A general framework for neural network models on censored survival data.

Elia Biganzoli¹, Patrizia Boracchi, Ettore Marubini

¹Unità Operativa di Statistica Medica e Biometria, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan, Italy. biganzoli@istitutotumori.mi.it

Neural Networks : the Official Journal of the International Neural Network Society

|May 23, 2002

Summary

Flexible parametric methods using feed forward artificial neural networks (FFANNs) offer advanced analysis for censored time data. This study introduces novel error functions and data representations for FFANNs to better model complex survival data outcomes.

Area of Science:

Statistics
Machine Learning
Biostatistics

Background:

Censored time data analysis is crucial in many fields, including medicine and economics.
Traditional regression models may struggle with complex, non-linear relationships in survival data.
Existing feed forward artificial neural network (FFANN) methods have limitations in handling censored time data.

Purpose of the Study:

To extend FFANNs for the statistical analysis of censored time data.
To develop methods for detecting complex non-linear and non-additive effects in survival analysis.
To fill the gap in accounting for censored times within FFANN models.

Main Methods:

Introduction of specific error functions tailored for censored survival data.
Development of novel data representation techniques for FFANNs.
Application to multilayer perceptron and radial basis function extensions of generalized linear models.

Main Results:

The proposed methods enable FFANNs to effectively analyze censored time data.
Demonstrated capability to detect intricate non-linear and non-additive dependencies.
Improved modeling of outcome dependence on continuous variables in survival analysis.

Conclusions:

The novel FFANN approach provides a flexible and powerful tool for censored time data analysis.
This work enhances the application of neural networks in survival modeling.
The introduced techniques offer a promising direction for future research in statistical modeling of time-to-event data.

Related Experiment Videos

A general framework for neural network models on censored survival data.

Related Concept Videos

Main Outcomes of the HEBE Trial: Improving Cardiorespiratory Fitness and Body Composition Through a Tailored Feasible Lifestyle Program.

Application of integrated nested Laplace approximation to identify hot spots of methylation heterogeneity in healthy individuals from the MAMELI cohort.

Use of robust regression methods to detect outliers and estimate parameters for Length-Weight relationships in fishes.

Mucosal Melanoma of the Head and Neck: A 45-Year Experience of a Tertiary Cancer Center.

Comparison of Morbidity Between Sentinel Lymph Node Biopsy and Elective Neck Dissection in Dogs With Head and Neck Malignancies.

The Reasons to Get Vaccinated: A Cross-Sectional Study on HPV Vaccination Adherence in a Northern Italian University.

Injecting reasoning into vision-language models via weight-decomposed merging.

A Non-equilibrium thermodynamic framework for neural networks: A principled correspondence and parameter dynamics.

Dual-pathway mask ranking guided selective fine-tuning for backdoor purification.

Discriminative transfer feature learning for unsupervised domain adaptation.

ARetinex-Net: Low-light image enhancement with adaptive retinex model and global illumination representation.

A Multi-Task Learning Framework with Physics Embedded for Signal Reconstruction and State Prediction in Underground Multi-Robot Systems.

Improving Parameter Recovery in Computational Models: Employing Outlier-Insensitive Loss Functions.

Predictors of survival and functional outcomes after spinal surgery for multiple myeloma: a nationwide study of 200 patients.

A comparison of missing data approaches for linear regression with missing not at random outcome and predictors.

Estimation of comparable standardized mean differences in cluster randomized trials with covariate adjustment.

Outlier detection in state-space models using mean-shift penalisation.

The Analysis of Risk Factors of Stroke-Associated Pneumonia in Patients with Acute Stroke Based on Lasso Regression and the Construction of a Nomogram Prediction Model.

Related Experiment Videos

A general framework for neural network models on censored survival data.

Area of Science:

Background:

Purpose of the Study:

Main Methods:

Main Results:

Conclusions:

Related Concept Videos

Related Articles