Re: [Phys-L] MOOCs & STEM

[brian whatcott <betwys1@sbcglobal.net>]

On 3/23/2015 1:37 PM, Bernard Cleyet wrote:
> On 2015, Mar 23, , at 07:38, brian whatcott <betwys1@sbcglobal.net> wrote:
>
> > Some graphic evidence on the conjunction of MOOCs with STEM:
> >
> > The code samples may not be helpful, but I found the three graphs interesting.
> > Go  Johns Hopkins!
> >
> > http://tinyurl.com/mh7tcm9
> >
> >
> > Brian Whatcott
>
> The red lines are not labeled.
>
> bc thinks Loren earns a D.  (bc possibly stupid?)
Perhaps it would help to say that Loren, a blogger, was presenting an 
analysis of publicly accessible massive datasets carried out by Toshi 
Takeuchi. Toshi's graph of  MOOC Sessions versus Courses carries three 
gradient lines which show that one unnamed university offering 5 such 
courses has repeated them  30 times, an average of 6 repeats or sessions 
per course. Johns Hopkins, in the same gradient band does almost as well 
with about 190 repeats of 35 courses - also in the band of 3 to 6 
sessions per course.

Talking of massive datasets, I found it tantalizing that in a recent TED 
talk by Jeremy Howard, he repeated that one or a few people working from 
15 minutes to a week or two, could match human expert level performance 
despite an absence of any domain knowledge, by allowing a"Deep Learning" 
application  to mull big datasets,  using the human interaction to 
simply identify features of interest in the recognition space of many 
dimensions.

http://tinyurl.com/ojbtrae

Brian Whatcott