tmap - Visualize big high-dimensional data

Getting started

NIPS conference papers visualized as a graph / spanning tree using tmap.

tmap is a very fast visualization library for large, high-dimensional data sets. Currently, tmap is available for Python.

NEW: We now provide a web-service that allows for the creaton of TMAP visualizations for small chemical data sets.

Try TMAP

Supported Environments

Language	Operating System	Status
Python	Linux	Available
	Windows	Available¹
	macOS	Available
R		Available²

¹Works with WSL
²Through the reticulate R interface to Python.

Installation

tmap is installed using the conda package manager. Don't have conda? Download miniconda.

conda install -c tmap tmap

We suggest using faerun to plot the data layed out by tmap. But you can of course also use matplotlib (which might be to slow for large data sets and doesn't provide interactive features). Following, an example plot produced with tmap and matplotlib.

tmap visualization plotted by matplotlib

pip install faerun
# pip install matplotlib

Laying out a Simple Graph

Even though TMAP is mainly targeted at tasks consisting of laying out very large data sets as trees, the simplest usage example is laying out a graph.

This example produces a matplotlib plot displaying the graph.

A graph layed out by tmap plotted with matplotlib

When laying out large graphs, it might be useful to discard some edges in order to create a more interpretable and visually pleasing layout. This is achieved using the (default) argument create_mst=True. Following, this is exemplified on a small graph.

This example produces a matplotlib plot displaying the graph and it's spanning tree. In the plot on the left, the spanning tree is highlighted by coloring the edges red. The thickness of the edges is inversely related to the edge weight.

A graph and its spanning tree layed out by tmap plotted with matplotlib

On a highly connected graph with 1000 vertices, the advantages of the tree visualizaton method applied by TMAP become obvious.

A large graph and its spanning tree layed out by tmap plotted with matplotlib

There are a wide array of options to tune the final tree layout to your linking. See Layout for the descriptions of all available parameters.

MinHash

To enable the visualization of larger data sets, it is necessary to speed up the k-nearest neighbor graph generation. While in general, any approach can be used to create this nearest neighbor graph (see Laying out a Simple Graph), tmap provides a built-in LSH Forest data structure, which enables extremely fast k-nearest neighbor queries.

In order to index data in the LSH forest data structure, it has to be hashed using a locality sensitive scheme such as MinHash.

tmap includes the two classes Minhash and LSHForest for fast k-nearest neighbor search.

The following example shows how to use the Minhash class to estimate Jaccard distances.

python minhash.py
0.390625
0.140625

LSH Forest

The hashes generated by Minhash can be indexed using LSHForest for fast k-nearest neighbor retreival.

python lsh_forest.py
Generating the data took 1164.400289999321ms.
Encoding the data took 155.53330996772274ms.
Adding the data took 17.799988971091807ms.
Indexing took 4.7386749647557735ms.
The kNN search took 0.26249000802636147ms.

After indexing the data, the 10 nearest neighbor search on a million 1,000-dimensional vectors took ~0.32ms. In addition, the LSHForest class also supports the parallelized generation of a k-nearest neighbor graph using the method get_knn_graph().

python lsh_forest_knng.py
Generating the data took 1171.7670540092513ms.
Adding the data took 189.09973296104ms.
Indexing took 5.959620990324765ms.
The kNN search took 311.5222750348039ms.

Using matplotlib / pyplot has tow main disadvantages: It is slow and does not yield interactive plots. For this reason, we suggest to use the Python package Faerun for large scale data sets. Faerun supports millions of data points in web-based visualizations.

Together with tmap, Faerun can easily create visualizations of more than 10 million data points including associated web links and structure drawings for high dimensional chemical data sets within an hour.

Examples

COIL 20

Data set dimensions: 1,440×16,384

The COIL-20 data set visualized as a graph / spanning tree using tmap.

View Download

Source

MNIST

Data set dimensions: 70,000×764

MNIST digitized digits visualized as a graph / spanning tree using tmap.

View Download

Source

Fashion MNIST

Data set dimensions: 70,000×764

Fashion MNIST fashion items visualized as a graph / spanning tree using tmap.

