Indirect dating with mixture density networks

<<< All talks

Indirect dating with mixture density networks

When	Wednesday, September 24, 2025, 11:00 AM – 12:00 PM (MT)
Where	LNCO 1100

Abstract

It is an astonishing fact about the world today that no one can say precisely how many people actually live on our planet, even though it would presumably be useful to have such information to plan for climate change and disaster risk management, among other things. Luckily for us, demographers and spatial data scientists are keenly aware of this problem and have devised sophisticated methods for interpolating population in these regions based on their built area, typically measured using remote sensing technology.

As it happens, this is exactly the reasoning applied by archaeologists seeking to reconstruct population sizes in the past. Unfortunately, archaeology faces an additional challenge here, since the archaeological record is a palimpsest of built area, representing continuous human settlement over decades, centuries, and sometimes even millennia. So, reconstructing population sizes across a region of interest requires that archaeologists also develop a chronology for that region at the same time.

A region’s chronology can be represented by a probability density function, p(t), with well-dated archaeological materials – like tree-rings and radiocarbon samples - assumed to be random draws from that distribution. Here, we propose to estimate p using a deep-learning extension to the mixture model known as a Mixture Density Network. With this model, we condition the chronology on diagnostic data X, treating mixture parameters as unknown functions of X that can be estimated using a simple multilayer perceptron. Specifically, we use the density of X in the area around each sampled date to estimate the mixture parameters. This allows us to interpolate dates at under-sampled sites and to build a better representation of the regional chronology, one that is based, in theory, on the totality of the archaeological record.

As an example, we fit an MDN to the distribution of tree-rings in the Mesa Verde region of southwestern Colorado using the spatial distribution of ceramics to estimate mixture parameters. An important ancillary goal of this research is to develop software tools scientists can use to train MDNs on their own data without also having to learn the intricacies of AI development and testing.

Speaker

Tags: geospatial

---

Part of the Data Science & AI Lecture Series. Something wrong on this page? Edit _data/talks/2025-09-24-kenneth-blake-vernon.toml.