BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Contributed Talk: Learning to Design Infrastructure Networks - Aka
 nksha Ahuja (University of Cambridge)
DTSTART:20260210T110000Z
DTEND:20260210T113000Z
UID:TALK242314@talks.cam.ac.uk
DESCRIPTION:Designing large-scale spatial infrastructure is a longstanding
  challenge. Infrastructure network topologies encode implicit design princ
 iples that can be inferred from data. They are multi-scale by construction
 \, with dense local connectivity and sparse long-range links constrained b
 y geography and cost.\n&nbsp\;\nIn optical backbones\, physical topology d
 irectly affects network performance which in turn affects communications a
 mong people\, business and organisations. Designing an optimal topology is
  complex due to scalability and computational constraints. Conventional gr
 aph generators rarely scale beyond ~40 nodes\, while manual or optimizatio
 n-based approaches require months of effort and become intractable for lar
 ge-scale networks. Consequently\, automating the design of optical network
 s remains an open challenge\, further compounded by limited access to real
 -world deployment data.\n&nbsp\;\nCan graph machine learning models infer 
 network design principles from past topologies to generate scalable topolo
 gies for the future? We introduce Topology Bench\, an open dataset of 105 
 real-world network topologies spanning diverse regions and scales. In this
  work\, we present Topology Architect\, the first application of graph mac
 hine learning for unsupervised topology generation. We evaluate the model 
 using graph similarity metrics\, achieving 95% spectral and 84% structural
  similarity to real networks as measured by Wasserstein distances. The mod
 el generates topologies in under one second by sampling edge probabilities
  from the latent space\, conditioned on user-specified parameters such as 
 node count and geographic locations. To assess latent space quality\, we c
 ompare clustering in Topology Bench using graph-theoretic properties and g
 raph embeddings from Topology Architect\, finding the latter 20 times more
  effective.\nThis research makes three key contributions: (i) provides an 
 open-access dataset of real-world network topologies\, (ii) captures the g
 raph-theoretic properties underlying topology design\, and (iii) automates
  topology generation. This further enables the synthesis of realistic topo
 logies in regions with sparse or proprietary data. While developed for opt
 ical networks\, this framework is transferable to large-scale infrastructu
 re systems\, including transportation\, urban planning\, and quantum commu
 nication.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR