Deep Learning
Hugo Larochelle, YouTube, January 23, 2014

I spent the better part of the day Thursday watching Hugo Larochelle's presentation on deep learning. This is basically the use of neural networks to support unsupervised machine learning. The morning covered deep learning basics, and the afternoon looked at text analysis using deep learning. This was really good stuff. Some of the mathematics I couldn't follow, but I got the gist of most of it. And anyhow, he's placed a mountain of videos online. My photo is from my 'photo-a-day' set, newly revived for 2014.

[Link] [Comment]

General Education's Remake
Paul Fain, Inside Higher Ed, January 23, 2014

To be clear, I don't think competency-based education is bad, per se, and I think it is certainly a better approach than what might be called curricular or subject-based edducation. I do have my doubts about the utility of breaking a larger piece of education into bite-sized competency nuggets, partially because it's pretty easy to lose the plot when negotiating these morsels, and partially because it doesn't actually make the learning (and especially the assessment) or the overall skill easier or more efficient. The other issue I have surrounds the remapping of the curriculum, especially when the objectives become to "develop the whole student for personal growth, economic productivity, and responsible citizenship." I do like the idea of "prepar[ing] students to tackle complex and unscripted problems – to apply evidence-based reasoning, judgment and ethical responsibility to questions where the answer is not known and the consequences matter." But this is only a small part of the 'integrative liberal learning' remapping. And I distrust the intentions of the Gates Foundation grant that is behind all this (check out these slides, for example). Related: the association’s Liberal Education and America’s Promise(LEAP), which promotes these essential learning outcomes. Maybe I'm too sceptical. I don't know.

[Link] [Comment]

Epidemiological modeling of online social network dynamics
John Cannarella, Joshua A. Spechler, arXiv, January 22, 2014

Well, this is putting your model to the predictive test. The authors write, "Extrapolating the best fit model into the future suggests that Facebook will undergo a rapid decline in the coming years, losing 80% of its peak user base between 2015 and 2017." I for one would not be surprised to see that happen (I wonder what the model says about Google - I'd say that in ten years they will be where Microsoft is today; Apple, meanwhile, will be far below where they are now, behind giants like Asus). Via Time. Related: the problem with Facebook is that it keeps most things from you.

[Link] [Comment]

To Each According to its Degree: The Meritocracy and Topocracy of Embedded Markets
J. Borondo, F. Borondo, C. Rodriguez-Sickert, C. A. Hidalgo, Nature.com, January 22, 2014

Interesting paper. "A system is said to be meritocratic if the compensation and power available to individuals is determined by their abilities and merits. A system is topocratic if the compensation and power available to an individual is determined primarily by her position in a network." Different structures create networks that are either meritocratic or topocratic. This paper introduces "a model that is perfectly meritocratic for fully connected networks but that becomes topocratic for sparse networks-like the ones in society." In particular, "the model predicts that meritocracy increases in societies that become better connected." This (to me) has implications for public policy (and education policy) - for example, open acccess (by implication) increases meritocracy.

[Link] [Comment]