View Download

Source

ChEMBL

Data set dimensions: 1,159,881×2³²

ChEMBL molecular assay data visualized as a graph / spanning tree using tmap.

View Request Data

Source

FDB17 and ChEMBL

Data set dimensions: 11,261,085×2³²

ChEMBL molecular assay data visualized as a graph / spanning tree using tmap.

View Request Data

Natural Product Atlas

Data set dimensions: 24,594×2³²

The Natural Product Atlas visualized as a graph / spanning tree using tmap.

View Download

Source

DSSTox

Data set dimensions: 848,816×2³²

The Distributed Structure-Searchable Toxicity (DSSTox) Database visualized as a graph / spanning tree using tmap.

View Download

Source

Protein DataBank

Data set dimensions: 131,236×136

Protein DataBank proteins visualized as a graph / spanning tree using tmap.

View Request Data

NIPS Conference Papers

Data set dimensions: 7,241×225,423

NIPS conference papers visualized as a graph / spanning tree using tmap.

View Download

Source

Drugbank

Data set dimensions: 9,300×2^32

Drugbank compounds visualized as a graph / spanning tree using tmap.

View Download

Source

Flowcytometry

Data set dimensions: 436,877×14

Flowcytometry data visualized as a graph / spanning tree using tmap.

View Download

Source

moleculenet.ai

Quantum Mechanics

QM8 Source Electronic spectra and excited state energy of small molecules calculated by multiple quantum mechanic methods.
QM9 source Geometric, energetic, electronic and thermodynamic properties of DFT-modelled small molecules.

Physical Chemistry

ESOL source Water solubility data(log solubility in mols per litre) for common organic small molecules.
FreeSolv source Experimental and calculated hydration free energy of small molecules in water.
Lipophilicity source Experimental results of octanol/water distribution coefficient(logD at pH 7.4).

Biophysics

PCBA (Due to the number of variables in this data set, please run the linked script to generated the visualizations locally) Selected from PubChem BioAssay, consisting of measured biological activities of small molecules generated by high-throughput screening.
MUVsource Subset of PubChem BioAssay by applying a refined nearest neighbor analysis, designed for validation of virtual screening techniques.
- MUV-466
- MUV-548
- MUV-600
- MUV-644
- MUV-652
- MUV-689
- MUV-692
- MUV-712
- MUV-713
- MUV-733
- MUV-737
- MUV-810
- MUV-832
- MUV-846
- MUV-852
- MUV-858
- MUV-859
HIV source Experimentally measured abilities to inhibit HIV replication.
PDBbind source Binding affinities for bio-molecular complexes, both structures of proteins and ligands are provided.
BACE source Quantitative (IC50) and qualitative (binary label) binding results for a set of inhibitors of human β-secretase 1(BACE-1).

Physiology

BBBP source Binary labels of blood-brain barrier penetration(permeability).
Tox21 source Qualitative toxicity measurements on 12 biological targets, including nuclear receptors and stress response pathways.
ToxCast source Toxicology data for a large library of compounds based on in vitro high-throughput screening, including experiments on over 600 tasks.
SIDER source Database of marketed drugs and adverse drug reactions (ADR), grouped into 27 system organ classes.
ClinTox source Qualitative data of drugs approved by the FDA and those that have failed clinical trials for toxicity reasons.

Documentation

Minhash

class tmap.Minhash(self: tmap.Minhash, d: int=128, seed: int=42, sample_size: int=128) → None¶

A generator for MinHash vectors that supports binary, indexed, string and also int and float weighted vectors as input.

Constructor for the class Minhash.

Keyword Arguments

d (int) – The number of permutations used for hashing
seed (int) – The seed used for the random number generator(s)
sample_size (int) – The sample size when generating a weighted MinHash

batch_from_binary_array(self: tmap.Minhash, arg0: List[tmap.VectorUchar]) → List[tmap.VectorUint]¶

Create MinHash vectors from binary arrays (parallelized).

