Mosquito Age Predictor to Speed Up Malaria Research
An AI tool was recently made to predict the age of mosquitoes.
Created on January 26|Last edited on January 26
Comment
A recent paper from BMC Bioinformatics shows that their AI tool can predict the age of mosquitoes (binary classification, as far as I'm aware) with up to 98% accuracy. From an initial glance, they trained with a simple MLP and with a CNN. They also applied dimensionality reduction (PCA or t-SNE) for the MLP case and no DR for the CNN experiment.
They also tested a wide variety of ML models like random forests, SVM, and XGBoost.
Interestingly, their best-performing pipeline was an XGBoost, termed XGB-6 for the version, with PCA dimensionality reduction and transfer learning. Their data was sourced from 2 places: Ifakara insectary and Glasglow insectary.
This best-performing pipeline achieved 90% accuracy on an Ifakara insectary test set and 98% on a Glasglow insectary test set! It also looks like PCA was much better than t-SNE.
Interestingly, CNN-2 from their results table shows even better performance, yet it was highlighted as the best-performing result. Perhaps there was a data leakage or some test set overfitting?
This paper demonstrates AI's versatility in other research fields like biology. Leveraging AI tools can help expedite the research rate and minimize the amount of monotonous or tedious work a researcher has to do.
Reference
X, Science. “A New AI Tool Can Predict Mosquitoes' Ages with 98% Accuracy to Speed Malaria Research.” Phys.org, Phys.org, 25 Jan. 2023
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.