T-distributed Stochastic Neighbour Embedding

hanaml.TSNE is a R wrapper for SAP HANA PAL T-distributed Stochastic Neighbour Embedding.

hanaml.TSNE(
  data = NULL,
  key = NULL,
  features = NULL,
  max.iter = NULL,
  obj.freq = NULL,
  dim = NULL,
  learning.rate = NULL,
  theta = NULL,
  perplexity = NULL,
  exaggeration = NULL,
  random.state = NULL,
  thread.ratio = NULL
)

Arguments

data	DataFrame DataFrame containting the data.
key	character, optional Name of the ID column. Defaults to the first column if not provided.
features	character of list of characters, optional Name of feature columns. If not provided, it defaults all non-key, non-label columns of data.
max.iter	integer, optional Specifies the maximum number of iterations for optimization process. Defaults to 250.
obj.freq	integer, optional Specifies the Frequency of calculating the objective function and putting the result into OBJECTIVES table. Defaults to 50.
dim	integer, optional Dimension of the embedded space. Value other than 2 or 3 is illegal. Defaults to 2.
learning.rate	double, optional Secifies the learning rate. Defaults to 200.0.
theta	double, optional The legal value should be between 0.0 to 1.0. Setting it to 0.0 means using the “exact” method which would run O(N^2) time, otherwise TSNE would employ Barnes-Hut approximation hich would run O(NlogN). This value is a tradeoff between accuracy and training speed for Barnes-Hut approximation. The training speed would be faster with higher value. Defaults to 0.5.
perplexity	double, optional The perplexity is related to the number of nearest neighbors. Larger value is suitable for large dataset. Make sure perplexity * 3 < [number of samples]. Default to 30.0.
exaggeration	double, optional The natural clusters would be more separated with larger value, which means there would be more empty space on the map. It specifies the value to be multiplied on pij before 250 iterations. Default to 12.0.
random.state	double, optional Specifies the seed for random number generation, where 0 means current system time s used as seed, and other values are simply real seed values. Defaults to 0.
thread.ratio	double, optional Controls the proportion of available threads that can be used by this function. The value range is from 0 to 1, where 0 indicates a single thread, and 1 indicates all available threads. Values between 0 and 1 will use up to that percentage of available threads.Values outside this range are ignored. Defaults to 0.

Value

Returns a list of DataFrame

• DataFrame 1
  Result of points in low-dimensional embedded space, structured as follows

  • ID: ID (correspond to input table).

  • x: Coordinate value of the 1st dimension.

  • y: Coordinate value of the 2nd dimension.

  • z: Coordinate value of the 3rd dimension.(NULL if the embedded space is 2 dimension)

• DataFrame 2
  Statistical info, structured as follows

  • STAT_NAME: Statistics name.

  • STAT_VALUE: Statistics value.

• DataFrame 3
  Recorded values of objective function for TSNE.

  • ITER: Iteration step.

  • OBJ_VALUE: Objective value of the iteration.

Details

This algorithm prepares the data for visualization of the data with the TSNE method. It returns a DataFrame of two-dimensional embeddings of the high-dimensional rows of input data.

Examples

> data$Collect()
  ID ATT1 ATT2 ATT3 ATT4 ATT5
1  1    1    2  -10  -20    3
2  2    4    5  -30  -10    6
3  3    7    8  -40  -50    9
4  4   10   11  -25  -15   12
5  5   13   14  -12  -24   15
6  6   16   17   -9  -13   18

Call the function:

> results <- hanaml.TSNE(data = data, perplexity = 1, max.iter = 500,
                         dim = 3, theta = 0, obj.freq = 50, random.state = 30)

Results:

> results[[1]]$Collect()
  ID          x         y         z
1  1   4.875853 -189.0905 -229.5364
2  2 -67.675459  213.6617  178.3976
3  3 -68.852910  162.7109  284.9663
4  4 -68.056108  193.1181  220.2754
5  5  76.524624 -189.8509 -227.6257
6  6 123.184000 -190.5492 -226.4772