BM 20160408B

Data On Nodes Thu May 12 22:41:17 CDT 2016 2659296c74bcec23c5dec54ac227ebf4d09dd60a

Bifurcating structure simulation

Pure birth process, 8 tips and 7 internal nodes, edge length \(t=50\)

Brownian data points on nodes

  • Starting from root \(\{Y_1 = 0, Y_2 = 0, \ldots, Y_K = 0\}_{K}\)
    • \(K\) data point (or “loci”) on each node
  • Simulate instances of Brownian motion \[Y_k^{(t)} := N(\mu t, \sigma^2 t)\] where \(t\) is edge length; data on node \(T\) \[Y_k^{(T)} = \sum_{t=1}^T Y_k^{(t)}\]
  • Repeat \(N\) times such that each node will have \(N\) “draws”

Simulation default setup: \(N = 100, K=100, \sigma = 0.1\)

Methods for Manifold learning

6 methods (or 9 considering variants on LLE)

  • Global distances:
    • PCA (covariance)
    • MDS (dissimilarities)
  • Local distances:
    • t-SNE (t-distributed Stochastic Neighbor Embedding)
    • Spectral embedding
    • Locally linear embedding including standard, LTSA, Hessian and modified LLE
    • LTSA: Local Tangent Space Alignment

Run

./tree-dmr.sos -c tree-dmr.yml --date 20160416 \
               --simulators bm_simulate bmf_simulate \
               --analysts manifold_analyze

Results

Brownian on nodes

Brownian on nodes

Brownian on nodes