5 Nov
2013
5 Nov
'13
5:04 p.m.
Hey, this is my first time writing to the Sudoroom list…
I’ve long been interested in the problem of automated species identification, the use of
mathematical methods (in practice, machine learning) to classify organisms on the basis of
morphological data. This is kind of a backwater, I think— DNA barcoding has stolen the
spotlight, as far as automated species identification is concerned. But barcoding has some
substantial problems, and there are groups it just can’t address very well, like fossils.
There’s no reason it can’t work, and in the studies where it’s been tried, it does work
pretty well. It’s just that biologists don’t have a keen interest in machine learning and
computer vision.
I got through most of the Coursera course in machine learning over the summer, and I… kind
of get it? I think? But I don’t understand the subject well enough to implement anything
much myself. My background is in biology, not math or computer science.
A recent paper in PLoS ONE (“Identification of Cichlid Fishes from Lake Malawi Using
Computer Vision,” http://www.plosone.org/article/info:doi/10.1371/journal.pone.0077686 )
illustrates what I’m talking about. The authors use a support vector approach and a random
forest approach to create a program that classifies images of cichlids based on color and
shape information. The paper mentions OpenCV and LIBSVM. Another example comes from a
University of Wisconsin-Madison group that has developed a bee classifier, idBee (Video
here: https://www.youtube.com/watch?v=t_WMYrriXGo , description here:
http://idbee.ece.wisc.edu/ ). Traditional bee classification is a real pain because it
relies on examination of bee mouthparts…
So, um, yeah, count me in as an interested student. I’m probably going to work on
extending the work in the PLoS paper to look at fossil cone snails. (Anyone who thinks
this sounds like a cool project can write me about it, obviously.)
—Tristan
On Nov 4, 2013, at 9:58 AM, Mike Schachter <mschachter(a)eigenminds.com> wrote:
Hi Max! I had some experience volunteering a few hours
a week at some great non-hackerspace nonprofits over the years - an adult literacy center
(Philadelphia Center for Literacy), another place that tutored homeless kids (LA School
on Wheels), and after that a place that focused mostly on writing for kids (826 Valencia
in SF). At Noisebridge, I gave classes on Machine Learning periodically for almost 2
years.
From those experiences, I found that being application driven (opposed to theorem
driven), and having personalized one-on-one or many-on-one instruction was pretty
effective. I expect the students to have diverse needs, from studying for the GED, to
learning enough math for parents to keep up with their child's homework, to people who
want to learn math just for fun. I'm also hoping to move people to the computer a bit
more as well, using python to construct and visualize basic functions (think like
f(x)=x^2), and to learn very basic programming techniques, like variables, arrays, and for
loops.
There are many massive online courses these days, and sometimes at Noisebridge ML people
have formed small groups to work through those courses, which is one way to
"scale" to the level of a hackerspace. I don't have a more specific strategy
right now, and am also comprised primarily of ears and interested in hearing other
people's ideas.
mike
On Mon, Nov 4, 2013 at 9:30 AM, Max Klein <isalix(a)gmail.com> wrote:
Mike,
I would be a potential student and teacher for your ideas. I wonder if you could tell of
any success you've had in the past with teaching Math in the Hackespace? I've been
a part on and off of Math Education at Sudo Room and have a Bachelor's in Math. From
my perspective, Mathematics, is really difficult to teach and learn outside of the School
or University, because it's so prerequisite based. Each new theorem builds on the
last. I feel like I personally have hit a wall with how we can teach math at a Hackerspace
that isn't one-on-one student/master. I'd love to see new approaches and break the
traditional academic mindset of how this is learned, and am all ears.
Max
On 1 November 2013 10:48, Mike Schachter <mschachter(a)eigenminds.com> wrote:
Hi sudoroom! A little while back I mentioned an interest in starting classes that work
towards data literacy, and wanted to give an update with some ideas and plans, with the
hope that people who are interested in helping out or already doing similar things would
come out of the woodwork.
First - I've heard about the morning math meetups through the list and think it
sounds great. My partner and I want to teach a hands-on curriculum for "Math
Literacy" to interested people of all educational backgrounds. It would include
everything from arithmetic to algebra to geometry, trigonometry, and calculus. I'm
hoping to even have a basic introduction to plotting functions and working with arrays
using web-based python terminals.
I want to find interested students and teachers, so if any of this appeals to you, please
let me know! I'm going to make a trip out to sudoroom probably in two weeks, perhaps
we could meet up and talk about it. I'll be at Noisebridge this Tuesday as well.
Second - I'm going to restart machine learning classes at Noisebridge, and am also
looking for interested students and teachers. The ML wiki page, along with a link to join
the mailing list, is here:
https://www.noisebridge.net/Machine_Learning
See you soon,
mike
_______________________________________________
sudo-discuss mailing list
sudo-discuss(a)lists.sudoroom.org
http://lists.sudoroom.org/listinfo/sudo-discuss
_______________________________________________
sudo-discuss mailing list
sudo-discuss(a)lists.sudoroom.org
http://lists.sudoroom.org/listinfo/sudo-discuss