Parameters: vec (List of VectorUchar) – A list of vectors containing binary values
Returns: A list of MinHash vectors
Return type: List of VectorUint

batch_from_int_weight_array(self: tmap.Minhash, arg0: List[tmap.VectorUint]) → List[tmap.VectorUint]¶

Create MinHash vectors from int arrays, where entries are weights rather than indices of ones (parallelized).

Parameters: vec (List of VectorUint) – A list of vectors containing int values
Returns: A list of MinHash vectors
Return type: List of VectorUint

batch_from_sparse_binary_array(self: tmap.Minhash, arg0: List[tmap.VectorUint]) → List[tmap.VectorUint]¶

Create MinHash vectors from sparse binary arrays (parallelized).

Parameters: vec (List of VectorUint) – A list of vectors containing indices of ones in a binary array
Returns: A list of MinHash vectors
Return type: List of VectorUint

batch_from_string_array(self: tmap.Minhash, arg0: List[List[str]]) → List[tmap.VectorUint]¶

Create MinHash vectors from string arrays (parallelized).

Parameters: vec (List of List of str) – A list of list of strings
Returns: A list of MinHash vectors
Return type: List of VectorUint

batch_from_weight_array(self: tmap.Minhash, vecs: List[tmap.VectorFloat], method: str='ICWS') → List[tmap.VectorUint]¶

Create MinHash vectors from float arrays (parallelized).

Parameters: vec (List of VectorFloat) – A list of vectors containing float values
Keyword Arguments: method (str) – The weighted hashing method to use (ICWS or I2CWS)
Returns: A list of MinHash vectors
Return type: List of VectorUint

from_binary_array(self: tmap.Minhash, arg0: tmap.VectorUchar) → tmap.VectorUint¶

Create a MinHash vector from a binary array.

Parameters: vec (VectorUchar) – A vector containing binary values
Returns: A MinHash vector
Return type: VectorUint

from_sparse_binary_array(self: tmap.Minhash, arg0: tmap.VectorUint) → tmap.VectorUint¶

Create a MinHash vector from a sparse binary array.

Parameters: vec (VectorUint) – A vector containing indices of ones in a binary array
Returns: A MinHash vector
Return type: VectorUint

from_string_array(self: tmap.Minhash, arg0: List[str]) → tmap.VectorUint¶

Create a MinHash vector from a string array.

Parameters: vec (List of str) – A vector containing strings
Returns: A MinHash vector
Return type: VectorUint

from_weight_array(self: tmap.Minhash, vec: tmap.VectorFloat, method: str='ICWS') → tmap.VectorUint¶

Create a MinHash vector from a float array.

Parameters: vec (VectorFloat) – A vector containing float values
Keyword Arguments: method (str) – The weighted hashing method to use (ICWS or I2CWS)
Returns: A MinHash vector
Return type: VectorUint

get_distance(self: tmap.Minhash, arg0: tmap.VectorUint, arg1: tmap.VectorUint) → float¶

Calculate the Jaccard distance between two MinHash vectors.

Parameters

vec_a (VectorUint) – A MinHash vector
vec_b (VectorUint) – A MinHash vector

Returns

float The Jaccard distance

get_weighted_distance(self: tmap.Minhash, arg0: tmap.VectorUint, arg1: tmap.VectorUint) → float¶

Calculate the weighted Jaccard distance between two MinHash vectors.

Parameters

vec_a (VectorUint) – A weighted MinHash vector
vec_b (VectorUint) – A weighted MinHash vector

Returns

float The Jaccard distance

LSHForest

class tmap.LSHForest(self: tmap.LSHForest, d: int=128, l: int=8, store: bool=True, file_backed: bool=False, weighted: bool=False) → None¶

A LSH forest data structure which incorporates optional linear scan to increase the recovery performance. Most query methods are available in parallelized versions named with a batch_ prefix.

Constructor for the class LSHForest.

Keyword Arguments

d (int) – The dimensionality of the MinHashe vectors to be added
l (int) – The number of prefix trees used when indexing data
store (bool) – Store the data for enhanced querying
file_backed (bool) – Whether to store the data on disk rather than in main memory (experimental)

add(self: tmap.LSHForest, arg0: tmap.VectorUint) → None¶

Add a MinHash vector to the LSH forest.

