liveBook · Manning

Problem definition

As explained in Machine Learning System Design

A concise written description of a problem, including its relevant goals, context, and important characteristics.

A problem definition is a concise description of the organizational problem, its context, and the outcome that matters before choosing an ML solution. Start with a helicopter-level statement that nontechnical decision-makers can understand, then add relevant goals, constraints, and characteristics. section 2.2

Use an inverted-pyramid process:

  • State the broad problem in plain language, such as fraud attacks, poor product margins, excessive manual tuning, or weak user engagement. section 2.2
  • Move from general understanding toward specific nuances through leading questions. section 2.2
  • Use the resulting understanding to identify which information and data are most valuable for project launch. section 2.2

From business concern to investigation

“There are fraudsters in our mobile app who try to attack our legitimate users” is a useful starting point because it communicates the business concern without prematurely prescribing an algorithm or architecture. section 2.2

Figure 2.3 The inverted pyramid scheme is an approach we recommend for gathering data required for a successful project launch.
figure

The inverted-pyramid diagram shows how to gather project-launch information by descending from broad problem understanding into detailed nuances. section 2.2

How Machine Learning System Design covers it

How deep it goes
Builds a top-down questioning process from helicopter-level statements to problem nuances and valuable data.
Examples
Uses fraud attacks, weak margins, manual tuning, and low user engagement as workplace problem statements.
What it assumes
Assumes broad domain understanding but does not require deep domain expertise or prior algorithm knowledge.
What you can do after
Evaluate/understand an ML initiative's broad goal and identify the information needed before implementation begins.
How it connects
Connects problem understanding with domain knowledge, data gathering, and ML system design.

Also explained in

Compare how 6 books explain problem definition →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning System Design. The section links point to the parts of the book it draws on.