Teach Machine Learning with Excel

Yumin Zhang

Download Paper | Permalink

Conference: 2020 ASEE Virtual Annual Conference Content Access
Location: Virtual On line
Publication Date: June 22, 2020
Start Date: June 22, 2020
End Date: June 26, 2021
Conference Session: Engineering Physics and Physics Division Technical Session 3
Tagged Division: Engineering Physics and Physics
Page Count: 10
DOI: 10.18260/1-2--35268
Permanent URL: https://peer.asee.org/35268
Download Count: 1911

Request a correction

Paper Authors

biography

Yumin Zhang Southeast Missouri State University

visit author page

Yumin Zhang is a professor in the Department of Engineering and Technology, Southeast Missouri State University. His research interests include semiconductor devices, electronic circuits, neural networks, and engineering education.

visit author page

Download Paper | Permalink

Abstract

Teach Machine Learning with Excel

Engineering and physics curricula are math intensive, and these math courses form an unsurmountable barrier for many Gen-Z students. Fortunately, there are a few handy math software packages, which can help students deal with this challenge. However, with the exponential growth of human knowledge and data, traditional math tools now become inadequate, so students need to learn new tools in dealing with the huge amount of data.

Machine learning (ML) now quickly becomes the new tools for engineers and physicists to extract essential information from large amounts of data, either from experiments or simulations. Actually, ML also provides an exciting opportunity to learn the models themselves. Beyond the realm of explicit knowledge that can be expressed with simple math formula, there is a much broader realm of implicit or dark knowledge, which cannot be analyzed in the traditional way.

Many concepts in ML were borrowed from statistical physics and information theory, such as cross entropy. In addition, many approaches in ML are also derived from physics and engineering, such as the backpropagation process. Therefore, students with engineering and physics background have little trouble in understanding the concepts in ML.

On the other hand, coding is a serious challenge for many students in learning ML. Therefore, it can be divided into two stages in learning machine learning: In the first stage students concentrate on the concepts and methods of ML with very simple programming tools, and in the second stage students will practice with more advanced tools, such as Tensorflow.

In Spring 2019 ML was introduced in a one credit hour course, and students met once a week in learning the related concepts and methods. The content of this course covered perceptron model and deep neural network in detail, and the more advanced topics were also introduced, such as CNN, RNN, and reinforcement learning. One of the most challenging topics was the backpropagation, since huge numbers of the weights can be modified just from the information of a single output. In order to lower the threshold of programming, Excel was used in training the simple neural networks in this course. Since the math formulae were very simple and the training process was also very short, Excel was capable to handle all the procedures. In addition, the data of every layer of neural network were able to be presented on the same page, and the algorithm could be revised very easily. It was proved to be a good platform to introduce ML to students without extensive programming background.

Citation
Format

Zhang, Y. (2020, June), Teach Machine Learning with Excel Paper presented at 2020 ASEE Virtual Annual Conference Content Access, Virtual On line . 10.18260/1-2--35268

TY - CPAPER
AB - Teach Machine Learning with Excel

Engineering and physics curricula are math intensive, and these math courses form an unsurmountable barrier for many Gen-Z students. Fortunately, there are a few handy math software packages, which can help students deal with this challenge. However, with the exponential growth of human knowledge and data, traditional math tools now become inadequate, so students need to learn new tools in dealing with the huge amount of data.

Machine learning (ML) now quickly becomes the new tools for engineers and physicists to extract essential information from large amounts of data, either from experiments or simulations. Actually, ML also provides an exciting opportunity to learn the models themselves. Beyond the realm of explicit knowledge that can be expressed with simple math formula, there is a much broader realm of implicit or dark knowledge, which cannot be analyzed in the traditional way.

Many concepts in ML were borrowed from statistical physics and information theory, such as cross entropy. In addition, many approaches in ML are also derived from physics and engineering, such as the backpropagation process. Therefore, students with engineering and physics background have little trouble in understanding the concepts in ML.

On the other hand, coding is a serious challenge for many students in learning ML. Therefore, it can be divided into two stages in learning machine learning: In the first stage students concentrate on the concepts and methods of ML with very simple programming tools, and in the second stage students will practice with more advanced tools, such as Tensorflow.

In Spring 2019 ML was introduced in a one credit hour course, and students met once a week in learning the related concepts and methods. The content of this course covered perceptron model and deep neural network in detail, and the more advanced topics were also introduced, such as CNN, RNN, and reinforcement learning. One of the most challenging topics was the backpropagation, since huge numbers of the weights can be modified just from the information of a single output.

In order to lower the threshold of programming, Excel was used in training the simple neural networks in this course. Since the math formulae were very simple and the training process was also very short, Excel was capable to handle all the procedures. In addition, the data of every layer of neural network were able to be presented on the same page, and the algorithm could be revised very easily. It was proved to be a good platform to introduce ML to students without extensive programming background.

AU - Yumin Zhang
CY - Virtual On line
DA - 2020/06/22
PB - ASEE Conferences
TI - Teach Machine Learning with Excel
UR - https://peer.asee.org/35268
DO - 10.18260/1-2--35268
ER -