Have you ever encountered a difficulty of setting learning a new skill as a top priority? In the data science world, learning/reading/researching is a daily habit in the industry. Since we’ve left school for years, it’s very common to see senior data scientists are completing 3–4 advanced online courses in each year. Updating knowledge is a key to success in machine learning, and it’s serious business to outgrow the industry standard. I don’t find myself being educated by a young graduate from college, because it shows a sign of incompetence. I have found that prioritizing task and time becomes a must-learn-skill, so a topic for today is to demonstrate the matrix which improved effectiveness in resource and time allocations.
Yesterday, a colleague of mines forwarded me a business article written by Harvard Business Review. It was a concise article, which taught me a few important lessons in managing data science skills, especially for the future when Microsoft created ML buttons on excel sheet. For example, how I arranged a priority between daily work and learning demanded tangible knowledge. I would not conclude the article the entire article into few sentences. However, I would apply the matrix to demonstrate the idea and apply it to my daily work and learning. I believed that the matrix helped me to achieve two manners: time management and learning improvement.
According to the matrix, I want to articulate the definition of “very useful” to make it crystal clear in my content. Usefulness is defined by how much one needs in order to achieve a perfect job. I’ve listed a couple of “very useful” candidates; such as pushing Mxnet framework on the production, reading NLP research papers, and finishing up advanced courses in statistics. In my case, I need to gain a deep understanding of the Mxnet framework that can lead to better performance and cost reduction. The Mxnet framework isn’t as popular as Tensorflow, but it’s a pythonic way. It’s feasible to learn, and plus reduce the workload on the production, by our test, over 4 times. Although Mxnet community is relatively smaller than Tensorflow’s, it’s a great framework I still highly recommend it to new data scientists. I score “usefulness” in the decreasing order: Mxnet Framework, NLP research paper, and advanced stats course. For my team, the priority is different. Putt, a senior data scientist, had to decide between a deploying a Thai chatbot on Google Dialogflow or improve the performance of the language model. Both tasks for Putt are equally important to our team. Since no right answer is given, we might need to consider another important factor: time. Some task is equally important to another, but require so much time to learn. I have to enroll one course in statistics, stochastic process, and it requires at least 4–5 hours for assignments. Since distribution is built on top of the statistics, it’s very useful to learn, but no time. I think that the matrix has helped me so much to just “plan” it on my schedule, and leave it until the end of the month to learn.
One hour, one day, or one week. How much time does the task require? One of the techniques I founded useful for time management is to partition one day into 3-time slots, each has 8 hours. For example, I put my time for sleep and entertainment into one slot. I usually measure how much time I sleep, and I sleep about 7 hours a day and enjoy the entertainment (mostly video games) for 1 hour. I have one slot dedicatedly for work (8 hours per day). The last slot is scheduled for eating, socializing, researching, and reading data science. Therefore, I have limited the time for “usefulness” for just 3–4 hours a day. I’m sure that everyone will share a different time slot, but more importantly, I can predict roughly how much time I can spend on each task. However, I also need to consider “Browse” of each task or how’s easy it’s easy to Google and find information out. Since the strong Googling skill requires, browse is the same. It should tell you how much time one really needs to make a learning happens. Now, I have a list of prioritized task and amount of time I can spend on, then the learning becomes feasible. I choose to learn about statistics as a “plan” for my last quarter of this year and push Mxnet framework as the top priority as it requires less time and gives a room for “browse.” I’ve found out that it’s useless to learning redundant courses like deep learning in Coursera or improving my understanding of data visualisation.
In conclusion, the 2x2 matrix may not fit all lifestyle, so there exist unexpected incidents such as family issue and urgent corporate meeting. However, I encourage everyone to try out the matrix to help you prioritize time better. In the future, I think data science is about a full set of computational skill (math and statistics) with minimal coding. I could be wrong, but it’s good to be prepared for the future.
Drop us a line and we will get back to you