Information, geometry, and pricing
Probability-valued firm information is used to study asset co-movement, construct interaction fields, and certify portfolio diversification.
Ask AI about our research:
Observable information → geometric structure → financial restriction
Language models can place heterogeneous documents in a common space. The next question is not only whether those representations predict a scalar outcome, but what mathematical structure they can provide for co-movement, peer interaction, and portfolio risk.
The three papers retain each firm's collection of article representations as a probability distribution. Transport geometry then plays three distinct roles: comparing two distributions, reconstructing a fixed target from several aligned peers, and aggregating dispersion under one coherent multi-firm law.
Read the blog postThe papers
Each paper changes which object is held fixed and which object is allowed to vary.
Quadratic Wasserstein geometry turns distributional separation into sharp conditional bounds on attainable systematic covariance under explicit information-to-exposure restrictions.
Target-anchored transport and simplex reconstruction turn distributions of firm information into a directed, bandwidth-free peer field for spatial exposure adjustment.
Multi-firm Wasserstein-2 dispersion yields a one-sided portfolio-risk certificate and an allocation rule that does not require cross-asset return covariances.
Animated companion
A narrated animation of the programme.
Narrated video · 05:40
The video moves from firms as probability distributions to couplings, covariance envelopes, target-anchored reconstruction, coherent multi-firm dispersion, and portfolio risk certificates.
Podcast explainer
A NotebookLM-generated podcast-style explainer of firms as news and probability distributions.
NotebookLM-generated podcast · 36:11
Interactive companion
A browser-native notebook where the W₂ separation you dial in caps how much the returns can co-move.
Runs in your browser · ~30s to start
Researchers
Work timeline
From language models to distributional geometry.
The programme takes shape around firm information, geometry, and pricing.
The first manuscript develops distributional geometry for systematic covariance.
Read the pre-printThe work reaches its first conference proceeding in Malta.
World Finance ConferenceResearcher
Department of Finance and Tax · University of Cape Town
Research at the intersection of quantitative finance, distributional geometry, language-model representations, and mathematical foundations.
Associate Professor
Department of Finance and Tax · University of Cape Town
Research in quantitative finance, financial econometrics, computational finance, stochastic processes, and financial risk management.
