Given that t-SNE is an highly demanding algorithm, the development team at De Novo Software made a great effort to improve the speed of running the algorithm with FCS Express.
To give you an idea of how t-SNE is performing within FCS Express, a performance test has been run comparing FCS Express and the independent R package in running both the exact tSNE algorithm and the Barnes-Hut approximation.
Here is a brief description of how the test was run:
File |
Machine |
FCS Express |
R |
•FCS format version: 3.0 •Total number of events: 500397 (different sample sizes were used in the performance test) •Total number of parameters: 40 (20 parameters have been selected in the performance test) |
•CPU: Intel Core i5-4200M CPU 2.5GHz •Memory: 8Gb •System type: Windows 8.1 64 bit |
•Version: 6.00.0045 (64 bit) |
R version 3.3.1 (2016-06-21) R Studio Version 0.99.903
|
The graph below (Figure 29.59) show the elapsed time (in minutes) for t-SNE calculation (using both the exact t-SNE and the Burnes-Hut Approximation (Amount of Approximation = 0.5)) in both FCS Express and R. A fixed number of parameters (i.e. n=20) and different numbers of events have been used. t-SNE was not estimated for unsampled events. Please refer to the Defining a tSNE transformation chapter for more information about all the above mentioned options.
Figure 29.59 - Performance comparison between FCS Express and R in running both the exact tSNE and the Barnes-Hut Approximation.