A Neural Network is Just a Glorified Math Equation

We’ll discuss loss, weights, saving a model, node, activation layer, and a lot more in such simple terms that you’ll think that a neural network is just a glorified math equation.

Image by author | Basic terms and symbols in a neural network

There’s a lot to unpack here. Let’s go through them one by one.

Let’s assume that you spied on your “friend” at work and created a dataset of the number of cups of coffee he gets in a day, along with a few other parameters you think might affect the number mentioned above of cups of coffee.

Image by author | Sample dataset

The above dataset is for five days, and you want to know how many cups of coffee he’ll drink on a day when he checks in at 10, closes two tickets and attends three calls. You can make an educated guess based on the 5 data points you have collected. But imagine if you had a year’s worth of data (Get ready to call a lawyer) and want to predict the same.

Humans are good at many things, but going through many data points and coming up with a pattern is not one of them. That’s why we have computers.

So in the above data set, Check-in time, tickets closed, and Calls/ Meetings are your input columns. Also called X. And Cups of coffee…