Dataset from files in TensorFlow - Model Pipeline Trace

Model Pipeline - Dataset from files

This pipeline shows how to load data from files, prepare it, train a simple model, and make predictions. It starts by reading data files, then processes the data, trains a model, and finally predicts new results.

Data Flow - 5 Stages

1Load data from files

N/A→Read CSV files containing 1000 rows and 5 columns each→1000 rows x 5 columns

Row example: [5.1, 3.5, 1.4, 0.2, 'Iris-setosa']

↓

2Preprocessing

1000 rows x 5 columns→Parse CSV lines, convert strings to numbers, separate features and labels→1000 rows x 4 feature columns + 1000 labels

Features: [5.1, 3.5, 1.4, 0.2], Label: 0 (encoded Iris-setosa)

↓

3Train/test split

1000 rows x 4 features + 1000 labels→Split data into 800 training and 200 testing samples→Training: 800 rows x 4 features + 800 labels; Testing: 200 rows x 4 features + 200 labels

Training feature sample: [5.1, 3.5, 1.4, 0.2], Training label: 0

↓

4Model training

800 rows x 4 features→Train a neural network classifier with 3 output classes→Trained model weights

Model learns to classify Iris species

↓

5Prediction

1 row x 4 features→Model predicts class probabilities for new sample→1 row x 3 class probabilities

Input: [5.0, 3.6, 1.4, 0.2], Output: [0.95, 0.03, 0.02]

Training Trace - Epoch by Epoch

Loss
1.1 |*       
1.0 | *      
0.9 |  *     
0.8 |   *    
0.7 |    *   
0.6 |     *  
0.5 |      * 
0.4 |       *
0.3 |        *
    +---------
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.05	0.60	Model starts learning with moderate accuracy
2	0.75	0.75	Loss decreases, accuracy improves
3	0.55	0.82	Model continues to improve
4	0.40	0.88	Good convergence observed
5	0.30	0.92	Training nearing completion with high accuracy

Prediction Trace - 3 Layers

Layer 1: Input layer

Layer 2: Hidden layer (ReLU activation)

Layer 3: Output layer (Softmax activation)

Model Quiz - 3 Questions

What is the shape of the data after preprocessing?

A 1000 rows x 4 features + 1000 labels

B 1000 rows x 5 features

C 800 rows x 4 features

D 200 rows x 3 features

Key Insight

Loading data from files and preprocessing it correctly is essential for training a model that learns well. The training trace shows how loss decreases and accuracy increases, indicating the model is learning. Softmax outputs give clear probabilities for classification.