Parameters: vecs (VectorUint) – A MinHash vector that is to be added to the LSH forest

batch_add(self: tmap.LSHForest, arg0: List[tmap.VectorUint]) → None¶

Add a list MinHash vectors to the LSH forest (parallelized).

Parameters: vecs (List of VectorUint) – A list of MinHash vectors that is to be added to the LSH forest

batch_query(self: tmap.LSHForest, arg0: List[tmap.VectorUint], arg1: int) → List[tmap.VectorUint]¶

Query the LSH forest for k-nearest neighbors (parallelized).

Parameters

vecs (List of VectorUint) – The query MinHash vectors
k (int) – The number of nearest neighbors to be retrieved

Returns

The results of the queries

Return type

List of VectorUint

clear(self: tmap.LSHForest) → None¶: Clears all the added data and computed indices from this LSHForest instance.

get_all_distances(self: tmap.LSHForest, arg0: tmap.VectorUint) → tmap.VectorFloat¶

Calculate the Jaccard distances of a MinHash vector to all indexed MinHash vectors.

Parameters: vec (VectorUint) – The query MinHash vector
Returns: The Jaccard distances
Return type: List of float

get_all_nearest_neighbors(self: tmap.LSHForest, k: int, kc: int=10) → tmap.VectorUint¶

Get the k-nearest neighbors of all indexed MinHash vectors.

Parameters: k (int) – The number of nearest neighbors to be retrieved
Keyword Arguments: kc (int) – The factor by which k is multiplied for LSH forest retreival
Returns: VectorUint The ids of all k-nearest neighbors

get_distance(self: tmap.LSHForest, arg0: tmap.VectorUint, arg1: tmap.VectorUint) → float¶

Calculate the Jaccard distance between two MinHash vectors.

Parameters

vec_a (VectorUint) – A MinHash vector
vec_b (VectorUint) – A MinHash vector

Returns

float The Jaccard distance

get_distance_by_id(self: tmap.LSHForest, arg0: int, arg1: int) → float¶

Calculate the Jaccard distance between two indexed MinHash vectors.

Parameters

a (int) – The id of an indexed MinHash vector
b (int) – The id of an indexed MinHash vector

Returns

float The Jaccard distance

get_hash(self: tmap.LSHForest, arg0: int) → tmap.VectorUint¶

Retrieve the MinHash vector of an indexed entry given its index. The index is defined by order of insertion.

Parameters: a (int) – The id of an indexed MinHash vector
Returns: VectorUint The MinHash vector

get_knn_graph(self: tmap.LSHForest, from: tmap.VectorUint, to: tmap.VectorUint, weight: tmap.VectorFloat, k: int, kc: int=10) → None¶

Construct the k-nearest neighbor graph of the indexed MinHash vectors. It will be written to out parameters from, to, and weight as an edge list.

Parameters

from (VectorUint) – A vector to which the ids for the from vertices are written
to (VectorUint) – A vector to which the ids for the to vertices are written
weight (VectorFloat) – A vector to which the edge weights are written
k (int) – The number of nearest neighbors to be retrieved during the construction of the k-nearest neighbor graph

Keyword Arguments

kc (int) – The factor by which k is multiplied for LSH forest retreival

get_weighted_distance(self: tmap.LSHForest, arg0: tmap.VectorUint, arg1: tmap.VectorUint) → float¶

Calculate the weighted Jaccard distance between two MinHash vectors.

Parameters

vec_a (VectorUint) – A weighted MinHash vector
vec_b (VectorUint) – A weighted MinHash vector

Returns

float The Jaccard distance

get_weighted_distance_by_id(self: tmap.LSHForest, arg0: int, arg1: int) → float¶

Calculate the Jaccard distance between two indexed weighted MinHash vectors.

Parameters

a (int) – The id of an indexed weighted MinHash vector
b (int) – The id of an indexed weighted MinHash vector

Returns

float The weighted Jaccard distance

index(self: tmap.LSHForest) → None¶: Index the LSH forest. This has to be run after each time new MinHashes were added.

is_clean(self: tmap.LSHForest) → bool¶

Returns a boolean indicating whether or not the LSH forest has been indexed after the last MinHash vector was added.

