This repository provides a collection of use cases and examples demonstrating the application of Curie for automated and rigorous scientific experimentation with AI agents.
The Curie framework is designed to facilitate scientific discovery across various domains by automating the experimentation process. This repository contains specific implementations and experiments that showcase how Curie can be applied to different problems, particularly in machine learning and stock prediction.
-
🤖 General Machine Learning Experiments: Contains use cases related to machine learning experiments.
q1_activation_func: Experiments with different activation functions in neural networks.q2_dog-breed-identification: Image classification task for identifying dog breeds.q3_siim-isic-melanoma-classification: Classification task for melanoma detection in medical images.q4_aptos2019-blindness-detection: Classification task for detecting diabetic retinopathy in eye images.q5_histopathologic-cancer-detection: Classification task for detecting metastatic cancer.
-
📈 Stock Price Prediction Use Case: Put machine learning to work on real-world financial data (note: raw data not published due to privacy constraints):
q0_general_optimize: Explore general optimization strategies to tune your models for peak performance.q1_optimize_hyperparameter: Test out different hyperparameter optimizations for stock prediction performance.q2_feature_selection: Test out different feature selection parameters to improve stock prediction accuracy.q3_feature_engineering: Test out different feature engineering techniques for stock prediction.q4_ensemble: Test out different ensemble methods for improving stock prediction accuracy.starter_code: Starter code and templates for stock prediction experiments.
-
🧬 Bioinformatics Experiments: Contains use cases related to bioinformatics and computational biology:
enhance_st_resolution: Experiments for enhancing spatial transcriptomics data resolution using deep learning approaches.infer_super_resol_tissue_arch: Experiments for inferring tissue architecture from super-resolution data.