Reliable neural networks for regression uncertainty estimation

Tohme, Tony; Vanslette, Kevin

doi:10.1016/j.ress.2022.108811

Cited by 9 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Extending our method to more general tasks with continuous action space would be a valuable direction for future investigation. To scale USN to continuous action space, regression uncertainty estimation methods might be helpful for sample selection (Gustafsson et al, 2023;Tohme et al, 2023). Secondly, USN only selects samples based on the positive correlation between loss and uncertainty estimations when the noise rate increases.…”

Section: Discussionmentioning

confidence: 99%

USN: A Robust Imitation Learning Method against Diverse Action Noise

Yu,

Han,

Tsang

2024

jair

View full text Add to dashboard Cite

Learning from imperfect demonstrations is a crucial challenge in imitation learning (IL). Unlike existing works that still rely on the enormous effort of expert demonstrators, we consider a more cost-effective option for obtaining a large number of demonstrations. That is, hire annotators to label actions for existing image records in realistic scenarios. However, action noise can occur when annotators are not domain experts or encounter confusing states. In this work, we introduce two particular forms of action noise, i.e., state-independent and state-dependent action noise. Previous IL methods fail to achieve expert-level performance when the demonstrations contain action noise, especially the state-dependent action noise. To mitigate the harmful effects of action noises, we propose a robust learning paradigm called USN (Uncertainty-aware Sample-selection with Negative learning). The model first estimates the predictive uncertainty for all demonstration data and then selects sampleswith high loss based on the uncertainty measures. Finally, it updates the model parameters with additional negative learning on the selected samples. Empirical results in Box2D tasks and Atari games show that USN consistently improves the final rewards of behavioral cloning, online imitation learning, and offline imitation learning methods under various action noises. The ratio of significant improvements is up to 94.44%. Moreover, our method scales to conditional imitation learning with real-world noisy commands in urban driving

show abstract

Section: Discussionmentioning

confidence: 99%

USN: A Robust Imitation Learning Method against Diverse Action Noise

Yu,

Han,

Tsang

2024

jair

View full text Add to dashboard Cite

show abstract

“…test cases), as well as more recent works (Stephens 2015;Cava et al 2019;La Cava et al 2019;Virgolin et al 2019;Cranmer et al 2020;Cranmer 2020;Kommenda et al 2020;de Franca & Aldeia 2021;Virgolin et al 2021). In addition, SR has been implemented using various methods ranging from brute force to (un-)guided Monte Carlo, all the way to probabilistic searches (McConaghy 2011;Jin et al 2019;Kammerer et al 2020;Brence et al 2021;Bartlett et al 2023), as well as through problem simplification algorithms (Luo et al 2022;Tohme et al 2023).…”

Section: Related Work-a Brief Survey Of Modern Srmentioning

confidence: 99%

Deep Symbolic Regression for Physics Guided by Units Constraints: Toward the Automated Discovery of Physical Laws

Tenachi,

Ibata,

Diakogiannis

2023

ApJ

View full text Add to dashboard Cite

Symbolic regression (SR) is the study of algorithms that automate the search for analytic expressions that fit data. While recent advances in deep learning have generated renewed interest in such approaches, the development of SR methods has not been focused on physics, where we have important additional constraints due to the units associated with our data. Here we present Φ-SO, a physical symbolic optimization framework for recovering analytical symbolic expressions from physics data using deep reinforcement learning techniques by learning units constraints. Our system is built, from the ground up, to propose solutions where the physical units are consistent by construction. This is useful not only in eliminating physically impossible solutions but also because the grammatical rules of dimensional analysis enormously restrict the freedom of the equation generator, thus vastly improving performance. The algorithm can be used to fit noiseless data, which can be useful, for instance, when attempting to derive an analytical property of a physical model, and it can also be used to obtain analytical approximations of noisy data. We test our machinery on a standard benchmark of equations from the Feynman Lectures on Physics and other physics textbooks, achieving state-of-the-art performance in the presence of noise (exceeding 0.1%) and show that it is robust even in the presence of substantial (10%) noise. We showcase its abilities on a panel of examples from astrophysics.

show abstract

“…Modern SR (Schmidt & Lipson 2009, 2011Kommenda et al 2020;Kammerer et al 2020;Bartlett et al 2023b;Brence et al 2021;Jin et al 2019;Luo et al 2022;Tohme et al 2023;Udrescu & Tegmark 2020;Kamienny et al 2022;Biggio et al 2020Biggio et al , 2021Vastl et al 2024;Kamienny et al 2023;Martius & Lampert 2017;Brunton et al 2016;Zheng et al 2022;Sahoo et al 2018;Petersen et al 2021a;Landajuela et al 2022;Holt et al 2023;Scholl et al 2023;Sousa et al 2024;Fiorini et al 2024;Shojaee et al 2024;Zhang & Lei 2024;Cheng & Alkhalifah 2024;He et al 2024;Makke & Chawla 2022;Angelis et al 2023;Faris et al 2024;Tian et al 2024;Michishita 2024;Melching et al 2024;Meidani et al 2024;Li et al 2024aLi et al , 2024bChen et al 2024) aims to use the immense computational resources at our disposal to search through possible analytic descriptions in terms of a set of functions and operators (e.g., x, +, −, ×, /, sin, cos, exp log, ...) to best fit some numerical data set (x, y) we wish to model. Concretely, one seeks s...…”

Section: Introductionmentioning

confidence: 99%

Class Symbolic Regression: Gotta Fit ’Em All

Tenachi,

Ibata,

François

et al. 2024

ApJL

View full text Add to dashboard Cite

We introduce “Class Symbolic Regression” (Class SR), the first framework for automatically finding a single analytical functional form that accurately fits multiple data sets—each realization being governed by its own (possibly) unique set of fitting parameters. This hierarchical framework leverages the common constraint that all the members of a single class of physical phenomena follow a common governing law. Our approach extends the capabilities of our earlier Physical Symbolic Optimization (Φ-SO) framework for symbolic regression, which integrates dimensional analysis constraints and deep reinforcement learning for unsupervised symbolic analytical function discovery from data. Additionally, we introduce the first Class SR benchmark, comprising a series of synthetic physical challenges specifically designed to evaluate such algorithms. We demonstrate the efficacy of our novel approach by applying it to these benchmark challenges and showcase its practical utility for astrophysics by successfully extracting an analytic galaxy potential from a set of simulated orbits approximating stellar streams.

show abstract

Reliable neural networks for regression uncertainty estimation

Cited by 9 publications

References 16 publications

USN: A Robust Imitation Learning Method against Diverse Action Noise

USN: A Robust Imitation Learning Method against Diverse Action Noise

Deep Symbolic Regression for Physics Guided by Units Constraints: Toward the Automated Discovery of Physical Laws

Class Symbolic Regression: Gotta Fit ’Em All

Contact Info

Product

Resources

About