Returns: True if index() has been run since MinHash vectors have last been added using add() or batch_add(). False otherwise
Return type: bool

linear_scan(self: tmap.LSHForest, vec: tmap.VectorUint, indices: tmap.VectorUint, k: int=10) → List[Tuple[float, int]]¶

Query a subset of indexed MinHash vectors using linear scan.

Parameters

vec (VectorUint) – The query MinHash vector
indices (VectorUint) –

Keyword Arguments

k (int) – The number of nearest neighbors to be retrieved

Returns

The results of the query

Return type

List of Tuple[float, int]

query(self: tmap.LSHForest, arg0: tmap.VectorUint, arg1: int) → tmap.VectorUint¶

Query the LSH forest for k-nearest neighbors.

Parameters

vec (VectorUint) – The query MinHash vector
k (int) – The number of nearest neighbors to be retrieved

Returns

The results of the query

Return type

VectorUint

query_by_id(self: tmap.LSHForest, arg0: int, arg1: int) → tmap.VectorUint¶

Query the LSH forest for k-nearest neighbors.

Parameters

id (int) – The id of an indexed MinHash vector
k (int) – The number of nearest neighbors to be retrieved

Returns

The results of the query

Return type

VectorUint

query_exclude(self: tmap.LSHForest, arg0: tmap.VectorUint, arg1: tmap.VectorUint, arg2: int) → tmap.VectorUint¶

Query the LSH forest for k-nearest neighbors.

Parameters

vec (VectorUint) – The query MinHash vector
exclude (List of VectorUint) –
k (int) – The number of nearest neighbors to be retrieved

Returns

The results of the query

Return type

VectorUint

query_exclude_by_id(self: tmap.LSHForest, arg0: int, arg1: tmap.VectorUint, arg2: int) → tmap.VectorUint¶

Query the LSH forest for k-nearest neighbors.

Parameters

id (int) – The id of an indexed MinHash vector
exclude (List of VectorUint) –
k (int) – The number of nearest neighbors to be retrieved

Returns

The results of the query

Return type

VectorUint

query_linear_scan(self: tmap.LSHForest, vec: tmap.VectorUint, k: int, kc: int=10) → List[Tuple[float, int]]¶

Query k-nearest neighbors with a LSH forest / linear scan combination. k`*:obj:`kc nearest neighbors are searched for using LSH forest; from these, the k nearest neighbors are retrieved using linear scan.

Parameters

vec (VectorUint) – The query MinHash vector
k (int) – The number of nearest neighbors to be retrieved

Keyword Arguments

kc (int) – The factor by which k is multiplied for LSH forest retreival

Returns

The results of the query

Return type

List of Tuple[float, int]

query_linear_scan_by_id(self: tmap.LSHForest, id: int, k: int, kc: int=10) → List[Tuple[float, int]]¶

Parameters

id (int) – The id of an indexed MinHash vector
k (int) – The number of nearest neighbors to be retrieved

Keyword Arguments

kc (int) – The factor by which k is multiplied for LSH forest retreival

Returns

The results of the query

Return type

List of Tuple[float, int]

query_linear_scan_exclude(self: tmap.LSHForest, vec: tmap.VectorUint, k: int, exclude: tmap.VectorUint, kc: int=10) → List[Tuple[float, int]]¶

Parameters

vec (VectorUint) – The query MinHash vector
k (int) – The number of nearest neighbors to be retrieved

Keyword Arguments

exclude (List of VectorUint) –
kc (int) – The factor by which k is multiplied for LSH forest retreival

Returns

The results of the query

Return type

List of Tuple[float, int]

query_linear_scan_exclude_by_id(self: tmap.LSHForest, id: int, k: int, exclude: tmap.VectorUint, kc: int=10) → List[Tuple[float, int]]¶

Parameters

id (int) – The id of an indexed MinHash vector
k (int) – The number of nearest neighbors to be retrieved

Keyword Arguments

exclude (List of VectorUint) –
kc (int) – The factor by which k is multiplied for LSH forest retreival

Returns

The results of the query

Return type

List of Tuple[float, int]

restore(self: tmap.LSHForest, arg0: str) → None¶

Deserializes a previously serialized (using store()) state into this instance of LSHForest and recreates the index.

