Description
Building a RNN with Keras
Objective: The goal of this assignment is to gain hands-on experience in designing and implementing a RNN using the Keras library in Python. You will select a dataset of your choice from kaggle and apply a deep learning model to it. The focus is on experimenting with different activation functions and optimization techniques to understand their impact on the model’s performance.
Requirements:
Dataset Selection:
Choose a dataset that interests you. It can be related to image classification, text analysis, or any other domain.
Perform necessary data preprocessing steps, such as normalization, encoding categorical variables, splitting into training and testing sets, etc.
Model Design:
Design a RNN using Keras.
Your network should have at least 3 hidden layers.
Activation Functions:
Experiment with three different activation functions: Sigmoid, Tanh, and ReLU.
Implement these activation functions in different models (or layers) to observe their effects on your network’s performance.
Optimization Technique:
Use any optimzer such as Adam, SGD as your optimization technique.
Discuss how your chosen optimizer influences the training process and outcomes.
Evaluation:
Evaluate the performance of your models using appropriate metrics (such as accuracy.).
Compare the results obtained with different activation functions and discuss your findings.
Report:
Write a report documenting your data preprocessing steps, model architecture, training process, results, and analysis.
Include visualizations such as loss and accuracy curves, confusion matrices, etc., where applicable.
Provide a link to the dataset.
Submission:
Submit your Jupyter Notebook containing the code, along with the report. Submit a ipynb file.
Note: Ensure that your code is well-commented and structured for readability and reproducibility. You will be evaluated based on the rubric provided.
