# Dimensionality Reduction

> Reduce feature dimensions: PCA, t-SNE, UMAP, SVD, and autoencoders. Use for visualization of high-dimensional data, noise reduction, feature compression, and as preprocessing for downstream ML models.

- **Type:** Skill
- **Install:** `agentstack add skill-ericwang915-data-scientist-skills-dimensionality-reduction`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [ericwang915](https://agentstack.voostack.com/s/ericwang915)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [ericwang915](https://github.com/ericwang915)
- **Source:** https://github.com/ericwang915/data-scientist-skills/tree/main/ds-machine-learning/skills/dimensionality-reduction

## Install

```sh
agentstack add skill-ericwang915-data-scientist-skills-dimensionality-reduction
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Dimensionality Reduction

## Purpose
Reduce the number of features while preserving important structure. Essential for visualization, denoising, and preprocessing.

## How It Works

### Method Selection

| Method | Preserves | Best For | Linear? |
|--------|-----------|----------|---------|
| PCA | Global variance | Feature compression, denoising | Yes |
| t-SNE | Local structure | 2D/3D visualization | No |
| UMAP | Local + global | Visualization, clustering prep | No |
| SVD | Variance | Sparse data, NLP (LSA) | Yes |
| LDA | Class separation | Supervised dimensionality reduction | Yes |
| Autoencoder | Learned representation | Complex non-linear compression | No |

### PCA Workflow
1. Standardize features
2. Compute covariance matrix and eigenvalues
3. Choose components: explained variance ≥ 85-95%
4. Transform and validate (scree plot, biplot)

### t-SNE / UMAP Workflow
1. Apply PCA first if >50 features (speed)
2. Tune perplexity (t-SNE) or n_neighbors (UMAP)
3. Generate 2D/3D embedding
4. Color by labels or clusters for interpretation

## Usage Examples

```
"Visualize this 50-feature customer dataset in 2D to see if
natural clusters exist"
```

```
"Reduce 200 features to the most important 20 using PCA
before training a model"
```

## Output Format

- **Method Choice**: Rationale for selected approach
- **Explained Variance**: Scree plot, cumulative variance
- **Visualization**: 2D/3D scatter plot of reduced space
- **Python Code**: sklearn / umap-learn implementation

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [ericwang915](https://github.com/ericwang915)
- **Source:** [ericwang915/data-scientist-skills](https://github.com/ericwang915/data-scientist-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-ericwang915-data-scientist-skills-dimensionality-reduction
- Seller: https://agentstack.voostack.com/s/ericwang915
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.