Fundamentals of AI and machine learning

Posted on 08/26/2020 05:40:34

In the next decade, pretty much everyone will both use and misuse AI/ML technologies. While the technology itself has been around for a long time, the buzz has intensified in the recent past. This has made everyone sit up and talk about how they are using artificial intelligence in their business (just like the ‘Digital Transformation’ buzz before it). I would guess that there are very few pitch-decks that miss out on using AI/ML as part of their strategy to impress investors. However, it is also important to understand the fundamentals of AI and Machine Learning in order to leverage them in our solutions in the right way.

AI/ML technologies are making a key difference in many industries and are offering radical innovations. The key (I think) is to remove the buzz and hype and focus on the impact that these technologies can have on the quality of life around us. To do that, let us begin by looking at what is Artificial Intelligence and its underlying concepts of machine learning.

Put very simply, machine learning is a set of statistical techniques that identifies patterns in data and uses the same to predict future instances of the pattern. There are a few broad outcomes that machine learning can deal with. As explained in this post, there are typically 5 questions that data science answers. These are

  • Is this A or B – classification
  • Is this different – anomaly detection
  • How much or how many – regression, forecasting
  • How are they organized – clustering, dimensional reduction
  • What should I do – prediction, reinforcement

Of course many other questions can be posed which are a variation or a combination of these leading us to more complex implementations. All the different algorithms can also be classified based on how they are trained – supervised, semi-supervised, unsupervised and reinforced. Some of the examples of machine learning implementations include the following :

  1. Supervised learning – where data sets with labelled data and desired outputs are provided based on which the algorithm learns to make future predictions
  2. Semi-supervised learning – where the datasets have both labelled and unlabelled data and the algorithm learns to self-label the data for its operations.
  3. UnSupervised learning – where there are no specific desired outputs, but the algorithm is left to classify large sets of data into logical groups based on identified patterns.
  4. Reinforcement learning – where the algorithm is provided with actions, parameters and end values and it learns and adapts different possibilities to wards and defined end goals.

Let us put all these together and start exploring the tools and the algorithms around AI and Machine Learning.

More Posts
Fundamentals of AI and machine learning
Applying Intelligence in the real world