Executive Briefing: A new taxonomy of machine learning

Description

Machine learning has taken hold of our imaginations as a general technique for solving difficult problems and offering disruptive opportunities across industries. While many of the models and approaches have been known and studied since the 1950s, the advent of next-gen data platforms over the past decade has finally made these data-intensive applications commercially viable. From that perspective, machine learning, as an enterprise pursuit, is in its infancy.

Along with any emerging technology comes a huge amount of confusion and noise. The total investments that companies both large and small are funneling into machine learning is well into the billions of dollars. However, for many industry observers, it has become almost impossible to distinguish between various software solutions that leverage machine learning with respect to impact and potential competitive advantage gains.

There are a number of interesting classifications of machine learning approaches: some researchers focus on the different types of learning or training algorithms; some focus on the types of mathematical models or functions that are being learned; and some focus on the computational complexity of the training algorithms (an area of study known as computational learning theory). However, none of these categorizations are helpful in determining the impact a machine learning solution may have on an industry or how much of a competitive advantage an organization could gain by implementing that solution. Simply put, some machine learning approaches are hugely disruptive and will precipitate the next generation of killer apps, while others are not that technically difficult or novel and will not be impactful.

Rachel Silver shares a new taxonomy of machine learning approaches that distinguishes between those that are providing enormous competitive advantage and those that represent merely small, incremental improvements on existing analytical tools and details a framework for evaluating ML approaches on several dimensions of complexity, including:

• The amount of data required (such as for training)
• The computational complexity of the training algorithm
• Real-time streaming requirements (versus just batch computing)
• Data throughput for the deployed model to process

Rachel explores examples of how to apply this framework to real-world machine learning approaches and highlights the technical requirements of supporting the most disruptive examples of ML solutions.