Machine Learning algorithms are being deployed in Department of Chemical Engineering classrooms for the first time.
'Black boxes' are a metaphorical description for machine learning systems in which the inner workings of the software is not well understood by the user. Often the algorithms at work in the 'boxes' require large amounts of time and knowledge to understand.
The potential of machine learning is rarely exploited in subjects not directly related to computer science, but many academics and teachers recognise that its data-crunching capabilities can be put to good use in a wide variety of situations. However these tools are rarely taught in engineering courses - Professor Erich Muller and Dr Lisa Joss, of the Department of Chemical Engineering, are two such pioneers.
What were the students asked to do?
In the 2017/18 academic year Professor Muller and Dr Joss decided to introduce a machine learning classroom exercise aimed at first-year undergraduate students.
Students were tasked to predict a correlation to calculate the normal boiling point of a fluid from a large data set of over 6000 compounds. In the first instance they were directed to use traditional methods.
The students then ran a subset of their data an artificial neural network (ANN), to produce an engineering-quality correlation. By the end of the task the students got a sense of the power of modern computers to manage large quantities of data. The data set was purposefully large enough to be impractical to sieve through manually.
"Recent advances in computer hardware and algorithms are spawning an explosive growth in the use of computer-based systems aimed at analyzing and ultimately correlating large amounts of data." Professor Erich Muller
Professor Muller said: “Recent advances in computer hardware and algorithms are spawning an explosive growth in the use of computer-based systems aimed at analyzing and ultimately correlating large amounts of data.
“As these machine learning tools become more widespread, it's important that scientists and researchers become familiar with them, both in terms of understanding the tools and the current limitations of artificial intelligence, and more importantly being able to critically separate the hype from the real potential.”
Professor Muller and Dr Joss are hopeful that there is a future for the incorporation of data science into the mainstream learning experience of Chemical Engineering students both at Imperial and beyond.
Dr Joss said: "This project may have been small scale, but it illustrates how to harness the power of machine learning to process large data sets in a simple and effective way in a classroom setting.
"The task exposes students to the vocabulary of machine learning and provides them with a satisfactory resolution of a real-life problem.
"Student feedback on this exercise has been exceptional. There's the recognition that this is an area that is becoming very topical and that the opportunities to incorporate these tools into design and engineering are flourishing."
"Machine Learning for Fluid Property Correlations: Classroom Examples with MATLAB" by Lisa Joss and Erich A. Müller, published 9 April in Journal of Chemical Education
Article text (excluding photos or graphics) © Imperial College London.
Photos and graphics subject to third party copyright used with permission or © Imperial College London.
Communications and Public Affairs