Skip to content

Instantly share code, notes, and snippets.

@ludovicc
Created March 24, 2016 14:18
Show Gist options
  • Save ludovicc/045ccfba464a99d7cff0 to your computer and use it in GitHub Desktop.
Save ludovicc/045ccfba464a99d7cff0 to your computer and use it in GitHub Desktop.
{
"name": "t_SNE",
"doc": "t-Distributed Stochastic Neighbor Embedding,\na technique for dimensionality reduction.\n",
"metadata": {
"docker_image": "hbpmip/r-tsne:latest"
},
"input": {
"type": "null"
},
"output": {
"doc": "New representation for the data",
"type": "array",
"items": {
"type": "record",
"name": "ReducedData",
"fields": [
{
"name": "1",
"type": "double",
"doc": "Reduced dimension 1"
},
{
"name": "2",
"type": "double",
"doc": "Reduced dimension 2"
},
{
"name": "prov",
"type": {
"type": "enum",
"name": "Enumprov",
"symbols": [
"AD1",
"AD2"
]
},
"doc": "Variable prov"
},
{
"name": "age",
"type": "double",
"doc": "Variable age"
}
]
}
},
"cells": {
"query": {
"type": {
"doc": "Definition of the query that has produced this model",
"name": "Query",
"type": "record",
"fields": [
{
"name": "variables",
"doc": "List of dependent variables",
"type": {
"type": "array",
"items": {
"type": "string"
}
}
},
{
"name": "covariables",
"doc": "List of explanatory variables",
"type": {
"type": "array",
"items": {
"type": "string"
}
}
},
{
"name": "grouping",
"doc": "List of groupings",
"type": {
"type": "array",
"items": {
"type": "string"
}
}
},
{
"name": "parameters",
"doc": "Additional parameters for tSNE function",
"type": {
"type": "record",
"name": "AdditionalParameters",
"fields": [
{
"name": "dims",
"type": "int",
"doc": "Output dimensionality"
},
{
"name": "initial_dims",
"type": "int",
"doc": "the number of dimensions that should be retained in the initial PCA step"
},
{
"name": "perplexity",
"type": "double",
"doc": "Perplexity parameter"
},
{
"name": "theta",
"type": "double",
"doc": "Speed/accuracy trade-off"
},
{
"name": "pca",
"type": "boolean",
"doc": "Whether an initial PCA step should be performed"
},
{
"name": "max_iter",
"type": "int",
"doc": "Maximum number of iterations"
}
]
}
},
{
"name": "sql",
"type": "string",
"doc": "SQL query"
},
{
"name": "count",
"type": "int",
"doc": "Number of records selected by the query"
}
]
},
"init": {
"variables": [
"age"
],
"covariables": [
"left_amygdala",
"left_splsupparlob",
"right_poparoper",
"right_pogpostcgyr"
],
"grouping": [
"prov"
],
"parameters": {
"dims": 2,
"initial_dims": 50,
"perplexity": 1,
"theta": 0.5,
"pca": true,
"max_iter": 1000
},
"sql": "select * from brain",
"count": 9
}
},
"reduced_data": {
"type": {
"doc": "New representation for the data, with the covariables projected to N dimensions, where N is the dims parameter\nand groupings added for visualisation and further processing\n",
"type": "array",
"items": {
"type": "record",
"name": "ReducedData",
"fields": [
{
"name": "1",
"type": "double",
"doc": "Reduced dimension 1"
},
{
"name": "2",
"type": "double",
"doc": "Reduced dimension 2"
},
{
"name": "prov",
"type": {
"type": "enum",
"name": "Enumprov",
"symbols": [
"AD1",
"AD2"
]
},
"doc": "Variable prov"
},
{
"name": "age",
"type": "double",
"doc": "Variable age"
}
]
}
},
"init": [
{
"1": -17.11878893,
"2": -102.96461404,
"prov": "AD1",
"age": 81.2
},
{
"1": 288.97005935,
"2": 198.74859944,
"prov": "AD1",
"age": 81.2
},
{
"1": -24.82965194,
"2": -84.44473342,
"prov": "AD1",
"age": 81.2
},
{
"1": -36.26630739,
"2": -56.97799786,
"prov": "AD1",
"age": 73.6
},
{
"1": 306.68069692,
"2": 243.7006467,
"prov": "AD1",
"age": 73.6
},
{
"1": 299.5901311,
"2": 225.99492791,
"prov": "AD1",
"age": 73.6
},
{
"1": -5.52277366,
"2": -130.82548993,
"prov": "AD1",
"age": 73.6
},
{
"1": -405.99128461,
"2": -137.37420226,
"prov": "AD2",
"age": 70.1
},
{
"1": -405.51208084,
"2": -155.85713656,
"prov": "AD2",
"age": 70.1
}
]
}
},
"action": [
{
"cell": "reduced_data"
}
]
}
@ludovicc
Copy link
Author

"items": "ReducedData" ... that looks too simple! Many thanks for spotting this one.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment