R/ClusterLhnData.R
ClusterLhnData.RdIn the following, T: the number of time bins, S: the number of odors, N: the number of cells, K: the number of clusters.
ClusterLhnData(Data, numClusters = 3, kalpha = 10, thalpha = 3/20, sdv0 = 0.1, taua = 1, taul0 = 0.5, minIters = 0, numIters = 10000, dt = 1e-05, seed = 0, initMode = "random", iclust = NULL, verbose = TRUE, timer = "OFF", slopeRatioToStop = 100, numSlopePoints = 20, checkToStopEvery = 100, keepHistory = NULL, keepHistoryAt = NULL, maxPreFitIters = 1)
| Data | FIXME - what should the input data look like?! |
|---|---|
| numClusters | The number of clusters to use. |
| kalpha | The shape parameter for the gamma prior on alpha. |
| thalpha | The scale parameter for the gamma prior on alpha |
| sdv0 | The standard deviation of the gaussian prior on membrane potential offset. |
| taua | The rate constant for the exponential prior on the drive parameters |
| taul0 | The rate constant for the exponential prior on l0. |
| minIters | The minumum number of iterations. |
| numIters | The maximum number of iterations. |
| dt | The time constant of the updates. |
| seed | The random seed to use. |
| initMode | The initialization mode for the clustering. Can be "random", "kmeans", or "kmeans++". |
| iclust | An initial clustering assignment, if any. |
| verbose | If TRUE will print out the progress of the algorithm and other diagonstic information. |
| timer | If "ON" will time different blocks of the code. |
| slopeRatioToStop | The ratio of the rate of change of the objective at the end to the start above which to terminate. |
| numSlopePoints | How many points to take to compute the slope of the objective |
| checkToStopEvery | How often to compute the stopping ratio. |
| keepHistory | A least of strings containing the variables to track. |
| keepHistoryAt | A list of iterations at which to record history. If NULL defaults to all. |
| maxPreFitIters | The maximum number of iterations to pre fit the cell-specific parameters to the clusters. If set to 0 will not prefit the parameters. |
A list consisting of
The random seed used.
A 2 x S x K array containing the learned drive parameters
A N x 1 vector containing the learned alpha values
A N x 1 vector containing the learned v0 parameters
A N x 1 vector containing the learned l0 parameters
A N x K matrix of the cluster responsibilities for each data point.
A T x S x N x K array containing the final lambda values
A T x S x N array containing the final lambda value for the most likely cluster for each fit.
The actual number of iterations that ran.
A numIters x 1 array containing the objective function as function of the number of iterations.
A N x 1 vector of cluster assignments.
A N x 1 vector of the probabilities of the cluster chosen.
A N x 1 vector of distances to its cluster center.
A string with the exit mode of the algorithm: "ITERS" if it hit the maximum number of iterations, "SLOPE_RATIO" if it exited early due to the slope ratio.
A list containing the values of the tracked variables for the specified iterations.
A miscellaneous list to hold other variables, used mainly for debugging.