Scribd
Upload
5 views5 pages

Lab 8

Uploaded by

raavimalleswari95
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
5 views5 pages

Lab 8

Uploaded by

raavimalleswari95
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 5

Module-4

Report for the MNIST Classification Program:


Below is the step-by-step approach to understanding, optimizing, and evaluating the given
MNIST classification program using TensorFlow and Keras.
Step 1: Necessary Imports and Configuration
 We import necessary libraries such as sys for system configurations, tensorflow for
building neural networks, and matplotlib for plotting accuracy graphs.
 The program forces UTF-8 encoding to ensure compatibility in different
environments.
Step 2: Loading and Preprocessing the MNIST Dataset
Load the Data:
 Use mnist.load_data() to load the dataset, which includes training and testing data
split into images (X_train, X_test) and corresponding labels (y_train, y_test).
Normalize the Input Data:
 Convert the pixel values to floats and normalize them to a range between 0 and 1 by
dividing by 255.0.
One-Hot Encode the Labels:
 Convert the integer class labels to one-hot encoded vectors using to_categorical,
which is essential for categorical cross-entropy loss during model training.
Step 3: Data Inspection
 Print the shapes of the training and testing datasets to ensure they are loaded and
preprocessed correctly.
Step 4: Model Creation and Summary
Initialize Sequential Model:
 Create a sequential model which allows stacking layers linearly.
Add Layers:
 Flatten Layer: Converts 2D images (28x28 pixels) into 1D vectors.
 Dense Layer: A fully connected layer with 128 units and ReLU activation for feature
extraction.
 Output Layer: Another dense layer with 10 units and softmax activation, representing
the 10 classes of digits (0-9).
Model Summary:
 Print the model summary to visualize the layer architecture and parameter counts.
Step 5: Model Compilation
 Compile the model with:
 Optimizer: adam for efficient training.
 Loss Function: categorical_crossentropy to match the one-hot encoded labels.
 Metrics: accuracy to measure how well the model is performing.
Step 6: Model Training
Train the Model:
 Fit the model on the training data for 10 epochs, using a batch size of 32.
 Split the training set further into an 80/20 split for training/validation during training.
Step 7: Model Evaluation and Accuracy Plotting
Evaluate Model:
 Use the test set to evaluate the model's performance with model.evaluate and print the
test accuracy.
Training History:
 Plot the training and validation accuracy over epochs using matplotlib to visually
inspect how well the model performed during training.
Optimization Considerations
Optimize Data Loading:
 Ensure the data loading process is efficient, and consider using data augmentation if
the model overfits.
Model Architecture:
 Experiment with deeper networks or more advanced architectures like Convolutional
Neural Networks (CNNs) for potentially better performance on image data.
Hyperparameter Tuning:
 Tune hyperparameters such as learning rate, batch size, and the number of epochs
through techniques like grid search or random search for optimal results.
Regularization:
 Implement dropout layers or L2 regularization to prevent overfitting, especially if
more complex models are used.
Learning Rate Schedulers:
 Utilize learning rate schedulers to adjust the learning rate dynamically during training
for better convergence.
Early Stopping:
 Implement early stopping to prevent over-training if the validation accuracy does not
improve over several epochs.
Conclusion
This program provides a basic yet effective approach to classifying MNIST digits using a
neural network. By considering the optimization suggestions, the performance and efficiency
can be improved, making the model more robust for competitive programming challenges.

Module-7
Report for the Iris Dataset Classification Program
Introduction:
The provided program demonstrates the process of loading the Iris dataset, preprocessing it,
and applying two types of machine learning models: Logistic Regression and Multi-layer
Perceptron (MLP) Neural Network. The program evaluates the models using accuracy
metrics and visualizes the results using confusion matrices.
Step-by-Step Approach
Import Necessary Libraries:
 The libraries numpy and pandas for data manipulation, scikit-learn for machine
learning models, and matplotlib and seaborn for plotting.
Load the Dataset:
 Use load_iris() from scikit-learn to load the Iris dataset into variables X (features) and
y (target labels).
Split the Dataset:
 Split the data into training and testing sets using train_test_split with a 70/30 train-test
ratio and a random state for reproducibility.
Standardize the Features:
 Standardize the feature values to have a mean of 0 and a standard deviation of 1 using
StandardScaler. This step is crucial for optimizing the performance of many machine
learning algorithms.
Logistic Regression Model:
Initialize and Train:
 Initialize LogisticRegression with a maximum of 200 iterations and train it using the
training data.
Predict:
 Predict the classes for the test data.
Evaluate:
 Print the confusion matrix, classification report, and precision and recall scores to
evaluate model performance.
Neural Network Model (MLP):
Initialize and Train:
 Initialize MLPClassifier with three hidden layers of 10 neurons each, and a maximum
of 1000 iterations.
Predict:
 Predict the classes for the test data.
Evaluate:
 Print the confusion matrix, classification report, and precision and recall scores to
evaluate model performance.
Visualization:
 Plot confusion matrices for both Logistic Regression and MLP models using seaborn
heatmaps for better visualization of the results.
Results and Evaluation
Logistic Regression:
Confusion Matrix and Classification Report:
 The confusion matrix shows the true positive, false positive, true negative, and false
negative counts for each class.
 The classification report provides detailed metrics such as precision, recall, and F1-
score for each class.
 Precision and recall scores are also printed for an overall understanding of model
performance.
Neural Network (MLP):
Confusion Matrix and Classification Report:
 Similarly, the MLP confusion matrix and classification report offer insights into the
model's performance across different classes.
Plots:
 Two confusion matrices are plotted side-by-side for easy comparison of the
performance of Logistic Regression and MLP models.
Optimization Considerations:
Hyperparameter Tuning:
 For both models, consider using techniques such as grid search or random search to
find the optimal hyperparameters.
Feature Engineering:
 Examine the features for possible transformations or interactions that could improve
model performance.
Advanced Models:
 Test more sophisticated models such as Support Vector Machines (SVM), Random
Forests, or Gradient Boosted Trees.
Cross-validation:
 Use cross-validation to ensure the model’s performance is robust and generalizes well
to unseen data.
Conclusion:
This program effectively demonstrates the process of loading, preprocessing, training,
evaluating, and visualizing the performance of machine learning models on the Iris dataset.
By considering optimization strategies, the performance can be further enhanced.

You might also like