Multi-input and multi-output models in TensorFlow - Model Pipeline Trace

Model Pipeline - Multi-input and multi-output models

This pipeline shows how a model can learn from two different inputs at the same time and give two different outputs. It’s like using two senses to understand something and then giving two answers based on that understanding.

Data Flow - 5 Stages

1Raw Data

1000 rows x 2 inputs→Two separate input features collected: Input A (10 features), Input B (5 features)→1000 rows x 2 inputs (10 features and 5 features)

Input A: [0.5, 1.2, ..., 0.3], Input B: [3, 0, 1, 2, 1]

↓

2Preprocessing

1000 rows x 2 inputs→Normalize Input A and one-hot encode Input B→1000 rows x 2 inputs (10 normalized features and 5 one-hot features)

Input A normalized: [0.05, 0.12, ..., 0.03], Input B one-hot: [0,1,0,0,0]

↓

3Feature Engineering

1000 rows x 2 inputs→Separate dense layers for each input to extract features→1000 rows x 2 feature vectors (each 8 features)

Feature vector A: [0.1, 0.2, ..., 0.05], Feature vector B: [0.3, 0.1, ..., 0.2]

↓

4Concatenate Features

1000 rows x 2 feature vectors→Combine feature vectors into one vector→1000 rows x 16 features

[0.1, 0.2, ..., 0.05, 0.3, 0.1, ..., 0.2]

↓

5Model Trains

1000 rows x 16 features→Train model with two outputs: Output 1 (regression), Output 2 (classification)→Two outputs: Output 1 shape (1000 rows x 1), Output 2 shape (1000 rows x 3 classes)

Output 1: [2.5, 3.1, ..., 1.2], Output 2: [[0.1, 0.7, 0.2], [0.8, 0.1, 0.1], ...]

Training Trace - Epoch by Epoch

Loss
1.2 |*       
1.0 | *      
0.8 |  *     
0.6 |   *    
0.4 |    *   
0.2 |     *  
0.0 +--------
      1 3 5 7 10 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.2	0.45	Loss starts high; accuracy low as model begins learning
3	0.8	0.60	Loss decreases; accuracy improves as model learns patterns
5	0.5	0.75	Loss continues to drop; accuracy rises steadily
7	0.35	0.82	Model converging; loss low and accuracy high
10	0.25	0.88	Training stabilizes with good accuracy and low loss

Prediction Trace - 4 Layers

Layer 1: Input Layer

Layer 2: Separate Dense Layers

Layer 3: Concatenate Features

Layer 4: Output Layers

Model Quiz - 3 Questions

Why does the model have two separate inputs?

A To reduce the number of features

B To learn from different types of data separately

C To make training faster

D To avoid using activation functions

Key Insight

Multi-input and multi-output models allow learning from different data sources simultaneously and produce multiple predictions. This helps solve complex problems where one input or output is not enough.