Difference between revisions of "Python-for-Machine-Learning/C3/Artificial-Neural-Networks/English"

Learning Objectives

• The fundamentals of Artificial Neural Networks (ANN)
• Implementing Multi- Layer Perceptron (MLP) classification
• Evaluating the performance of a trained MLP model.

• Ubuntu Linux 24.04
• Jupyter Notebook IDE

The learner must have basic knowledge of Python.

For pre-requisite Python tutorials, please visit this website.

Code-Files

• The files used in this tutorial are provided in the Code files link.
• Please download and extract the files.
• Make a copy and use them while practicing.

Artificial Neural Network

• Artificial Neural Networks are models inspired by the human brain’s processing.
• Neurons receive inputs, adjust weights, and pass outputs to other neurons.
• ANNs have input, hidden, and output layers to process and learn patterns.

Artificial Neural Network

• Each connection in ANN has a weight that helps it to make better predictions.
• ANNs learn by adjusting weights based on errors to improve accuracy.

Multi-Layer Perceptron

• MLP is a type of ANN with multiple hidden layers that enhance feature learning.
• It processes data through interconnected neurons in a forward direction.
• MLP is widely used for classification and regression tasks in machine learning.

ANN Architecture

arch.png

The Input Layer receives the data from the dataset.

The number of input neurons matches the number of input features.

Hidden Layers process these inputs through weighted connections.

These layers help the network learn complex patterns.

The number of hidden neurons depends on task complexity.

ANN Architecture

arch.png

The number of output neurons matches the number of output classes.

Weights are updated during training to optimize performance.

Artificial Neuron Model

neuron.png

Each neuron receives inputs along with a bias value.

Each input has a weight that determines its importance.

The summation function adds the weighted inputs and bias.

The activation function decides the neuron’s output.

The output is then passed to the next layer in the network.

Type conda activate ml

Press Enter

Activate the machine learning environment as shown.

Type cd Downloads

Type jupyter notebook

Please navigate to the respective folder of your code file location.

Then type, jupyter space notebook and press Enter.

Artificial Neural Networks.ipynb file

Click the Artificial neural networks dot ipynb file to open it.

Note that each cell will have the output displayed in this file.

import pandas as pd import numpy as np import matplotlib.pyplot as plt import seaborn as sns

Make sure to Press Shift and Enter to execute the code in each cell.

We will use the Breast Cancer Wisconsin dataset from sklearn library.

The dataset has 30 features describing breast tumor characteristics.

The target variable is 0 for malignant tumors and 1 for benign tumors.

data = load_breast_cancer()

Then, we create a dataframe using a pd dot dataframe for data handling.

bcancer_df.tail()

Highlight

bcancer_df.shape

print("Feature names:", data.feature_names)

print("Target names:", data.target_names)

plt.figure(figsize=(12,6)) plt.show()

Malignant tumors are shown in red and benign tumors in green.

Outliers are values far from the rest of the data and appear as small dots.

In this plot, the malignant class has a higher mean radius with more variation.

The benign class has a lower mean radius with fewer outliers.

Highlight

scaler = RobustScaler() df_scaled = pd.DataFrame( scaler.fit_transform(bcancer_df.iloc[:, :-1]),

To handle outliers, we normalize the features using Robust Scaler.

It scales values based on the median and IQR, making it resistant to outliers.

X = bcancer_df.drop(columns=["target"]) y = bcancer_df['target']

The y variable stores the target classes for classification.

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)

mlp_relu = MLPClassifier(hidden_layer_sizes=(100, 50), activation='relu', max_iter=1000, random_state=42)

mlp_relu.fit(X_train, y_train)

Using the MLPClassifier function, we define mlp underscore relu.

It has two hidden layers, the first with 100 neurons and second with 50 neurons.

The model uses the Rectified Linear Unit that is ReLU activation function for faster convergence.

ReLU helps the MLP to train effectively.

The model is trained for a maximum of 1000 iterations to optimize performance.

y_train_pred_relu = mlp_relu.predict(X_train)

print("Training Accuracy (ReLU):", format(accuracy_score(y_train, y_train_pred_relu), ".3f"))

The result is formatted to three decimal places and printed.

This indicates that the model has learned well from the training data.

plt.plot(mlp_relu.loss_curve_, label="Training Loss", color="blue")

plt.xlabel("Iterations")

plt.ylabel("Loss")

The x-axis represents iterations, while the y-axis shows the loss value.

A decreasing loss curve indicates effective training.

Initially, the loss is high but quickly drops, showing rapid learning.

After 20 iterations, the loss stabilizes, indicating model convergence.

Model convergence means training has optimized weights, with minimal further gain.

y_pred_relu = mlp_relu.predict(X_test)

accuracy_relu = accuracy_score(y_test, y_pred_relu)

We calculate and print the testing accuracy.

This indicates strong generalization to unseen data.

y_probs = mlp_relu.predict_proba(X_test)[:, 1]

fpr, tpr, _ = roc_curve(y_test, y_probs)

roc_auc = auc(fpr, tpr)

The curve shows the balance between true positive rate and false positive rate.

print("\nMLP with ReLU activation - Classification Report:")

The other metrics like recall and F1-score suggest that it correctly identifies most positive instances.

Thus, the model has learnt to classify whether the patient has breast cancer or not.

In this tutorial, we have learnt about

• The fundamentals of Artificial Neural Networks (ANN)
• Implementing Multi- Layer Perceptron (MLP) classification
• Evaluating the performance of a trained MLP model

Assignment

• Use activation='logistic' instead of activation='relu'
• Evaluate the performance using accuracy and classification report for test set

Assignment Solution

FOSSEE Forum

Machine Learning, visit the FOSSEE forum and post your question.

Thank You

Thanks for joining.