The 15th Annual Conference of the
Bayesian Network Modelling Association (BNMA2023)
The 2023 Bayesian Network Modelling Association (BNMA) conference has been announced for the 1st - 3rd of November in Brisbane! In a nod to the Association's rebranding (formerly ABNMS, the Australasian Bayesian Network Modelling Society), this year's exciting conference theme is "Expanding Horizons - BNs and Beyond". We still love Bayesian networks, but let's not forget they are just one technique in an ever-expanding landscape of options for modelling the world we live in. How are these other techniques different to BNs? Are BNs ready to play nice with deep learning? Can they compete with (or complement) systems dynamics, structured equation modelling, Bayesian hierarchical models and classical statistical models?
Our Brisvegas conference venue will be the stunning University of Queensland St Lucia Campus, a short bus ride from Central Brisbane. As well as the ever-popular introduction to BN workshops on the 30th and 31st of October, there will also be a student session for those who aren't quite ready to present, but would still be excited to informally discuss their work and get feedback and ideas from (very friendly!) experts in the field.
- Workshop program
- Conference program
- Key dates
- Code of conduct and meeting rules
Talk: Partitioning directed acyclic graphs into clusters to minimise arbitrary dependent costs
Inference of Directed Acyclic Graphs (DAGs) such as Bayesian Networks (BNs) can be supported and in many cases made more efficient by partitioning (i.e. mapping) the DAG into clusters. However, optimal partitioning is NP-hard, and the challenge is exacerbated by statistical inference as the cost to be optimised is dependent on both nodes within a cluster, and the mapping of clusters connected via parent and/or child nodes, which we call dependent clusters. We discuss a novel algorithm called DCMAP which can, given an arbitrarily defined, positive cost function, iteratively and rapidly find near-optimal, then optimal cluster mappings. Shown analytically to converge to optimal solutions using dynamic programming, we also find empirically that the algorithm is time-efficient for a DBN model of a seagrass complex system using a computation cost function. For a 25 and 50-node DBN, the search space size was 9.91x109 and 1.51x1021 possible cluster mappings, respectively, but near-optimal solutions with 88% and 72% similarity to the optimal solution were found at iterations 170 and 865, respectively. The first optimal solution was found at iteration 934 (95% CI 926,971), and 2256 (2150,2271) with a cost that was 4% and 0.2% of the naive heuristic cost, respectively.
BNMA is offering four travel grants to attend the BNMA2023 Tutorials and Conference. Awards will be for $500 to contribute to the costs of travel to and from the conference. Grants will be awarded to students and new members of BNMA who will be presenting and have not attended an BNMA conference in the past. Existing members who will be presenting at the conference may also be eligible depending on grant availability.
Applications for grants should include a brief statement indicating the type of grant being applied for (student travel, new member travel or existing member travel), justifying the application, along with an estimate of travel costs, a scanned signed statement from an academic supervisor verifying student status (if in application for one of the student travel grants), and a CV. An abstract of the talk to be presented at the conference must be submitted in parallel (see above). Student travel grants will only be awarded if the recipient registers for the conference and tutorials; registration for tutorials is optional for other grantees.
Travel grant applications must be submitted by email to firstname.lastname@example.org by 30th September 2023.
If you have any questions regarding BNMA2023, please email email@example.com.
Partners & Sponsors
BNMA would like to gratefully acknowledge our conference partners and sponsors: