Where to continue after finishing the tutorials?

After reading through the tutorials, you should now have a decent idea of how GigaSOM.jl works internally. With real data, you may encounter situations that require deeper digging in parameters and GigaSOM internals. We list several of the most frequently used ones:

  • You may want to increase the size of SOM in initGigaSOM to get more precise clusters.
  • You may speed up the computation a lot by using neighborhood-indexing structures – see package NearestNeighbors and parameters knnTreeFun of functions trainGigaSOM and embedGigaSOM
  • It is adviced to try different settings of SOM training – of the arguments of trainGigaSOM, try modifying the starting/finishing radius (rStart, rFinal), using a different radius decay (parameter radiusFun, try e.g. linearRadius) or try a different neighborhood (kernelFun and somDistFun) or a completely different metric.
  • You can get a sharper or smoother embedding by varying the amount of neighbors (k) and smoothing of the neighborhood (smooth) of embedGigaSOM.
  • For plotting of really huge data, you may want to try GigaScatter.