CombustionNet

Design a Benchmark Dataset for Deep Learning in Combustion Research Inspired by ImageNet. This project is motivated to design a benchmark dataset serves as MNIST and ImageNet for the research and education of deep learning in combustion. The problem should be easy to understand and anyone without much combustion knowledge can set up the problem within couple hours. Meanwhile, the problem should be challanging in a sense that a mature deep learning model won't work that much perfert, such that there is a lot of room to improve the model.

Collection of existing dataset

** ECNet: scalable, retrainable and deployable machine learning projects for fuel property prediction **

Brief descriptions of the databases:

• cn_database_v1.0.csv: cetane number database containing 482 molecules from 11 compound groups, each with an experimental cetane number and 5305 QSPR descriptors calculated using alvaDesc
• cn_database_v1.1.csv: cetane number database containing 482 molecules from 11 compound groups, each with an experimental cetane number and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• cp_database_v1.0.csv: cloud point database containing 43 molecules, each with an experimental cloud point value and 5305 QSPR descriptors calculated using alvaDesc
• cp_database_v1.1.csv: cloud point database containing 43 molecules, each with an experimental cloud point value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• kv_database_v1.0.csv: kinematic viscosity database containing 216 molecules, each with an experimental kinematic viscosity value and 5305 QSPR descriptors calculated using alvaDesc
• kv_database_v1.1.csv: kinematic viscosity database containing 216 molecules, each with an experimental kinematic viscosity value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• mon_database_v1.0.csv: motor octane number database containing 308 molecules, each with an experimental MON value and 5305 QSPR descriptors calculated using alvaDesc
• mon_database_v1.1.csv: motor octane number database containing 307 molecules, each with an experimental MON value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• pp_database_v1.0.csv: pour point database containing 41 molecules, each with an experimental pour point value and 5305 QSPR descriptors calculated using alvaDesc
• pp_database_v1.1.csv: pour point database containing 41 molecules, each with an experimental pour point value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• ron_database_v1.0.csv: research octane number database containing 308 molecules, each with an experimental RON value and 5305 QSPR descriptors calculated using alvaDesc
• ron_database_v1.1.csv: research octane number database containing 307 molecules, each with an experimental RON value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• s_database_v1.0.csv: octane sensitivity database containing 308 molecules, each with an experimental octane sensitivity value and 5305 QSPR descriptors calculated using alvaDesc
• s_database_v1.1.csv: octane sensitivity database containing 307 molecules, each with an experimental octane sensitivity value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• ysi_database_v1.0.csv: unified yield sooting index database containing 421 molecules, each with an experimental unified yield sooting index value and 5305 QSPR descriptors calculated using alvaDesc
• ysi_database_v1.1.csv: unified yield sooting index database containing 421 molecules, each with an experimental unified yield sooting index value and 1875 QSPR descriptors calculated using PaDEL-Descriptor
• db_template.csv: ECNet-formatted database template