Pure vs Applied Mathematics: A Conversation at the Boundary

Mon, 01 Jun 2026 00:00:00 +0000

I recently had a fascinating conversation with a pure mathematician specializing in homotopy theory about the relationship between abstract and applied mathematics. He struggled to understand why statisticians rely on results such as the Law of Large Numbers, the Central Limit Theorem, and even conventional rules of thumb, like using $30$ as a minimum sample size in certain statistical tests.

To him, these choices seemed grounded in “experience” rather than strict reasoning, and therefore susceptible to bias. While I reassured him that the foundational theorems of statistics are undeniably rigorous, he remained cautious. From his purist viewpoint, applied mathematics is risky: models can fail in practice, assumptions are never fully checked, and empirical heuristics feel uncomfortably loose.

His perspective reminded me of something one of my PhD supervisors once told me:

“Proving something theoretically doesn’t guarantee it will work in your numerical experiments.”

This tension captures the heart of the pure–applied divide: theoretical certainty versus empirical risk.

In this article, I explore the relationship between pure and applied mathematics, briefly examining the origins of applied mathematics and statistics, the motivations behind their methods, and the surprising ways abstract theory and real-world problems continually shape each other.

Applied Mathematics: Where Theory Meets Practice

Applied mathematics and statistics have always been deeply connected to real-world problems. Some key early motivations include astronomy and celestial mechanics, demography and life tables, gambling and probability games, measurement error and scientific observation, and physical sciences and engineering challenges.

Over time, these needs gave rise to statistical modeling, hypothesis testing, and a vast ecosystem of methods designed to quantify uncertainty and extract structure from data. While pure mathematics often pursued generalization and abstraction, applied mathematics pressed toward utility.

For many of us, myself included, the practical side of mathematics is what makes the field compelling. Applications make mathematical ideas accessible to a wider audience, highlight where theory succeeds or fails, motivate new mathematical discoveries, and provide reality checks for abstract assumptions.

As a student, I found the most captivating subjects were those connected to real phenomena, where mathematics revealed something useful, surprising, or intuitive about the world.

Why Statistics? Origins and Purpose

Statistics sits at the intersection of mathematics and the scientific method. It involves the collection, preparation, analysis, interpretation, and visualization of data. It also provides the tools for making inferences under uncertainty — placing it squarely on the applied side of mathematics, though firmly grounded in probability theory.

Historically, probability emerged in the 17th century from games of chance, statistical inference matured in the late 19th and early 20th centuries driven by needs in agriculture, biology, physics, and the social sciences, and modern statistics now spans machine learning, computational methods, and high-dimensional analysis.

Statistics is thus both a theoretical discipline and an engineering practice: rigorous in principle, approximate in application.

The Number 30: A Case Study in Practical Heuristics

One of my colleague’s frustrations was the widely used rule of thumb that a sample size of $30$ is “large enough” for certain statistical tests.

This convention traces, loosely, to the Central Limit Theorem (CLT). The CLT states that for sufficiently large samples drawn from a distribution with mean $\mu$ and variance $\sigma^2$, the distribution of the sample mean approaches a normal distribution.

The key issue is the phrase “sufficiently large.” It is not universally defined. The rate of convergence depends on the skewness of the original distribution, the presence of heavy tails, the desired accuracy, and the specific application.

In many practical scenarios, around $30$ observations give a reasonable approximation. But this is not guaranteed. It is a guideline formed from experience, simulation, and decades of applied practice — not a strict mathematical threshold.

This gap between rigorous asymptotic theorems and finite-sample pragmatism is precisely what can trouble pure mathematicians.

Pure Mathematics: The World of Homotopy

My colleague’s background is in homotopy theory, a branch of topology concerned with understanding when one continuous function can be continuously deformed into another. If such a deformation exists, the two functions are said to be homotopic.

Homotopy theory seeks to classify spaces and structures based on these deformation relationships. It is abstract, elegant, and profoundly removed from everyday empirical concerns.

But even this deeply theoretical domain finds its way into applications.

Homotopy in Applied Settings

The first time I encountered the word homotopy was in a very different context: Hocking et al. (2011), “Clusterpath: An Algorithm for Clustering using Convex Fusion Penalties.”

In this paper, homotopy algorithms are used to compute continuous regularization paths for convex clustering problems. The key idea: as tuning parameters vary, the solution evolves smoothly, mirroring the continuous deformations that homotopy theory formalizes.

This connection was not lost on me — it directly inspired part of the methodological foundation of my own PhD work on MGLASSO, where convex clustering via group-fusion penalties plays a central role in simultaneous network inference and multi-scale clustering.

This is one of many examples where abstract mathematical structures inspire practical computational tools. The path from pure theory to application is often indirect, but it exists.

Where Pure and Applied Mathematics Meet

The interaction between pure and applied mathematics is not a hierarchy but a conversation.

Pure mathematics builds general frameworks, focuses on proof, consistency, and abstraction, and explores ideas independent of immediate application.

Applied mathematics and statistics embrace uncertainty, approximation, and empirical risk, prioritize usefulness and interpretability, and test theory against the friction of real-world data.

The two disciplines continuously influence each other: applications inspire new theoretical questions, pure theories eventually find unexpected applications, practical challenges push for new mathematical tools, and abstractions create the foundation for robust methodologies.

Rather than separate worlds, they are complementary perspectives on the same pursuit: understanding patterns, structures, and truths — whether in thought or in nature.

The conversation that inspired this article revealed something important: both pure and applied mathematicians are striving for clarity, but in different ways. Pure mathematics seeks clarity through abstract certainty; applied mathematics seeks clarity through practical insight.

Neither approach is superior. Both are essential. And the most exciting mathematical ideas often arise precisely where these two worlds meet — as anyone working at the boundary of graphical models, optimization, and biological data knows well.

Reference

T. Hocking et al., Clusterpath: An Algorithm for Clustering using Convex Fusion Penalties, 2011.

Mathematics | Edmond Sanou