From System Engineer to Data Scientist
My career in defense R&D started in 2010 after obtaining my M.Sc. Electronics from Nanyang Technological University (NTU). Starting from FPGA programming, I progressed to system engineering, then to a team lead managing a group of system engineers. In 2019, I took the leap of faith and went for my post-graduate studies in National University of Singapore (NUS), to pursue a PhD in Computer Science — specialization in Artificial Intelligence.
Many questioned the soundness of my decision. Colleagues applauded my courage in making the switch. Some of my closest friends expressed concern about the steep learning curve due to my limited knowledge in computer science.
A 35 year old NUS Undergraduate
1 year on. I’m glad that I’ve survived the SQL, NP-hardness, Turing Test, and proceeded to complete 2 modules on Data Mining and Neural Network.
I remember my first lecture on Computational Complexity, seated among my peers, all of them at least a decade younger than me, excited that I’m back to school as a student again. Since I’m the only mid-30s around, there is no reason for me to be reserved. I will raise any question that I have, to learn the most from the lesson, and facilitate the class learning.
Then the Professor started his lecture, he briefly spoke about Space Bounded Classes, Graph Reachability, Isolation Lemma - and many other terms that, when separated, each is a word that I understood well; but when combined, hits me like a squash ball against a wall.
I started taking notes on the terms that I will enquire about after the lesson: Space Bounded Classes, Graph Reachability, … By the term Isolation Lemma, I decided that I will google and get myself up to speed by the next lesson. When I hear BBRS Algorithm, I concluded that this steep learning curve is insurmountable, at least not within the short span of 1 week.
I dropped the module. Instead, I registered for several undergraduate modules for my first semester in NUS. Together with my peers, who are now younger than most of the junior staff back at work, we pick up Python, Data Structures, Foundations of Artificial Intelligence (AI), and Machine Learning. With that, I began my first semester back in school as a student.
Why the Switch
It was a Saturday evening in 2018 when I chanced upon an article about an AI technique that produces higher resolution images from lower resolution ones. They term it super-resolution.
I initially dismissed the article as yet another technology article. It was only moments later that I realized the novelty of what was presented. If what was claimed is true (and it indeed is), that technique is effectively creating something out of nothing. That was when I started noticing AI.
Baader-Meinhof Phenomenon (also Frequency Bias) is the phenomenon where something you recently learned suddenly appears ‘everywhere’. Since the super-resolution article, I started noticing articles after articles about AI and how it will change the way we function.
Crawling (in a literal, not cybersecurity sense) through the overwhelming loads of information online. I quickly became overwhelmed by the vastness of what AI is about. From different sources, I read about Python, Pytorch (is it part of Python?), Activation Functions, Neurons, CNN, RNN, NN, ANN, Gradient Descent. It felt like I was trying to fix a 1000 pieces jigsaw puzzle in my mind; while I continue to juggle with work commitment, family duties.
One year later, that jigsaw puzzle remained marginally completed, probably only the frames of the puzzle was complete. I learnt about various terminology in AI, where does deep learning fit in AI, basics of many algorithms (especially neural network related), and the immense potential that AI have. Within the frame of the puzzle, there were multiple fragments pieced together: That we need to use gradient descent as a backward pass for the learning process, that Pytorch and Tensorflow were dominant libraries used for the algorithms, that Pytorch is a library that can be used when we code in Python language. But, as I read on, there were also more mysteries: Anaconda? Conda?
I made the decision to switch in Jan 2019. Granted a scholarship by my company, I was given the opportunity to choose, and to convince the management, on my field of study. Choosing Computer Science, with AI specialization, came almost as a default. Without much fanfare, the decision was made and I would return to school starting August 2019.
Why this Blog
One year on, while still not an expert, I’m no longer considered a novice in the area of AI. The many hours of lecture videos, Google, Coursera and Github has finally borne fruit.
This blog serves as a medium (yes, medium) for me to consolidate the knowledge gained, and that which will be gained in future. It serves three purposes:
1.As my own reference material
Tony Buzan’s “Recall After Learning Graph” shows the retention rate of an individual over time. Without review, an individual retains only 20% of new or unfamiliar information on the next day. *Try recalling the name of the organ that transport food from our throat to our stomach.*
“What I remember now, I will (almost) no longer remember in a year” – myself
If I can spend minutes trying to find my iPhone after each shower, only to find it (almost always) lying on the sofa; I reasoned that the only logically next step is to keep a record of what I’ve learnt. A record that serves as my cue-cards for future stages in life.
2.As my answer to friends’ questions
The expression of surprise upon knowing my return to life as a student often come with questions like:
- “What made you decide on the switch?”,
- “Is AI really that powerful? Is it difficult to learn? How should I start?”,
- “I’m in charge of digital transformation. How should AI be applied?”, and
- “I hear that Deep Learning is able to automatically derive features for classification. Is it true?”
Being able to devote quality time to go back to school is a luxury that I’m extremely thankful for. This same luxury is not extended to all who wants to learn. This blog is my way of contributing back to my social circle, to provide some pointers, direction and hopefully answers to the questions that they may have.
3.As my contribution back to the AI community
The extensive sharing among the AI community made my learning journey so much easier. For every question, there will be an answer somewhere, the quality of the answer a function of how thorough is our search. For every error message, there will be a stackoverflow post to address the issue, or at least to suggest a workaround.
Reciprocity: A social norm of responding to a positive action with another positive action
I intend to reciprocate through this blog. Written in a not-so-technical manner, this blog is targeted for readers who will prefer an intuitive, gentle guide to the knowledge in AI. Readers who are experts in the AI are strongly welcomed to provide feedback and suggestions.
This post signals the start of my journey as a writer. Having not written non-technical material for at least a decade, there are many ways that I will be able to improve all. I will like to solicit your feedback to improve on my writing.
I will also like to connect with like-minded readers who will like to share and explore ideas in the field of AI together.
Thank you.
By the way, that organ that transport food from our throat to our stomach is the “oesophagus”.
