Evidential Conditional Neural Processes
This work addresses uncertainty quantification for few-shot learning, which is important for model training and decision-making in data-scarce domains, though it represents an incremental improvement over existing CNP models.
The paper tackles the problem that Conditional Neural Process models only capture overall uncertainty without distinguishing between epistemic and aleatoric uncertainty in few-shot learning. The proposed Evidential Conditional Neural Processes achieve uncertainty decomposition and demonstrate robustness to noisy training tasks through extensive experiments on synthetic and real-world data.
The Conditional Neural Process (CNP) family of models offer a promising direction to tackle few-shot problems by achieving better scalability and competitive predictive performance. However, the current CNP models only capture the overall uncertainty for the prediction made on a target data point. They lack a systematic fine-grained quantification on the distinct sources of uncertainty that are essential for model training and decision-making under the few-shot setting. We propose Evidential Conditional Neural Processes (ECNP), which replace the standard Gaussian distribution used by CNP with a much richer hierarchical Bayesian structure through evidential learning to achieve epistemic-aleatoric uncertainty decomposition. The evidential hierarchical structure also leads to a theoretically justified robustness over noisy training tasks. Theoretical analysis on the proposed ECNP establishes the relationship with CNP while offering deeper insights on the roles of the evidential parameters. Extensive experiments conducted on both synthetic and real-world data demonstrate the effectiveness of our proposed model in various few-shot settings.