​​ 11 ppl attended ​

Open exchange

​

show casing of some databases

• Robert shows the “Heurist” database by ian johnson database: archaeology, social science
  • is easy to upload data (even 3D data)
  • heurist.huma.num.fr
  • built in record sheets, map, etc, quite general purpose
  • https://heurist.huma-num.fr/heurist/startup/
• node-goat: graph database
  • https://nodegoat.net
  • easy to use for network analysis
• arches database
  • https://eamena-project.github.io/eamena-arches-dev/talks/2023-gmpca/pres/#/arches-it-stack ​

computational environments

• recent article about computational environments https://doi.org/10.1038/d41586-023-01469-0
• discusses the renv R package, Ben Marwick was interviewed briefly for it, along with many others ​

new methods

• most archaeological data is visual data
• triple IF to describe image (https://iiif.io/)
• JSON format
• 3Dhop will develop sth similar for 3D ​

CAA wrap up

• SIG SSLA had a good meeting in person
• SIG SSLA had a successful session
• CAA ethics officer is being needed! ​

very high CAA conference fees

• CAA steering committee is aware it
• task force is being developed to discuss new ideas about that
• at the moment mostly men, middle-aged and from Europe
• call to diversify! if you are interested, please join
• discussion:
  • needs to be more accessible
  • needs more outreach, esp. to Asia and Africa
  • more accessibility via hybrid
  • “international” is in the name
  • professional conference centers are the main trouble (high fees, equipment is expensive) –> go somewhere else, esp. universities
  • in European context it’s unusual not to be at university, in other areas (North America) it is more usual to be at a conference center
  • more open about the size of the bursary for students, so they can plan according
• Jim and Martin may be part of the task force, don’t hesitate to contact them ​

SIG SSLA computational archaeology map

• https://sslarch.github.io/MapofComputationalArchaeology/
• filtering and searching mechanism is more user-friendly now (tags are click-able), search more inclusive (across all fields)
• please: everyone add new entries
• will be tweeted about soon!
• map is written in Elm, may need some familiarity to get used to it, if you are interested to check it out ​

geometric morphometrics

​

talk by David Matzig

• on his PhD to use Geometric Morphometric Methods (GMM) and phylogenetic comparative methods (PCM) -> to assess stone tool typologies and study the cultural evolution
• part of CLIOARCH
• traditional morphometrics and landmark based GMM use some human-chosen feature points for comparison
• outline based GMM -> equally spaced points + elliptic fourier analysis –> doesn’t need human decision process and can be automated
• package outlineR -> extracts outlines from pictures
• Momocs: elliptic fourier transformation and PCA
• -> this can be clustered
• paper with Hussain and Riede in Journal of Paleolithic Archaeology 2021
  • compared whole-outline GMM to typology based study
  • Nicolas 2017 data (typologcal methods) compared with output from this workflow
  • -> quite successful in recreating typochronology and clustering of regions
• future work:
  • use bayesian phylogenies
  • inferred by a tree-partite models (Watnock & Wright 2020)
  • planned outcome: tree typology which can be used to infer new
  • planned outcome: extinction rates, technological evolution random or not?
  • will be tested for correlation with climate, hunting strategies etc ​

talk by Robert Bischoff

• plans a regional study in SW of USA (between Arizona and New Mexico)
• projectile point typologies are not well developed and not comparable within the area
• GMM used for re-analysis:
  • Fourrier does not capture variation in side notches, which are shallow + many points are broken
  • therefore used landmark analysis with equal distance of points
  • tips are not included in the analysis (no landmarks at the points)
  • program: tpsDig2
  • generalised procrastis alignment –> so that all points are on the same plane
• PCA, hierarchical clustering similar to Davids method
• network by PCA distances and prject the outline of the points -> easier to see clusters
• many similarities, also between clusters
• use this approach to compare sites (similar/distance based)
• compare different networks (architecture, pottery, projectile points)
• network as a graph tool
• preprint: https://osf.io/preprints/socarxiv/a6wjc/ ​

discussion

• nearest neighbor, split graph (https://en.wikipedia.org/wiki/Split_graph) to work with the tree -> is a mixture of a tree and a network
• Thomas Huet uses Davids data for teaching: http://shinyserver.cfs.unipi.it:3838/teach/stats/upv3/_site/#/dim-data-shape
• visualisation of networks: sometimes readers mix up the weight of the edge by colour with the distance between the nodes “in print”
• careful with displaying algorithms
• good idea to sometimes use landmarks, well reasoned
• canonical analysis, redundancy analysis can be used for to check for clusters
• alternatives to PCA?
  • David: without PCA, just Fourrier -> works the same, but PCA gives a nice visual summary
  • Martin: does not highlight small changes as well, used t-SNE (https://en.wikipedia.org/wiki/T-distributed_stochastic_neighbor_embedding)
  • David: yeah, similar problem, but maybe more a problem with Fourrier?
  • Georg: bit more skeptical, because some input needed (perplexity) and there is no good way to choose which ​- posterior cluster check?
  • elbow and silhouette
  • visualisation
  • Georg: silhouette is probably the best
• what do you use for a priory knowledge in the tree building?
  • distribution of priors
  • x million reruns -> finds its own priors
  • C14-dates of projectiles added
  • there is no implementation of bayesian phylogenetics in R
  • RevBayes: Bayesian Phylogenetic Inference Using Graphical Models and an Interactive Model-Specification Language https://doi.org/10.1093/sysbio/syw021 ​

next meeting 07th June