Max number of iteration cylces 
Number of cycles, FItSNE tries to optimize positioning the output vectors. A larger number of iteration may generate a more accurate map, in case FItSNE finds a local optimal solution. Most simply, run FItSNE with a decent number of iterations (e.g. 250). In case the mapping looks non informative, try to run it again with more iterations (e.g. 500 next round 100, ...). Better, run FItSNE on the command line and see its progress dependong on the iterations. 
Perplexity:  A larger perplexity will tend to spread the data clouds. Try different values and observe the outcome. But: recommended Perplexity = (Number_of_Features_Conditions  1) / 3). I..e. 30 samples => Perplexity ~10. SUMO automatically adjusts a too high perplexity accordingly. 
Dimensions:'  Generate a 2 dimensional embedding => data clouds in a 2dplain (commonly used model) or a 1dimensional embedding => stacked bars 
Number of vectors  Legth of vectors  Consumed RAM (GB)  Elapsed time (s) 

100000  1000  2.8  270 
10000  1000  0.3  30 
1000  1000  0.03  15 
1000  10000  0.3  45 
1000  100000  2.5  65 
conda install c condaforge umaplearn
conda install c condaforge/label/cf201901 umaplearn
proxy_servers:
http: http://user:pass@corp.com:8080
https: https://user:pass@corp.com:8080
Number of nearest neighbors 

Minimal distance  
Dimensions:  Define number of dimensions for embedding: 1  dimensioanl embedding 2  dimensional embedding => data clouds in a 2Dplain (commonly used model) 3  dimensional embedding => data clouds in a 3Dspace 
Metric  Method how to compute similarity / distance between data vectors. Choice of a metric may fudamentally alter resulting embedding. Thus consider carefully which metric to use depending on the data, the experimental design and the scientific question. 
Annotation column ID for grouping 
Data viewers will use this annotation column to auto group (genes/conditions). The annotation column should contain a small set (~≤<100) of "class identifiers" (e.g. treatment names). Specify "0" or leave empty to avoid autogrouping. 
Anaconda path  Specify location of your ANACONDA, or any other Python installation Click the ... button to open a file system browser and navigate to the repective folder. 
Number of vectors  Legth of vectors  Elapsed time (s) 

100  100  5.27 
1000  100  7.63 
10000  100  31.67 
100000  100  171.53 
100  1000  6.11 
100  10000  11.83 
100  100000  68.28 
1D Here, YDimension is meaningless. 
2D 
3D 
Euclidean 
Canberra 
Cosine 
Correlation 