What is a Neural Network?
Table of contents
Artificial Intelligence?
Artificial intelligence describes the field where computers are programmed to learn and that neural network are a subfield of the former. In fact, in contrast to traditional programming where a computer is instructed what to do with a specific input, in artificial intelligence, the computer is instructed to learn the relationship between an input and an expected output. Eventually, the computer is able to interpret an input it has never seen before and give an output.
What Are Objects?
Objects are what is being analyzed by the neural network. In our case, a neural network will be used to identify the mitosis stage of multiple cells in a picture; the cells are the objects.
How Do Computers Learn?
Here is a very simplified example of how a neural network that identifies cats and dogs as objects is trained1:
- The computer is presented with a series of dog images and a series of cat images.
- Cats and dogs are labeled on each image
- The computer learns what constitutes a dog and a cat in terms of numbers by analyzing the images
- Imagine each image as black and white, and as being constituted of pixels each with a value between 0 and 1; 0 is black and 1 is white.2 The computer, in a way, tries to determine what arrangement of numerical values constitues a dog versus a cat.
- Once trained, the computer should be able to identify a dog or a cat in a picture even if the latter has never been seen before by the computer.