Machine Learning

Machine learning in nutshell

          My dear friend “Google search” was, as usual, very useful in giving me tons of material on GenAI. However, most of the articles didn’t explain what ‘machine learning’ is. Fortunately I had already done the course on ML so I knew what it is :-)

And that is what prompted me to write a small post on how machine learns from data.

 Let’s say, five years back I started a shop to sell chips packets. So, I have sales data of last 5 years available with me.

Formula that gives the output closer to an expected output as in above table" is as below

Sales for month X = (X + Y) Z / C P (hypothetical formula only :-) )

Where,

X = Number indicating a month (1-12)

Y = Number of holidays in the month X

Z = Annual % population increase

P = Current population

C = Some constant (4.765)

For last 5 years I compared the result given by above formula with the actual sale that has happened. It looks to give me the 70% accurate result. And hence I am ok with the value of constant C (4.765)

As the data for 2024 also became available, I adjusted the value of C to 3.901 to give me 74% accuracy

As the data for 2025 also became available, I adjusted the value of C to 3.892 to give me 76% accuracy

As the data for 2026 also became available, I adjusted the value of C to 3.492 to give me 80% accuracy

As more and more data became available, I was in a better position to adjust the value of “C”

So, here I am learning to predict more accurately as more and more data is becoming available.

If you replace I with machine / software program, it becomes, machine is learning.  

Only difference is software programs (neural networks) written can adjust the value of constant C automatically with more and more data availability.