Well, tomorrow and accident, you’ll never know which comes first.
The reason or the excuse why I didn’t post regularly is that I am now taking an internship in a company. The working hours are from 10 am to 9 pm Monday to Friday. Technically, I work for 8 hours per day because we have a two-hour break in the afternoon and a one-hour break in the evening. However, I do think that the intensity of this schedule is comparable to the 996 working hour system. Whenever I get back home after work, it is nearly 9:30 pm. Eight hours of work have already cost me too much concentration, and I simply don’t want to do anything but lie on my couch and watch some Youtube videos. This is the only time for me to relax, but on the other hand, once I reflect on myself and think about it, it is actually the only period for me to improve myself. I would love to get up early to study, but everyone knows that this is hard.
At the same time, this internship opportunity indeed pushed me really hard. I was recruited like a blank paper, and now this blank paper turns into at least a chapter of a book. On the first day at work, I was directly assigned to a task from scratch. The manager gave me some datasets and asked me to build a deep learning neural network by myself to solve the problem. I indeed learned some basics of neural networks and optimization from a mathematical point of view, but I am still essentially a student studying pure mathematics! It is like asking a bartender to make a boba milk tea for you. Fortunately, it is not as hard as I thought.
I used two weeks to learn a list of things:
- Complete two courses in Coursera.org
- Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning by DEEPLEARNING.AI
- Advanced Computer Vision with TensorFlow by DEEPLEARNING.AI
- Four chapters in the book Deep learning
- Chapter 5: Machine Learning Basics
- Chapter 6: Deep Feedforward Networks
- Chapter 7: Regularization for Deep Learning
- Chapter 9: Convolutional Networks
- The use of Jupyter Notebook, Linux server, Tensorflow.
I firmly believe that the materials above are even easier than undergraduate analysis. However, as deep learning being an experimental science subject, its difficulty lies in its experimental performance and applications but not in a theoretical form. I used two weeks to learn and successfully constructed a working network, but its performance is not going very well anywhere expected. Initially, the neural network did have some good improvements after the two days, but I found a fatal mistake made by me, so I had to get rid of all my experiments before.
In my opinion, if I was a computer science student, I may perform better during this period. This is a direct consequence of the fact that I spent most of my time on coding and debugging rather than thinking or any other things related to mathematics. I consider this as a problem because I didn’t utilize my strength. From another aspect, I did get to know the industry and a more practical side behind my mathematical background. However, as I have already noticed, the learning curve of this internship may suffer from the effect of diminishing returns soon. At that time, I will have to decide whether it is worth staying in this company or search for a better place.
Coming back to the reading plan. As I explained before, I didn’t adjust myself enough to make use of the time available in the morning and in the evening. Meanwhile, I highly douted that it is a common situation happened in the industry. Therefore, in order to improve myself and surpass others, I need to do what others can’t do: use this time starting from yesterday. It is going to be hard, but I hope I can persist.
Leave a Reply