Date | Time
12/06/2015 | 2 h 00 min - 6 h 00 min
Le prof. Antony D. Rollett donne un séminaire sur le sujet
« Data mining for correlation research in métallurgie »
The Materials Genome argues for the use of data science to accelerate materials development. Although there has been much discussion of “big data”, the reality is that materials development will mostly be done with normal, i.e. small data sets. Nevertheless data science has many tools to offer us that can help us analyze and understand our data sets, especially when the number of parameters is more than, say, three (think of the number of columns in a spreadsheet). Seeking correlations between variables is a formal way of asking whether, say, yield stress depends on fraction recrystallized. Finding combinations of variables that can be linked to another variable is another natural analysis, which in traditional metallurgy appears in the form of equations that link, say, martensite start temperature to a combination of composition variables. Such analyses can be performed with the help of standard, open source statistical analysis tools such as Canonical Correlation Analysis and Principal Component Analysis. A convenient framework for such analyses is the open source R package. Students are invited to install R on their laptops and come prepared to step through the examples that will be demonstrated in this seminar.
Link to the package to install: www.r-project.org
The seminar will be available by Internet connection to Adobe-Connect, on simple PC or laptop at the address:
Connection will be open from 13H on the 12th June.