UMAP

Description

UMAP (Uniform Manifold Approximation and Projection) is a non-linear dimensionality reduction technique that creates 2D or 3D visualisations of high-dimensional data by constructing a mathematical model of the data's underlying manifold structure. Unlike t-SNE, UMAP preserves both local neighbourhood relationships and global topology more effectively, using techniques from topological data analysis and Riemannian geometry. This approach often produces more interpretable cluster layouts while maintaining meaningful distances between clusters, making it particularly valuable for exploratory data analysis and understanding complex dataset structures.

Example Use Cases

Explainability

Analysing single-cell RNA sequencing data to visualise how different cell types cluster based on gene expression patterns, revealing developmental trajectories and identifying previously unknown cell subtypes in tissue samples.

Exploring customer segmentation by reducing hundreds of behavioural and demographic features to 2D space, showing how different customer groups relate to each other and identifying transition zones where customers might move between segments.

Limitations

Hyperparameter choices (n_neighbors, min_dist, metric) significantly influence the embedding structure and can lead to very different interpretations of the same data.
While preserving global structure better than t-SNE, distances in the reduced space still don't directly correspond to distances in the original feature space.
Performance can be sensitive to the choice of distance metric, which may not be obvious for complex or mixed data types.
Like other manifold learning techniques, it assumes the data lies on a lower-dimensional manifold, which may not hold for all datasets.

Resources

lmcinnes/umap

Software Package

UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

Research Paper•Leland McInnes, John Healy, and James Melville•Feb 9, 2018

Uniform Manifold Approximation and Projection (UMAP) and its Variants: Tutorial and Survey

Documentation•Benyamin Ghojogh et al.•Aug 25, 2021

How UMAP Works — umap 0.5.8 documentation

Tutorial

Related Techniques

Name	Description	Assurance Goals
Classical Attention Analysis in Neural Networks	Classical attention mechanisms in RNNs and CNNs create alignment matrices and temporal attention patterns that show how models focus on different input elements over time or space. This technique analyses these traditional attention patterns, particularly in encoder-decoder architectures and sequence-to-sequence models, where attention weights reveal which source elements influence each output step. Unlike transformer self-attention analysis, this focuses on understanding alignment patterns, temporal dependencies, and encoder-decoder attention dynamics in classical neural architectures.	Explainability
Coefficient Magnitudes (in Linear Models)	Coefficient Magnitudes assess feature influence in linear models by examining the absolute values of their coefficients. Features with larger absolute coefficients are considered to have a stronger impact on the prediction, while the sign of the coefficient indicates the direction of that influence (positive or negative). This technique provides a straightforward and transparent way to understand the direct linear relationship between each input feature and the model's output.	Explainability Transparency
DeepLIFT	DeepLIFT (Deep Learning Important FeaTures) explains neural network predictions by decomposing the difference between the actual output and a reference output back to individual input features. It compares each neuron's activation to a reference activation (typically from a baseline input like all zeros or the dataset mean) and propagates these differences backwards through the network using chain rule modifications. Unlike gradient-based methods, DeepLIFT satisfies the sensitivity property (zero input gets zero attribution) and provides more stable attributions by using discrete differences rather than gradients.	Explainability Transparency
Attention Visualisation in Transformers	Attention Visualisation in Transformers analyses the multi-head self-attention mechanisms that enable transformers to process sequences by attending to different positions simultaneously. The technique visualises attention weights as heatmaps showing how strongly each token attends to every other token across different heads and layers. By examining these attention patterns, practitioners can understand how models like BERT, GPT, and T5 build contextual representations, identify which tokens influence predictions most strongly, and detect potential biases in how the model processes different types of input. This provides insights into positional encoding effects, head specialisation patterns, and the evolution of attention from local to global context across layers.	Explainability Fairness Transparency
Model Distillation	Model distillation transfers knowledge from a large, complex model (teacher) to a smaller, more efficient model (student) by training the student to mimic the teacher's behaviour. The student learns from the teacher's soft predictions and intermediate representations rather than just hard labels, capturing nuanced decision boundaries and uncertainty. This produces models that are faster, require less memory, and are often more interpretable whilst maintaining much of the original performance. Beyond compression, distillation can improve model reliability by regularising training and enable deployment in resource-constrained environments.	Explainability Reliability Safety
Generalized Additive Models	An intrinsically interpretable modelling technique that extends linear models by allowing flexible, nonlinear relationships between individual features and the target whilst maintaining the additive structure that preserves transparency. Each feature's effect is modelled separately as a smooth function, visualised as a curve showing how the feature influences predictions across its range. GAMs achieve this through spline functions or other smoothing techniques that capture complex patterns in individual variables without interactions, making them particularly valuable for domains requiring both predictive accuracy and model interpretability.	Transparency Explainability Reliability