Cito Workshop

Install packages:

devtools::install_github('citoverse/cito')
devtools::install_github(repo = "TheoreticalEcology/EcoData", dependencies = F, build_vignettes = F)

1 - Introduction to cito

• How to fit simple fully-connected neural networks
• Adjusting learning rate and epochs to improve convergence
• Regularization in cito
• Automatic hyperparameter tuning in cito

01_Intro.R

2 - Inference with cito (using explainable AI)

• PDP, ALE, Feature Importance, and Average conditional effects in cito
• Confidence intervals

02_xAI.R

3 - Convolutional neural networks

• Multi-class task - predicting flower species from images

03_CNN.R

Trouble shooting for the flower dataset:

Sometimes EcoData::dataset_flower() cannot download the data automatically (depends on the OS and security settings). The dataset can also be downloaded manually from https://www.dropbox.com/s/slgml71w22of4n4/flower.rda?raw=1

4 - Bonus

• Embeddings
• Likelihoods and residual checks

04_Embeddings.R

• Multi-Species SDM

05_MultiSpecies.R