Does Data Science relate to Machine Learning and Deep Learning? How?

Vijay Vignesh
5 min readFeb 23, 2020
Image representation of various tools and languages of data science

Over 98% of the customers who indulged in a mobile phone purchase online is likely to buy a case for it as his next purchase, says data. Today, the online retailers bring just those right recommendations of mobile phone cases right after a mobile phone purchase and this purchase pattern prediction is a very simple case example of extensive data science usage in action.

While they say that data is the new oil, unless it is refined it cannot be used and remains invaluable. That’s exactly what data science is all about.

In definition, data science is the science of collecting, storing, processing, describing and modeling data which helps extraction of useful insights and information.

If this is what data science is all about, where does it even relate to Machine Learning (ML) and Deep Learning (DL)? Well, let’s side-track a bit and also see what Artificial Intelligence (AI) and the fuss around it is all about.

Artificial Intelligence or AI as been around at least for some time but has lately gained popularity over the technological advancements and capabilities unleashed in the computing powers. AI is about building systems or agents that demonstrate “intelligence”. It is the ability of the computers of these machines to understand data, learn from data and make patterns or inferences from the data which could otherwise be very difficult for humans. AI also enables these computers to learn from these patterns and “fine-tune” their knowledge based on new inputs that were not part of data used for training these machines.

In essence, the tasks that constitute AI are as follows:

  • Problem solving
  • Knowledge representation & reasoning
  • Decision making
  • Communication, Perception and Actuation.

While problem solving and knowledge representation & reasoning deal with search algorithms and proportional & first order logic respectively, it is in the decision making and communication, perception and actuation steps where AI intersects hugely with data science.

Decision making in AI back in the day was predominantly enabled by expert systems. These are designed to solve complex problems by reasoning through bodies of knowledge, represented mainly by a set of written rules rather than through conventional procedural code. Writing these rules were exhaustive and complicated, inexpressible and unknown that served as their limitations. Thus dawned the era of data driven decision making.

Data driven decision making framework

Utilizing data to develop insights and arrive at conclusions is essentially done in three different approaches namely:

  • Machine Learning
  • Deep Learning
  • Reinforcement Learning

These approaches are a subset of AI in itself and data science processes of collecting, storing, processing, describing and modeling data are used vividly in the above said approaches. It is this data driven part of AI that intersects with the world of data sciences.

It is this data driven part of AI that intersects with the world of data sciences.

Machine Learning (ML) can be said as one of the implementations of AI. As the name goes, ML is used in situations where we want the machines to learn from huge amounts of data and apply that knowledge on new pieces of data that streams into the system. Some of the applications of ML that apply data sciences are as follows:

  • Virtual Assistants (Eg. Siri, Alexa, Google Assistant etc.) where data science processes of collecting and processing information based on customer’s previous involvement with these assistants is performed.
  • Commuting predictions where traffic updates are thrown ahead of the journey and online transportation networks where ride fares and surge pricing models are decided based on the route and rider demand.
  • Social media services that include tools such as People You May Know, Face Recognition, Similar pins etc.
  • E-Mail spam and malware filtering based on user inputs and multi-layer decision tree coding pattern.
  • Online customer support that features a variety of NLP programmed chatbots at the front end of customer service.
  • Product recommendations based on user’s search interests and shopping history.

Deep Learning (DL) on the other hand is a further subset and more so an advancement of Machine Learning. Deep Learning sets foot in places where the data sets are vastly huge and accuracy is of paramount importance. It is also important to note that deep learning requires much powerful hardware to run on (mostly GPUs are used), it takes significantly more time to train your models, and it is generally more difficult to implement compared to ML.

Applications of DL is spread across various fields, notable ones are the self driving technology of autonomous cars that is fast-developing in today’s world. Some of the other real world examples include Netflix’s program recommendations based on inputs from complex algorithms of user engagement and search history.

Reinforcement learning deals with the actuation part of AI where the reinforcement agent decides what to do to perform the given task. In the absence of a training dataset, the agent learns from its experience and applies the knowledge to perform its task. This approach seeks applications in robotics for industrial automation and in large environments where model of the environment is available but solution is not known and the only way to resolve is by interacting and learning from the environment.

All of these apply data science at its core to collect, process, describe and model data based on whose results decision making is done through one of Machine Learning or Deep Learning approaches as the application deems it to be. In other words, data science stands for an all-inclusive term that consists of aspects of ML & DL for functionality. Indeed, ML, DL and AI are a part and parcel of data science.

--

--