Semi-supervised learning basics in ML Python - Model Pipeline Trace

Model Pipeline - Semi-supervised learning basics

Semi-supervised learning uses a small amount of labeled data and a large amount of unlabeled data to train a model. It helps the model learn better when labeling data is expensive or slow.

Data Flow - 5 Stages

1Input data

1000 rows x 10 columns→Dataset contains 100 labeled rows and 900 unlabeled rows→1000 rows x 10 columns

First 100 rows have labels like 'cat' or 'dog'; remaining 900 rows have no labels

↓

2Preprocessing

1000 rows x 10 columns→Normalize features and handle missing values→1000 rows x 10 columns

Feature values scaled between 0 and 1

↓

3Feature extraction

1000 rows x 10 columns→Extract meaningful features or embeddings→1000 rows x 5 columns

Reduced features representing shapes or colors

↓

4Model training

100 labeled rows x 5 columns + 900 unlabeled rows x 5 columns→Train model using labeled data and pseudo-labels from unlabeled data→Trained model

Model learns to classify cats and dogs using both labeled and guessed labels

↓

5Evaluation

Test set 200 rows x 5 columns→Measure accuracy and loss on unseen labeled data→Accuracy and loss scores

Accuracy = 85%, Loss = 0.35

Training Trace - Epoch by Epoch

Loss
0.9 |****
0.7 |*** 
0.5 |**  
0.3 |*   
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.85	0.55	Model starts learning from labeled and pseudo-labeled data
2	0.65	0.68	Loss decreases as model improves predictions
3	0.50	0.75	Accuracy improves steadily
4	0.40	0.80	Model benefits from unlabeled data guidance
5	0.35	0.85	Training converges with good accuracy

Prediction Trace - 4 Layers

Layer 1: Input sample

Layer 2: Feature transformation

Layer 3: Model prediction

Layer 4: Decision

Model Quiz - 3 Questions

Why does semi-supervised learning use unlabeled data?

A To replace labeled data completely

B Because unlabeled data is always more accurate

C To help the model learn patterns when labeled data is limited

D To increase the number of features

Key Insight

Semi-supervised learning improves model accuracy by using both labeled and unlabeled data, making it useful when labeled data is scarce or costly to obtain.