The reason you should STOP checking time when studying

Machine learning in daily life [1]

If you count the hours you spend working to achieve your future goals, you’ll likely remain as a worker receiving hourly pay!

 In recent years  I majored in Korean Literature, with a double major in Philosophy. When I was in my third year of University, I began to feel a detachment between myself and a life in which I’m preparing to get a job. I started to feel the harsh reality of the working world, which Conan O’Brien talked about at the Dartmouth college Graduation ceremony, “the only place to get a job is Ancient Greece.”

If your child majored in fine arts or philosophy, you have a good reason to be worried.

 I followed a strange but popular advice: “choose the area first, and prepare to get a job”, so I chose “marketing” and began preparing to get a job. While doing so, I accidentally got into data analysis, which is one of the sectors within the marketing area, and one day I found myself studying machine learning, statistics and writing python. That’s how I started my one and a half years’ journey preparing to enter a lab, researching Artificial Intelligence via a Computer Science master’s degree, which was quite a random choice for me. In the meantime, studying AI from basic calculus to deep learning models, I frequently experienced some daily insights into humanity from an engineering and technological perspective. In this blog category “Technology feat. Liberal Arts”, I will share my thoughts each week about the insights into humanity, derived from technology. This first post is about the importance of choosing a ‘metric’ to evaluate the prediction model in machine learning. This led me to think about my own daily life experience as well, “when I train my brain by studying, how do I evaluate the outcome?”

 In terms of Deep learning, of course, large and manageable data sets are highly important. However, there is another important matter which is choosing the way to evaluate the model, meaning, which metric to use in evaluating the data when training a model. Usually, it is fine to use the basic metric ‘Accuracy’ which is simply the percentage of correctly predicted cases divided by all cases.

 This would be the worst option, however, when it comes to the case of predicting cancer patients, who are less than 1%. Due to the model having more than 99% ‘Accuracy’ when repeating the prediction “not cancer”, the model will be trained to parrot that predictive pattern by the process of mathematical optimization (the goal of which is to achieve a higher Accuracy). So in this case false negative happens, which means classifying cancer patients as healthy. Dealing with the extremely skewed / unbalanced data, therefore, we need to use ‘Recall rather than ‘Accuracy’, (In this case, the ratio of positive and negative patients is skewed). ‘Recall’ evaluates ‘how many predictions were correct among only the cancer patients, rather than among all cases. In other words, ‘Recall’ may create errors classifying some healthy people as being cancer patients. The focus of this model is on not missing any cancer patients. In a hospital setting, it’s obvious which model is more needed; a model which makes some false alarms but is always able to distinguish those patients with cancer, rather than a model with amazingly high ‘Accuracy’, but never distinguishes cancer patients.

 Deep learning (sometimes referred to as a deep artificial neural network) was created based on  how the neural network of a human brain operates. It is a type of machine learning, stacking ‘neuron layers’ and putting data through it (forward propagation), and training the model with ‘back propagation’ which is learning from reverse circulation feedback. With respect to the huge success of deep learning, neuroscience itself is also in the limelight as of recent. One approach is that deep learning structures can be helpful in understanding the human brain reversely. There is a continuous process of training taking place in our brains, so it is possible to affect this process through the use of ‘meta-cognition’ or, “thinking about thinking”. Specifically when it comes to studying and learning, the decisive factor is the same as machine learning- which metric should I use’.

 It’s very easy to see how influential the metric is in our daily lives. If you check your weight every day, you’ll likely lose measurable weight, but if you monitor what you eat and take photos of yourself every day, you’ll get a visibly healthy and slimmer body.

 Similarly, in the matter of the human brain, choosing which way to evaluate your learning process has a huge impact on the outcome of the training. T. Harv Eker wrote in his book Secrets of the Millionaire Mind, that self-made millionaires prefer to choose being paid by their results rather than being paid by each hour that they work. In most cases, being paid by hour was chosen due to the fear of the individual’s value being tested within the harsh market. If you count the hours you spend working to achieve your future goals, you’ll likely remain as a worker receiving hourly pay. Although it’s also beneficial to treasure your time, if your focus is too heavily placed on the ‘time spent’ as the metric by which to evaluate the amount of self-improvement, your improvement model will evolve to result in a higher amount of working hours.

 However, if you focus on the output of ‘your volume of achievements / how much work you have completed’ as the metric, your learning model will result in achieving a better outcome, regardless of time constraints. If you want to check the amount of effort you have already put in and adjust your plan for the future, don’t be satisfied by recording by time, for example ‘studied 6 hours’. Instead, you should evaluate the exact outcome you’ve achieved by doing so, in terms of how much you have completed, measured by both qualitatively and by volume. Maybe you could have done it in a much shorter time if you paid more attention, maybe you could achieve more goals in future by working harder for shorter periods of time. By utilising this metric, you can precisely evaluate the effort’s exact impact on your initial specific goal. 

How to use the outcome metric

I listed below a couple of ways you can utilise this method, to help you get started in making the transition from recording your time to recording your effort. I hope these will be beneficial in you beginning to feel satisfied with your working efforts!! :) 

1. Record the amount of work you completed
ex. Today I wrote a 4 page essay for my class (NOT Today I wrote for 6 hours)

2. Record your level of focus on your work
ex. 7 /10 - (this is one way of doing it, but feel free to use whichever is best for you!)

3. Record how your studying has improved your ability in regards to achieving your goal
ex. I read an article about a nice metric - I can now apply this to my daily life! Yey!

I hope you enjoyed this article as much as I did, and I hope you come back again for more!! :) 

댓글 쓰기

5 댓글

  1. 답글
    1. Thank you very much! I'll keep posting every week. Hope you enjoy other posts as well :)

  2. It would be really helpful for anyone. Great insight, dude.
    Even I dont have any idea of machine learning tho, It was easy to understand and interesting.
    At the end of essay, I could find it cute as well where Emily gives some examples :D I'll look forward to the next.

    1. Wow thank you very much for the best comment! :D
      I'll keep posting this series and others, hope you enjoy them as well. Really glad that you liked this.