AI: regression v classification & supervised v unsupervised

The learning is the key element of AI. A simple way of categorising is classification and regression. You decide which one it is by what you are trying to predict. If you are try to predict whether it is a cat or a dog from a bunch of photographs then it is classification. 

If you are trying to predict a continuous value e.g. how the temperature, water and light affects plant growth. Then it is regression. You will input the temperature, watering pattern and light after gathering as much data as possible in the first place and you should be able to predict how well the plant will grow. 

Also the learning can be split into two parts, unsupervised and supervised learning. This is more relevant to the type of data that you have. For instance if you have two sets of images of dogs and cats. They may be labelled as either a picture of a cat or a dog or just the pictures with no labels at all. 

In supervised learning, the AI has labelled images on which to be trained. This is obviously very efficient and effective but the downside is that someone has to label every image correctly. With cats and dogs that should’ve mean any errors but if you were doing breeds of dogs and cats there will also be some errors. 

Unsupervised learning is where the AI is presented with those images without labels and will try to group them according to the features that it can recognise. This will need a lot of data (images). There is always the chance that it will get it wrong but in some cases there isn’t always an obvious right answer anyway so you might just be looking for patterns. 

So in summary you have to decide whether your model is classification or regression and then whether you are going to use labelled data (supervised) or unlabelled data (unsupervised). Obviously the above examples are very simple but non the less these are just some of the questions (and there are a lot more) that you will need to ask yourself when developing AI. 

The skill required is more of an art than science. Often there isn’t a single way of tackling a problem and in some instances a simpler algorithm is all that is needed especially if time and resources are limited. So AI isn’t always the solution. People head down that path without thinking when there are cheaper and simpler solutions available. 

Saying that it is tool in your tool box and having it there is always going to stand you in good stead. The concepts may be unfamiliar but they aren’t as difficult as you might think.