What is a Tensor (Primary Concepts)

1. Imagine a Box of Numbers

  • A single number is like a dot → called a 0D tensor (scalar)
  • A list of numbers is like a line → called a 1D tensor (vector)
  • A grid of numbers (rows and columns) is like a sheet of paper → a 2D tensor (matrix)
  • A bunch of sheets stacked (like a book or cube) → 3D tensor
  • And it can go even higher (4D, 5D…) for complex data

In a Neural Network:

When data (like images, sentences, or numbers) goes into a neural network, it’s always converted into these tensors — because that’s how the network “sees” the world.

Real-life Examples:

  1. mage: A colored photo is a 3D tensor

    height × width × color channels (like red, green, blue)

  2. Text: A sentence can be converted to a 2D tensor

    number of words × word features (like embedding vectors)

  3. Table of numbers (like Excel rows): That’s a 2D tensor

    rows = data points, columns = features

Why Tensors Matter in Neural Networks?

  • Neural networks do math on these tensors.
  • The layers of the network are like machines that twist, stretch, and squeeze these tensors to extract meaning.

Tensor (Primary Concepts) – Tensor example with Simple Python