UMAP for dimensionality reduction in ML Python - Model Pipeline Trace

Model Pipeline - UMAP for dimensionality reduction

This pipeline uses UMAP to reduce the number of features in data while keeping its important structure. It helps us see and understand complex data by turning many features into just two or three.

Data Flow - 3 Stages

1Input Data

1000 rows x 50 columns→Raw data with 50 features per sample→1000 rows x 50 columns

[[0.5, 1.2, ..., 0.3], [1.1, 0.7, ..., 0.9], ...]

↓

2Preprocessing

1000 rows x 50 columns→Standardize features to zero mean and unit variance→1000 rows x 50 columns

[[-0.3, 0.8, ..., -1.1], [1.2, -0.5, ..., 0.4], ...]

↓

3UMAP Dimensionality Reduction

1000 rows x 50 columns→Reduce features from 50 to 2 using UMAP→1000 rows x 2 columns

[[1.5, -0.7], [0.3, 1.2], ...]

Training Trace - Epoch by Epoch

Loss
1.0 | *       
0.8 | **      
0.6 | ***     
0.4 | ****    
0.2 | *****   
    +---------
     1 2 3 4 5
     Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.85	N/A	Initial embedding with high loss, structure not clear
2	0.60	N/A	Loss decreased, clusters start to form
3	0.45	N/A	Better separation of groups visible
4	0.35	N/A	Embedding stabilizes, loss decreases slower
5	0.30	N/A	Final embedding with clear cluster structure

Prediction Trace - 3 Layers

Layer 1: Input Sample

Layer 2: Standardization

Layer 3: UMAP Projection

Model Quiz - 3 Questions

What does UMAP do in this pipeline?

A Increases the number of features

B Removes rows from the dataset

C Reduces data from many features to fewer features

D Changes data labels

Key Insight

UMAP helps us see complex data by turning many features into just a few while keeping the important patterns. Standardizing data first helps UMAP work better. The training loss going down shows the model finds a clearer view of the data step by step.