
>
> What kind of simple learning processes do you model? It sounds like
> interesting work.
>
> I'm coming from the bottom-up, and worked in Neuroscience to model
> individual brain cells in the retina as electrical circuits. Some years
> later,
> I'm working on higher level simplified models of individual brain cells and
> on the side trying to better understand network behavior, which is a part
> of the reason I'm focusing on artificial neural networks, as you've noted,
> from the perspective of universal approximation of multivariate functions.
>
> I'm also pretty interested in recurrent networks and adaptive filters.
>
>   mike
>
>
>
> On Thu, Jan 6, 2011 at 8:18 PM, Brian Morris <cymraegish at gmail.com> wrote:
>
>> just for reference: my interest is more in modeling simple learning
>> processes rather than detailed brain structure / function to help understand
>> learning in general including biological and artificial and possibly
>> composite / interfaced systems.  Some people do try and model the brain from
>> bottom up but this is really not practical at least for another several
>> decades and possibly much longer.
>> Even with top-down modeling though, it is important to be faithful to the
>> science.
>>
>> Artificial life and evolutionary algorithms are another way of looking at
>> things.  Natural evolution can hardly be said to be Supervised, at least
>> from a scientific point of view. I had thought some ways Neural Nets were
>> like Artificial Selection, so actually can be beneficial to look at from an
>> EA perspective.
>>
>>
>>
>> Brian
>>
>>
>> On Thu, Jan 6, 2011 at 7:58 PM, Mike Schachter <mike at mindmech.com> wrote:
>>
>>> Hi Brain, George,
>>>
>>> I attended a talk once by a Neurosurgeon who used deep brain stimulation
>>> with electrodes to help Parkinson's patients. It seemed very effective,
>>> but
>>> implanting an electrode into someone's brain is a pretty serious thing to
>>> do.
>>>
>>> Actual brain networks are far more complex than artificial neural
>>> networks.
>>> As I mentioned in another email, there are similarities between the two,
>>> and
>>> studying artificial neural networks can help to understand the brain, but
>>> as
>>> far as I've seen, it's difficult, if not impossible, to replicate
>>> something like
>>> epilepsy in a feedforward artificial neural network. There are highly
>>> detailed
>>> biological models of neural networks which can, but they're not the
>>> subject
>>> of the workshop :)
>>>
>>>  mike
>>>
>>>
>>>
>>>
>>> On Wed, Jan 5, 2011 at 9:40 PM, Brian Morris <cymraegish at gmail.com>wrote:
>>>
>>>> There are pacemaker like devices now which can be used to modulate
>>>> higher level brain activity in cases of gross disfunction such as epilepsy,
>>>> and severe depression. I don't know too much about them, or when something
>>>> like that could be used to treat schizophrenia.
>>>>
>>>> As far as I know, neural nets are used to build some modular functions
>>>> such as object recognition, control as in robotics, memory, and so forth;
>>>> rather than higher level processing.
>>>>
>>>> Also not so easily analyzed themselves in terms of how a given network
>>>> does its thing. So modeling when things go haywire what is going on ==
>>>> pretty hard with these sorts of techniques.
>>>>
>>>> Am I pretty much correct or maybe not ?
>>>>
>>>>
>>>>
>>>> On Wed, Jan 5, 2011 at 8:51 PM, Mike Schachter <mike at mindmech.com>wrote:
>>>>
>>>>> Hi George,
>>>>>
>>>>> The talk is more of a workshop and won't focus too much on the
>>>>> high level math.
>>>>>
>>>>> If you're analyzing EEG output, then yes, neural networks might
>>>>> be able to help you um, well, do something with data. I don't know
>>>>> what you mean by "gaining control of wild thoughts and impulses".
>>>>> Are you talking about some sort of bio-feedback where people learn
>>>>> to control their alpha waves?
>>>>>
>>>>> You should come out either way, I'll be sure to make it appeal to
>>>>> a broader audience and not have too much painful math.
>>>>>
>>>>>   mike
>>>>>
>>>>>
>>>>>
>>>>> On Wed, Jan 5, 2011 at 8:42 PM, george pope <gpope111 at gmail.com>wrote:
>>>>>
>>>>>> If this talk is all about advanced math I doubt that I'll take much
>>>>>> in. If however if I can learn even a little about new ways of gaining
>>>>>> control of wild thoughts and impulses? using "headsets" equipment
>>>>>> brain wave monitors enabling a person to read and act upon tiny brain
>>>>>> signals sent out by our "radio heads?" If so I'll come early to be
>>>>>> sure to get a seat. A valued friend has been addicted to various
>>>>>> delusions for many years.
>>>>>>
>>>>>> george pope
>>>>>> san mateo ca
>>>>>>
>>>>>> On Wed, Jan 5, 2011 at 7:19 PM, Brian Morris <cymraegish at gmail.com>
>>>>>> wrote:
>>>>>> >
>>>>>> >
>>>>>> > ---------- Forwarded message ----------
>>>>>> > From: Mike Schachter <mike at mindmech.com>
>>>>>> > Date: Mon, Jan 3, 2011 at 1:19 PM
>>>>>> > Subject: [ml] Neural Net Workshop on 1/26/2010
>>>>>> > To: "<noisebridge-discuss at lists.noisebridge.net>"
>>>>>> > <noisebridge-discuss at lists.noisebridge.net>
>>>>>> > Cc: ml <ml at lists.noisebridge.net>
>>>>>> >
>>>>>> >
>>>>>> > Hi everyone!
>>>>>> >
>>>>>> > The Machine Learning group at Noisebridge wants to teach
>>>>>> > you! We're holding a workshop on constructing and using
>>>>>> > neural networks, to raise Neural Network Awareness (NNA)
>>>>>> > and money for Noisebridge:
>>>>>> >
>>>>>> > https://www.noisebridge.net/wiki/Neural_Network_Workshop
>>>>>> >
>>>>>> > We hope to combine theory and application in a way that is
>>>>>> > satisfying to both prick mathematicians and laid-back-not-
>>>>>> > gonna-bother-proving-that-shit programmers.
>>>>>> >
>>>>>> > It'll be held on Wednesday, 1/26/2011 around 7:00pm. Bring
>>>>>> > your laptops! I'll post more information closer to the actual
>>>>>> > date.
>>>>>> >
>>>>>> >   mike
>>>>>> >
>>>>>> >
>>>>>> > _______________________________________________
>>>>>> > ml mailing list
>>>>>> > ml at lists.noisebridge.net
>>>>>> > https://www.noisebridge.net/mailman/listinfo/ml
>>>>>> >
>>>>>> >
>>>>>> >
>>>>>> > _______________________________________________
>>>>>> > sf-lug mailing list
>>>>>> > sf-lug at linuxmafia.com
>>>>>> > http://linuxmafia.com/mailman/listinfo/sf-lug
>>>>>> > Information about SF-LUG is at http://www.sf-lug.org/
>>>>>> >
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.noisebridge.net/pipermail/ml/attachments/20110113/83a1411b/attachment.htm 


More information about the ml mailing list