MultiTimesProject

This project implements a contrastive learning framework for time-series signal (process signal), with specific downstream classification and regression tasks. The repo is organized into several modules to facilitate pretraining, classification, and regression workflows.

---

Table of Contents

1. Overview
2. Dependencies and Setup
3. Modules and Usage
   • Contrastive Learning Pretraining
   • Transformer Classifier Training
   • Contrastive Model Finetuning for Classification
   • Contrastive Model Finetuning for Regression
4. Results and Metrics
5. Future Work

---

Overview

This project leverages contrastive learning to pretrain representations for time-series signals and applies them to downstream classification and regression tasks. The key modules of this project are:

• Contrastive Learning Pretraining: To learn representations of time-series signals.
• Classifier Training from Scratch: To train a classifier model without using pretraining.
• Finetuning Pretrained Models: To finetune pretrained contrastive learning models on specific tasks like classification and regression.

---

Dependencies and Setup

Ensure you have the following installed:

• Python 3.11
• Necessary libraries (see at requirements.txt)

Set up the environment by cloning this repository and navigating to the project folder:

git clone <repository-url>
cd <repository-folder>
pip install -r requirements.txt

## Modules and Usage

### 1. Contrastive Learning Pretraining

To perform contrastive learning pretraining, run:

```bash
python runCL.py

### 2. Transformer Classifier Training
To train a transformer-based classifier from scratch, run:

python runPretrainClassifier.py