qwertyboss

• “We don’t choose our earliest habits, we imitate them. We follow the script handed down by our friends and family, our church or school, our local community and society at large. In many ways, these social norms are the invisible rules that guide your behaviour each day.”

• “We imitate the habits of three groups in particular:
The close.
The many.
The powerful.”

• The human mind knows how to get along with others. It wants to get along with others. This is our natural mode. You can override it—you can choose to ignore the group or to stop caring what other people think—but it takes work. Running against your culture requires extra effort.

• Once we fit in, we start looking for ways to stand out. This is one reason we care so much about the habits of highly effective people. Many of our daily habits are imitations of people we admire or envy.

Anyone knows about any platform for accountability group like MakerLog but for posting your learning/habits/side projects in public?

A development analogy would be
- logging level "INFO" is Twitter/Linkedin
- I am looking for a logging level "DEBUG" platform

4 / 20 hours
Training on 100 GB of data is turning out to be difficult than expected.

Found an interesting resource for a few algorithms -> https://dafriedman97.github.io/mlbook/content/c1/construction.html

3 / 20 hours
Current: Baseline model
Next: CV strategy

> Temptation building
• Desire is the engine that drives behaviour. Every action is taken because of the anticipation that precedes it. It is the craving that leads to the response. It is as if the brain is saying, “See! I knew I was right. Don’t forget to repeat this action next time.

• Habits are dopamine driver feedback. Every behaviour that is habit-forming - taking drugs, eating junk, playing video games, watching porn, browsing social media is associate with a high level of dopamine. And whenever dopamine rises, so does your motivation to act.”

• Example: Use The habit stacking + temptation bundling formula.

“After [CURRENT HABIT WHICH I DO], I will [HABIT I NEED TO DO].
After [HABIT I NEED TO DO], I will [HABIT I WANT TO DO].”

“If you want to watch sports, but you need to make sales calls:
After I get back from my lunch break, I will call three potential clients (need).
After I call three potential clients, I will check ESPN (want).”

I had late meetings so had to do cheat workouts in between the calls, only 1 set per muscle group.

2 / 20 hours done

Setting up the data pipeline and experimenting the right validation strategy.

> Cue-induced wanting

“The people with the best self-control are typically the ones who need to use it the least”

“People who appear to have tremendous self-control, it turns out those individuals aren’t all that different from those who are struggling. Instead, “disciplined” people are better at structuring their lives in a way that does not require heroic willpower and self-discipline”

Note: It is important to break the symmetry while weight initialization or else all the neurons will be activated in the same fashion.

- Data Cleanup: Encode features, Standardise features, etc.
This is necessary because the sigmoid function squashes really small and really large inputs.

- Weight Initialization
If the weight is too small it will bring you to the vanishing gradient problem if the weight is too large it will bring you to the exploding gradient problem. Instead, initialize the weight with a specific range.

I didn't have enough energy so did a cheat workout with only 1 set per muscle group.

This is an experimental challenge I will be only working a total of 20 hours on this project (excluding model training) and let's see how much progress can I make.

Competition link: https://www.kaggle.com/c/landmark-recognition-2021/overview/evaluation

Last Date: 2nd October

Create a list of basic ML topics which you want to learn the theory, code and maths.