Skip to content

Latest commit

 

History

History
80 lines (60 loc) · 3.26 KB

README.md

File metadata and controls

80 lines (60 loc) · 3.26 KB
Raw

Neural Network for Regression Task

Overview

This repository contains an implementation of a modular neural network for a regression task, with a focus on predicting the median house value using the California House Prices Dataset. The project is divided into two main parts:

  1. Neural Network Mini-Library: A low-level neural network implementation built from scratch using NumPy.
  2. Regression Model: A neural network architecture designed for predicting house prices, trained using either PyTorch or the custom neural network library developed in Part 1.

Data

The dataset used for the regression task is the California House Prices Dataset, which contains information on various attributes of block groups in California, including:

  • Longitude
  • Latitude
  • Housing median age
  • Total rooms
  • Total bedrooms
  • Population
  • Households
  • Median income
  • Ocean proximity
  • Median house value (target variable)

The dataset is provided in CSV format and can be loaded and preprocessed using Pandas.

Features

Part 1: Neural Network Mini-Library

The first part of the project involves building a modular neural network library using NumPy, including the following components:

  • Linear Layer: Implements an affine transformation, with methods for forward pass, backward pass, and parameter updates.
  • Activation Functions: Includes Sigmoid and ReLU activation layers with forward and backward methods.
  • Multi-Layer Network: Combines multiple linear layers and activations into a neural network architecture.
  • Trainer: A class that handles data shuffling, minibatch gradient descent, and training the network over multiple epochs.
  • Preprocessor: A class that implements min-max scaling for input normalization.

Part 2: Regression Model

In the second part, a neural network is created and trained to predict house prices using either PyTorch or the custom mini-library. Key components include:

  • Data Preprocessing: Handling missing values, normalizing numerical data, and encoding categorical features (such as ocean proximity) using one-hot encoding.
  • Model Training: The model is trained using gradient descent, with options for minibatch training and shuffling.
  • Evaluation: Methods for model evaluation and prediction, using standard regression metrics.

Hyperparameter Tuning

A hyperparameter search is performed to find the optimal configuration for the neural network, including parameters such as learning rate, number of layers, and the number of neurons per layer.

Usage

  1. Clone the repository and navigate to the project directory.
  2. Ensure you have the necessary dependencies installed: 
    pip install -r requirements.txt
  3. Preprocess the dataset and train the model: 
    from src.part2_house_value_regression import Regressor
regressor = Regressor(x_train, nb_epoch=10)
regressor.fit(x_train, y_train)
  4. Save the trained model: 
    from src.part2_house_value_regression import save_regressor
save_regressor(regressor)
  5. Use the trained model to make predictions: 
    predictions = regressor.predict(x_test)

Dependencies

  • Python 3.x
  • NumPy
  • Pandas
  • Scikit-learn (for preprocessing utilities)
  • PyTorch (if using PyTorch for training)