Parameters: path (str) – The path to the file which is deserialized

size(self: tmap.LSHForest) → int¶

Returns the number of MinHash vectors in this LSHForest instance.

Returns: The number of MinHash vectors
Return type: int

store(self: tmap.LSHForest, arg0: str) → None¶

Serializes the current state of this instance of LSHForest to the disk in binary format. The index is not serialized and has to be rebuilt after deserialization.

Parameters: path (str) – The path to which to searialize the file

Layout

layout_from_lsh_forest(lsh_forest: tmap::LSHForest, config: tmap.LayoutConfiguration=tmap.LayoutConfiguration(), create_mst: bool=True, clear_lsh_forest: bool=False, weighted: bool=False) → Tuple[tmap.VectorFloat, tmap.VectorFloat, tmap.VectorUint, tmap.VectorUint, tmap.GraphProperties]

Create minimum spanning tree or k-nearest neighbor graph coordinates and topology from an LSHForest instance.

Parameters

lsh_forest (LSHForest) – An LSHForest instance

Keyword Arguments

config (LayoutConfiguration, optional) – An LayoutConfiguration instance

create_mst (bool): Whether to create a minimum spanning tree or to return coordinates and topology for the k-nearest neighbor graph clear_lsh_forest (bool) – Whether to run clear() on the LSHForest instance after k-nearest neighbor graph and MST creation and before layout

Returns

The x and y coordinates of the vertices, the ids of the vertices spanning the edges, and information on the graph

Return type

Tuple[VectorFloat, VectorFloat, VectorUint, VectorUint, GraphProperties]

layout_from_edge_list(vertex_count: int, edges: List[Tuple[int, int, float]], config: tmap.LayoutConfiguration=tmap.LayoutConfiguration(), create_mst: bool=True) → Tuple[tmap.VectorFloat, tmap.VectorFloat, tmap.VectorUint, tmap.VectorUint, tmap.GraphProperties]

Create minimum spanning tree or k-nearest neighbor graph coordinates and topology from an edge list.

Parameters

vertex_count (int): The number of vertices in the edge list edges (List of Tuple[int, int, float]) – An edge list defining a graph

Keyword Arguments:

config (LayoutConfiguration, optional) – An LayoutConfiguration instance

create_mst (bool) – Whether to create a minimum spanning tree or to return coordinates and topology for the k-nearest neighbor graph

Returns

The x and y coordinates of the vertices, the ids of the vertices spanning the edges, and information on the graph

Return type

Tuple[VectorFloat, VectorFloat, VectorUint, VectorUint, GraphProperties]

tmap.mst_from_lsh_forest(lsh_forest: tmap::LSHForest, k: int, kc: int=10, weighted: bool=False) → Tuple[tmap.VectorUint, tmap.VectorUint]¶

Create minimum spanning tree topology from an LSHForest instance.

Parameters

lsh_forest (LSHForest) – An LSHForest instance
k (int) – The number of nearest neighbors used to create the k-nearest neighbor graph

Keyword Arguments

kc (int) – The scalar by which k is multiplied before querying the LSH forest. The results are then ordered decreasing based on linear-scan distances and the top k results returned
weighted (bool) – Whether the MinHash vectors in the LSHForest instance are weighted

Returns

the topology of the minimum spanning tree of the data indexed in the LSH forest

Return type

Tuple[VectorUint, VectorUint]

class tmap.ScalingType(self: tmap.ScalingType, arg0: int) → None¶

The scaling types available in OGDF. The class is to be used as an enum.

Notes

The available values are

ScalingType.Absolute: Absolute factor, can be used to scale relative to level size change.

ScalingType.RelativeToAvgLength: Scales by a factor relative to the average edge weights.

ScalingType.RelativeToDesiredLength: Scales by a factor relative to the disired edge length.

