AI-ordlista
Den kompletta ordlistan över AI
Deep Survival Analysis
Application of deep neural networks to model survival data, enabling the capture of complex non-linear relationships between covariates and event risk.
Time Censoring
Phenomenon where the event of interest is not observed for some units before the end of the study, creating incomplete data that requires specific analysis methods.
Hazard Function
Function describing the instantaneous rate of event occurrence at time t, conditional on survival up to that time, modeled differently in deep learning approaches.
DeepHit
Deep neural network architecture that directly models the discrete survival distribution without parametric assumptions on the shape of the hazard function.
Right Censoring
The most common type of censoring in survival analysis where the survival time is known to be greater than a certain observed value, but the exact value is unknown.
Survival Loss
Loss function specifically designed for survival analysis models, accounting for both observed event times and censored data in optimization.
Deep Cox Network
Extension of the Cox proportional hazards model using a neural network to learn a non-linear representation of covariates while maintaining the proportional hazards assumption.
Neural Kaplan-Meier
Deep learning approach that estimates the survival function by combining the flexibility of neural networks with the non-parametric properties of the Kaplan-Meier estimator.
Informative Censoring
Situation where the censoring mechanism is related to the event risk, violating the non-informative censoring assumption and requiring more sophisticated survival models.
Time-to-Event Prediction
Prediction task aimed at estimating the time until the occurrence of a specific event, using deep learning models to handle the complexity and censoring of data.
Survival Curve
Graphical representation of the probability of survival over time, estimated by deep learning models from training data and individual predictions.
Concordance Index (C-index)
Evaluation metric specific to survival analysis measuring the model's ability to correctly order event times between pairs of individuals.
Dynamic Deep Survival Models
Deep learning models that incorporate longitudinal or sequential data to continuously update survival predictions over time.
Competing Risks
Situation where multiple types of mutually exclusive events can occur, requiring multi-task deep learning models to estimate cause-specific risks.
Recurrent Neural Networks for Survival
Use of RNNs or LSTMs to model sequential survival data where covariates evolve over time before the occurrence of the event.
Attention-based Survival Models
Deep learning architectures using attention mechanisms to identify the most influential covariates on survival risk at different time points.
Frailty Models in Deep Learning
Extension of deep learning survival models incorporating random effects (frailty) to capture unobserved heterogeneity between individuals or groups.
SurvivalGAN
Generative adversarial networks specifically designed to synthesize realistic survival data preserving censoring characteristics and time distributions.
Multi-Task Survival Learning
Learning approach where the survival model is trained simultaneously on multiple related tasks to improve generalization and capture shared relationships.