Shadow Learning
Abstract
This article discusses a systematic approach I created to optimize efficiency of learning theoretical concepts based on notes I took in a Shadow Day event for Software Engineering program in University of Waterloo. This article approaches the essence of learning methods and its implications to learning in 21th century with a focus on its usage in software engineering and computer science. The essential techniques of shadow learning are covered in details, and comparisons with other learning methods are explored. Last, we will discuss its implication to information processing and artificial intelligence. Shadow Learning, being a framework in essence, does not cover the whole learning cycle. It is a technique to structure knowledge to improve efficiency.
Introduction
Throughout my self-learning journey, I developed a learning methodology that helps me to learn theoretical concepts more quickly and efficiently from any medium of information — — the Shadow Learning Approach, which was inspired by my notes taken at a shadow day in Waterloo Software Engineering.
Learning Methods
I define learning approaches as the systematic and creative operations to acquire knowledge and skills efficiently and to reinforce this knowledge and skills until one reaches a relatively high proficiency in the skills and knowledge they are learning. Note that successful learning approaches should also ensure that the learner can apply theoretical knowledge in various different contexts. Learning approaches are essential for learners because they not only guide them through the learning process but also ensure they learn more efficiently. In this digital age, especially in the field of Software Engineering and Computer Science, fast learning and self-learning are required for one to be able to survive the fast pace of iteration in technologies. Personalized learning approaches can help one to learn fast and apply the knowledge and skills to achieve great results in various fields. For students, these learning approaches will enable them to learn technical skills fast to participate in technical extracurricular activities. I would accredit most of my achievements in high school to the two learning approaches I developed — — Shadow Learning and the Banana Bread Approach. With the development of ML and AI, these learning approaches also become the basis of the cutting edge AI and ML research. I believe these learning approaches are not only an effective tool for my learning in university, but also a starting point for me in my journey to pursue Machine Learning and Artificial Intelligence.
Introduction to Shadow Learning
Very often, the lecturers might not provide clear notes or slides for students. There also might be a time that a COOP student is required to learn a framework in a short period of time from a disorganized documentation. These are the times to apply Shadow Learning in your studies or your work. I usually create a semantic network that quickly grasps all the concepts in a lecture or textbook, which can be as simple as bullet-point lists. Then elaborate on the details in each concept such as definitions, properties, applications. The key to this semantic network, which I call “shadowing”, is to understand the relationship between concepts and how each of them fits into the semantic network. The essence behind it is to organize the information in a meaningful way so that our brain can process it more quickly.
Shadow Learning Essentials
We are always encouraged to look at “the bigger picture.” The idea of a “bigger picture” is the core of the Shadow Learning approach as it helps us to look at individual concepts at a holistic level and be aware of their roles in the system. The best way to use this strategy is actually to focus on the relationship between concepts and use the simplest and fastest way to translate the idea onto the paper — — our written language, discipline-specific notations, and graphical representation.
To best represent the relationship between concepts, I use the directory structure which is widely used in the file system as it is the best way to store information up till now. Another benefit of a directory structure is that it presents the information in a structured way so that we can find the relationship between concepts more quickly. The semantic network represented in this way consists of all the key points in the source with each node representing the point and its siblings including the details of this concept such as definition, property, application, graphical representation, essence. With this approach, I actually stored all the information in a way that my brain understands. When I want to access it, by looking at the “shadow,” my brain will carry out a search very similar to a search in the tree and access all the details corresponding to the concept while understanding its position in the semantic network.
Applications
Shadow learning is very helpful in learning theoretical concepts. I used shadow learning to prepare for my Organic Chemistry unit in Chemistry 4U. Organic Chemistry involves a great deal of functional groups and complicated reactions between them. I first organized the unit in two categories: Functional Groups and Reactions. Then I make a list of the functional groups and insert these functional groups into a directory structure. Next, the details of each group, such as nomenclature, shapes, were added in the child nodes of each functional group. The Reactions chart is made in a similar method.
This approach is extremely helpful when learning a new framework at the job or for projects. A lot of the new frameworks these days involves new structure, concepts, or tools that are unique to the framework, and shadow learning is extremely effective in learning a family of new tools that are connected together.
Inspiration
The Shadow Learning approach is inspired by the Mind Map developed by Tony Buzan in the 1950s. The shadow learning approach is based on a similar concept as the Mind Map except it uses a file directory structure and uses less graphic representations. The Mind Map is very useful in brainstorming, but it may not be entirely suitable for Engineering and Math lectures as these lectures can be very abstract and logic-based. The downside of a mind map is that it will be inefficient if a student doesn’t have their own symbolic representation for the mind map. According to my observation in my school, most students spend too much time figuring out what symbol or graph to use to represent a concept and eventually cannot keep up with the lecture. Also, it will be more difficult to find a symbolic representation of abstract concepts in advanced math and engineering courses.
Implications to Information Processing
As I use this strategy more in my learning, I find that it actually resembles the way the human brain processes information. When we first encounter new concepts, our brain will do a query into the knowledge base to search for the definition or explanation related but cannot find it. Oftentimes, most of the students will be confused and just try to memorize it. Our brain tends to memorize new concepts when it cannot find a better way to interpret it. Even though a student may memorize all the concepts, these concepts often are stored discretely. Usually, after some time, they will likely forget about the new concepts if they don’t take action to actively understand them. Shadow Learning provides students with a way to establish connections between concepts and look at the bigger picture.
Implications to Artificial Intelligence
After developing this strategy and promoting them in my school, I also grew more interested in artificial neural networks. Shadow Learning has transformed my understanding of the procedure of acquiring new knowledge and skills. After watching a few video series on artificial neural networks, I start to see the connection between the Shadow learning and how neural networks function.
I am planning to take CS 486 and CS 480 as one course in AI Specialization in my upper-year to gain a more mature understanding of AI to help me open up a career path in intelligent systems or web machine learning.