ScalingType.RelativeToDrawing: Scales by a factor relative to the drawing.

class tmap.Placer(self: tmap.Placer, arg0: int) → None¶

The places available in OGDF. The class is to be used as an enum.

Notes

The available values are

Placer.Barycenter: Places a vertex at the barycenter of its neighbors’ position.

Placer.Solar: Uses information of the merging phase of the solar merger. Places a new vertex on the direct line between two suns.

Placer.Circle: Places the vertices in a circle around the barycenter and outside of the current drawing

Placer.Median: Places a vertex at the median position of the neighbor nodes for each coordinate axis.

Placer.Random: Places a vertex at a random position within the smallest circle containing all vertices around the barycenter of the current drawing.

Placer.Zero: Places a vertex at the same position as its representative in the previous level.

class tmap.Merger(self: tmap.Merger, arg0: int) → None¶

The mergers available in OGDF. The class is to be used as an enum.

Notes

The available values are

Merger.EdgeCover: Based on the matching merger. Computes an edge cover such that each contained edge is incident to at least one unmatched vertex. The cover edges are then used to merge their adjacent vertices.

Merger.LocalBiconnected: Based on the edge cover merger. Avoids distortions by checking whether biconnectivity will be lost in the local neighborhood around the potential merging position.

Merger.Solar: Vertices are partitioned into solar systems, consisting of sun, planets and moons. The systems are then merged into the sun vertices.

Merger.IndependentSet: Uses a maximal independent set filtration. See GRIP for details.

class tmap.LayoutConfiguration(self: tmap.LayoutConfiguration) → None¶

A container for configuration options for layout_from_lsh_forest() and layout_from_edge_list().

int k

The number of nearest neighbors used to create the k-nearest neighbor graph.

Type: int

int kc

The scalar by which k is multiplied before querying the LSH forest. The results are then ordered decreasing based on linear-scan distances and the top k results returned.

Type: int

int fme_iterations

Maximum number of iterations of the fast multipole embedder.

Type: int

bool fme_randomize

Whether or not to randomize the layout at the start.

Type: bool

int fme_threads

The number of threads for the fast multipole embedder.

Type: int

int fme_precision

The number of coefficients of the multipole expansion.

Type: int

int sl_repeats

The number of repeats of the scaling layout algorithm.

Type: int

int sl_extra_scaling_steps

Sets the number of repeats of the scaling.

Type: int

double sl_scaling_min

The minimum scaling factor.

Type: float

double sl_scaling_max

The maximum scaling factor.

Type: float

ScalingType sl_scaling_type

Defines the (relative) scale of the graph.

Type: ScalingType

int mmm_repeats

Number of repeats of the per-level layout algorithm.

Type: int

Placer placer

The method by which the initial positons of the vertices at eachlevel are defined.

Type: Placer

Merger merger

The vertex merging strategy applied during the coarsening phaseof the multilevel algorithm.

Type: Merger

double merger_factor

The ratio of the sizes between two levels up to which the mergingis run. Does not apply to all merging strategies.

Type: float

int merger_adjustment

The edge length adjustment of the merging algorithm. Does notapply to all merging strategies.

Type: int

float node_size

The size of the nodes, which affects the magnitude of their repellingforce. Decreasing this value generally resolves overlaps in a verycrowded tree.

Type: float

Constructor for the class LayoutConfiguration.

__init__(self: tmap.LayoutConfiguration) → None¶: Constructor for the class LayoutConfiguration.

class tmap.GraphProperties(self: tmap.GraphProperties) → None¶

Contains properties of the minimum spanning tree (or forest) generated by layout_from_lsh_forest() and layout_from_edge_list().

mst_weight¶

The total weight of the minimum spanning tree.

Type: float

n_connected_components¶

The number of connected components in the minimum spanning forest.

Type: int

n_isolated_vertices¶

Type: int

degrees¶

The degrees of all vertices in the minimum spanning tree (or forest).

Type: VectorUint

adjacency_list¶

The adjaceny lists for all vertices in the minimum spanning tree (or forest).

Type: List of VectorUint

Constructor for the class GraphProperties.

__init__(self: tmap.GraphProperties) → None¶: Constructor for the class GraphProperties.