https://github.com/prachi-77/nas_reinforcementlearning
Demonstrate how Q-Learning, a reinforcement learning technique, can be used for Neural Architecture Search (NAS) to improve the performance of a CNN model.
https://github.com/prachi-77/nas_reinforcementlearning
Last synced: 3 months ago
JSON representation
Demonstrate how Q-Learning, a reinforcement learning technique, can be used for Neural Architecture Search (NAS) to improve the performance of a CNN model.
- Host: GitHub
- URL: https://github.com/prachi-77/nas_reinforcementlearning
- Owner: prachi-77
- Created: 2025-02-12T07:53:27.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-02-12T08:00:17.000Z (3 months ago)
- Last Synced: 2025-02-12T08:51:21.641Z (3 months ago)
- Language: Jupyter Notebook
- Size: 0 Bytes
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Neural Architecture Search (NAS) with Q-Learning
This notebook demonstrates how Q-Learning, a reinforcement learning technique, can be used for Neural Architecture Search (NAS) to improve the performance of a FashionMNIST CNN model.
## Key Steps in This Notebook:
### 1. Train a Baseline CNN Model:
- Load and preprocess the FashionMNIST dataset.
- Define and train a simple CNN model.
- Evaluate and visualize the initial model performance.
### 2. Define the Q-Learning Setup:
- Define the state space (different CNN architectures).
- Define the action space (modifying filter size, kernel size).
- Initialize the Q-table randomly.
### 3. Implement Q-Learning for NAS:
- Use an epsilon-greedy policy to balance exploration and exploitation.
- Iterate over 10 episodes to update architectures based on validation accuracy.
- Update the Q-table using the Bellman equation.
### 4. Visualizations and Results:
- Compare the pre-NAS and post-NAS architectures.
- Display the Q-table evolution over episodes.
- Show improvements in model accuracy.
By the end of this notebook, you will see how Q-learning helps dynamically search for an optimal CNN architecture to improve classification performance on FashionMNIST.