optimization

🌸 DATA EXPANSION

Current data volume is not enough, and the split of ages is very rough, with 10 years as a group. Hence, acquisition of more data and a more precise age split are required in the future.

🌸 MODEL OPTIMIZATION

Parameters of the neural network (learning rate, activation function, regularization...), as well as the verbal repository of the genetic programming could be furthur optimized and added.

🌸 ENSEMBLE LEARNING

Application of more than one learners or GP functions to improve the model structure.

🌸 GENE INTERACTIONS

Integration of gene interactions as input features to get closer to internal conditions.

prospects

🌸 horizontal mining

Simimlar strategies can be applied to different tissues or organs, and even metabolic pathways. With adequate storage, models on expression of metabolic pathway related genes along ageing can be generated for the whole body from organic levels.

🌸 vertical mining

We can also consider adding interfering factors into the model or explore the expression level changes of transcriptiion factors along ageing to furthur investigate the model, revealing the global patterns of gene expression changes and the mechanisms beneath.

acknowledgement

Sincere gratitude to Professor Xiu-jie Wang for her tutelage.

Created by Oren | All Rights Reserved