20 Nov 2020

My ground zero on machine learning

Challenges I have on my path of learning and applying machine learning from school and work are <\br>

  1. clarify/reframe the problem
  2. easily understand the data
  3. quickly coming out/implementing/testing the new idea
  4. not actively invovled in a machine learning community where people learn from each other.

I have considered taking a research route(phd) by which I can relearn some foundamental knowledge and working on a few advanced topics within acedemia enviroment. However, when I rethink about this crazy plan. It’s probably not worthy for me. Yes, it definely will helps me with getting the dream job. However, I have to quit my fulltime job and be away from family for years. Not even to mention the amount of prework I need to finish in order to get into a credited program. My current goal is getting better in applying machine learning, and competing in kaggle makes perfct sense for me. it removes a lot of barriers and creates a ‘safe’ environment for people to learn/fail.<br>

Goal: Be Kaggle Expert in Competition and (Notebook or Discussion)

Timeline: 6 months(05/20/2021)

Currenty where I am:

  • Have fundamental understanding in statistics and machine learning.
  • Enjoy coding with different language but struggle to quickly implement ideas
  • Like to do exploratory analysis on data and get some basic understanding of the data

Problem I have:

  • Persistency, I have started kaggle very long time, one issue standing out is that I’m easily got distracted by something else, like defeat from the last idea tried or just bad mood from life.
  • Scientics strategy/plan(probably there isn’t one), I usually go with random.

Rules I must follow:

  1. Focus on one competition at a time.
  2. There must be a time placehold for kaggle every week in my calendar.
  3. Post update every week on the blog.
  4. Have fun in applying idea and seeing results regardless good or bad.

Current Competition:

  • Predict futures sales
  • Timeline: 2 months(01/20/2021)
  • Current progress: Baseline model with 0.96 mre.
  • Next step: read https://www.kaggle.com/c/competitive-data-science-predict-future-sales/discussion/74835

Tags: