Hyperparameter Tuning and Sensitivity Analyses

library(discent)

hyperparameter tuning of learning rate, lambda
add some text about adam optimizer and more conservative vs aggressive inputs sensitivity analysis for starting parameter guesses

library(ggplot2)

# volcano is a matrix; convert to data frame
volcano_df <- as.data.frame(as.table(volcano))
names(volcano_df) <- c("x", "y", "z")

ggplot(volcano_df, aes(x, y, z = z)) +
  geom_contour_filled(bins = 15) +
  labs(fill = "Elevation") +
  theme_minimal()

# need to convert loss from final fis and M at various different levels
# and capture that landscape
# from your expand.grid start params