Choosing the right techniques for your AI/ML experiments

Early this morning, I was sitting on the balcony, thinking about choosing the topic for today. The atmosphere was very pleasant, trees were swaying with the cool breeze, and birds were chirping and jumping around. Suddenly I noticed a light-colored butterfly flying around swiftly from place to place. I don’t know what it was looking for, but it continued to fly around randomly till it decided to fly away, and it disappeared as suddenly as it showed up.

I kept wondering about why exactly I noticed and remembered the butterfly amongst so many other things around.

Was it the color?

Was it the sudden movement?

Was it the sound?

Was it the flipping of wings?

Was it because of the short stay?

Was it because of the sudden appearance and disappearance?

Was it because of my interest in butterflies?

Was it because of the small size?

Would I even notice it, if I was not present there at the right time?

What exactly do we look for when we analyze the data to figure out something meaningful about it? Do we really need to know what to look for? And what is a good time to look at it?

Had I been watching a tree with that intensity instead of the butterfly, the observations would have been entirely different, and so are my thoughts – the outcome.

Often the outcomes we see are prominently affected by the things we focus on, or in terms of machine learning, we can say the data & approach we take.

This is why it is important In machine learning experiments, to be careful about the techniques we are deploying based on –

What are we trying to learn?

• How much and what type of data is available?
• How much processing power and time can be used?
• When do we need the results? Real-time? Or on a schedule?

How do you ensure that you have chosen the right techniques for your AI/ML experiments?