docs: add roadmap for advanced survival models and extend TODO

Author: DiogoRibeiro7

TODO.md

@@ -60,4 +60,17 @@ This document outlines future enhancements, features, and ideas for improving th
 
 - [ ] Add performance benchmarks for each model
 - [ ] Improve PyPI discoverability (add keywords)
-- [ ] Create a Streamlit or Gradio live demo
+- [ ] Create a Streamlit or Gradio live demo
+
+---
+
+## 🧠 8. New Survival Models to Implement
+
+- [ ] Accelerated Failure Time (AFT) models:
+  - [ ] Log-Normal AFT
+  - [ ] Log-Logistic AFT
+  - [ ] Weibull AFT formulation
+- [ ] Piecewise Exponential Model
+- [ ] Competing Risks simulation
+- [ ] Recurrent Events simulation
+- [ ] Mixture Cure Model

docs/source/theory.md

@@ -81,6 +81,37 @@ $$
 
 ---
 
+
+## 5. Accelerated Failure Time (AFT) Models
+
+AFT models assume that the effect of covariates accelerates or decelerates time to event directly, rather than the hazard.
+
+### Log-Normal AFT
+
+The model assumes:
+
+$$
+\log(T_i) = X_i \beta + \varepsilon_i, \quad \varepsilon_i \sim \mathcal{N}(0, \sigma^2)
+$$
+
+Thus:
+
+$$
+T_i \sim \text{Log-Normal}(X_i \beta, \sigma^2)
+$$
+
+The survival function is:
+
+$$
+S(t \mid X) = 1 - \Phi\left(\frac{\log(t) - X \beta}{\sigma}\right)
+$$
+
+Where:
+
+- \( \Phi \) is the standard normal cumulative distribution function (CDF)
+
+This model is especially useful when the proportional hazards assumption is not valid and provides interpretable effects in the time domain.
+
 ## Notes
 
 All models support censoring: