Uncertainty Estimation and Calibration with Finite-State Probabilistic RNNs
This work addresses the problem of reliable uncertainty quantification and calibration for RNNs, which is important for building trustworthy machine learning systems across various applications.
This paper proposes a method to estimate uncertainty in recurrent neural networks (RNNs) by introducing stochastic discrete state transitions. The model's uncertainty is quantified by repeatedly sampling from these transitions during prediction, leading to varied outcomes when uncertainty is high.
Uncertainty quantification is crucial for building reliable and trustable machine learning systems. We propose to estimate uncertainty in recurrent neural networks (RNNs) via stochastic discrete state transitions over recurrent timesteps. The uncertainty of the model can be quantified by running a prediction several times, each time sampling from the recurrent state transition distribution, leading to potentially different results if the model is uncertain. Alongside uncertainty quantification, our proposed method offers several advantages in different settings. The proposed method can (1) learn deterministic and probabilistic automata from data, (2) learn well-calibrated models on real-world classification tasks, (3) improve the performance of out-of-distribution detection, and (4) control the exploration-exploitation trade-off in reinforcement learning.