Henry Wolf IGERT fellow in Neurobiology of Language at the University of Connecticut. Founder of chaoticneural.com and wolfnotes.org.

once more, with feeling

1 min read

I found the first two chapters of An Introduction to Statistical Learning with Applications primarily material covered in the first week of the Machine Learning course on Coursera or the two statistics courses the three of us have taken at the University of Connecticut. I suspect that the third and fourth chapters will be similar in this regard, but more applied to machine learning. The lab portion did provide me with a few new R functions that I had not used previously, so that was quite beneficial. I feel the exercises are likely where we will get the most benefit and I plan to answer at least one of each type of question, conceptual and applied.

“Science is the systematic classification of experience.” – G. H. Lewes

Conceptual #4. (a) One real-life application for classification might be classifying ESL students into classes based on number of years studied, aptitude, grades, and test scores. This may be an inference question, because we want to see which of the independent variables will have an ultimate effect on student performance in their assigned classes. The actual dependent variable could be an expected score in each level, based upon the profiles of past students and their post-class performance. A similar system could help assign appropriate prerequisite courses at a university, based on the performance of students who have or have not take certain classes and the year in school. (b) Regression could be useful in predicting the degree of activation, measured using EEG, a specific point in the brain will have to stimuli. The response would be the brain activation; the predictors would be aspects of the stimuli (e.g. number of letters in a word). Perhaps jumping a bit beyond the scope of this book, this seems like an instance where deep learning would be rather useful. Depending on the stimuli, the goal could be inference or prediction. As scientists, our goal is generally inference. (3) Cluster analysis may be useful analyzing fMRI data. One could identify which brain regions work in concert across time, potentially providing insights that are not obvious to a human observer.

Applied questions were skipped this week due to their being rather elementary.

Henry Wolf IGERT fellow in Neurobiology of Language at the University of Connecticut. Founder of chaoticneural.com and wolfnotes.org.

individual differences in sensitivity to homophony in visual word…

I successfully defended my MS thesis, titled Individual Differences in Sensitivity to Homophony in Visual Word Recognition, on 24 April 2017. Follow the link...
Henry Wolf
20 sec read

new blog series incoming

While there are a large number of useful tutorials that explain machine learning to coders, most seem to lack a basis in curriculum development. By...
Henry Wolf
43 sec read

Leave a Reply

Your email address will not be published. Required fields are marked *