Ensuring fairness in machine learning models is a critical challenge. Existing debiasing methods often compromise performance, rely on static correction strategies, and struggle with data sparsity, particularly within minority groups. Furthermore, their utilization of sensitive attributes is often suboptimal, either depending excessively on complete attribute labeling or disregarding these attributes entirely. To overcome these limitations, we propose FairNet, a novel framework for dynamic, instance-level fairness correction. FairNet integrates a bias detector with conditional low-rank adaptation (LoRA), which enables selective activation of the fairness correction mechanism exclusively for instances identified as biased, and thereby preserve performance on unbiased instances. A key contribution is a new contrastive loss function for training the LoRA module, specifically designed to minimize intra-class representation disparities across different sensitive groups and effectively address underfitting in minority groups. The FairNet framework can flexibly handle scenarios with complete, partial, or entirely absent sensitive attribute labels. Theoretical analysis confirms that, under moderate TPR/FPR for the bias detector, FairNet can enhance the performance of the worst group without diminishing overall model performance, and potentially yield slight performance improvements. Comprehensive empirical evaluations across diverse vision and language benchmarks validate the effectiveness of FairNet. Code is available at \url{https://github.com/SongqiZhou/FairNet}.

Employing machine learning (ML) models to accelerate experimentation and uncover biological mechanisms has been a rising tendency in genetic engineering. However, effectively collecting data to enhance model accuracy and improve design remains challenging, especially when data quality is poor and validation resources are limited. Active learning (AL) addresses this by iteratively identifying promising candidates, thereby reducing experimental efforts while improving model performance. This review highlights how AL can assist scientists throughout the design-build-test-learn cycle, explore its various practical implementations, and discuss its potential through the integration of cross-domain expertise. In the age of genetic engineering revolutionized by data-driven ML models, AL presents an iterative framework that significantly enhances the functionalities of biomolecules and uncovers their intrinsic mechanisms, all while minimizing expenses and efforts.

Fast charging has attracted increasing attention from the battery community for electrical vehicles (EVs) to alleviate range anxiety and reduce charging time for EVs. However, inappropriate charging strategies would cause severe degradation of batteries or even hazardous accidents. To optimize fast-charging strategies under various constraints, particularly safety limits, we propose a novel deep Bayesian optimization (BO) approach that utilizes Bayesian recurrent neural network (BRNN) as the surrogate model, given its capability in handling sequential data and providing uncertainty quantifications for the output. In addition, a combined acquisition function of expected improvement (EI) and upper confidence bound (UCB) is developed to better balance the exploitation and exploration. The effectiveness of the proposed approach is demonstrated on the PETLION, a porous electrode theory-based battery simulator. Our method is also compared with the state-of-the-art BO methods that use Gaussian process (GP) and non-recurrent network as surrogate models. The results verify the superior performance of the proposed fast charging approaches, which mainly results from that: 1) the BRNN-based surrogate model provides a more precise prediction of battery lifetime than that based on GP or non-recurrent network; and 2) the combined acquisition function outperforms traditional EI or UCB criteria in exploring the optimal charging protocol that maintains the longest battery lifetime. Note to Practitioners—This study is motivated by the need to develop fast-charging strategies for batteries to reduce the charging time while maintaining the safety and longevity of batteries. Traditional methods to optimize battery fast-charging protocols require either solving complex battery models or conducting tremendous repetitive and costly cycling experiments. In addition, the majority of works in this direction mainly focus on minimizing the charging time without considering the battery degradation caused by the designed charging protocol. To address these issues, we present a data-driven optimization technique to rapidly discover the optimal fast-charging strategies that not only decrease the charging time but also slow down the battery degradation. Our method is efficient and can find the optimal solution within tens of iterations (or repetitive experiments), thus greatly reducing the cost and efforts in optimizing battery charging protocols.

Optimizing the fast-charging of lithium-ion batteries to minimize charging time while limiting battery degradation is a crucial challenge for the battery community. In this article, we propose an attention-enhanced deep reinforcement learning (AtDRL) approach for fast charging optimization under conditions of partial battery states observable. By incorporating a self-attention mechanism within the actor-critic network architecture, the AtDRL model can dynamically shift its focus across different parts of the input sequence during prediction. This is particularly advantageous for processing input sequences of varying lengths and utilizing incomplete observations to improve sample efficiency, thereby improving the performance of fast-charging optimization. In addition, a crafted reward function is further put forward to meticulously control battery current, temperature, and voltage to mitigate battery degradation. The effectiveness of the proposed AtDRL approach is demonstrated on a porous electrode theory-based battery simulator. The results show that, compared with the conventional DRL-based charging method and Bayesian optimization-based charging method, the proposed AtDRL approach possesses enhanced charging performance and higher sample efficiency under the scenarios of incomplete state observations.

The fast charging problem of lithium-ion batteries with minimum charging time while limiting battery degradation is receiving increasing attention and is a critical challenge to battery community. Difficulties in this optimization lie in that: 1) the parameter space of charging strategies is high dimensional, while the budget of the experimental cost is often limited; 2) the evaluation of charging strategies' performance is expensive; and 3) the degradation process of the battery is strongly nonlinear, and multiple degradation mechanisms occur simultaneously leading to difficulties for establishing accurate first-principle models. Current methods to address these difficulties are mainly electrochemical-model-based optimization and grid search, which are rarely adaptive to battery degradation and/or are of low sample efficiency. In this article, we propose an adaptive model-based reinforcement learning (RL) approach for fast-charging optimization while limiting battery degradation, in which a probabilistic surrogate model of differential Gaussian process (GP) is adopted to adaptively describe the degradation of cells. The effectiveness of the proposed approach is demonstrated on PETLION, a high-performance porous-electrode-theory-based battery simulator. The results show that 1) compared with the model-free RL method, the proposed adaptive GP-based RL approach possesses superior charging performance and high sample efficiency and 2) the proposed method performs well in the handling of degradation constraints on voltage and temperature for dynamically aging batteries with its adaptability to the variations of environment.

Data-driven fault diagnosis has attracted increasing research interest with a recent trend of aiming at large-scale and complex systems. In this article, we propose a method under a probabilistic framework, named dynamic Bhattacharyya bound (DBB), to extract features for fault diagnosis. An information criterion is adopted to determine the order of dimensionality reduction and time lags when applying the proposed approach. Compared with conventional diagnostic approaches, the proposed DBB approach has several advantageous features. First, the DBB approach minimizes an upper bound of the Bayes error which is a direct manifestation of the misclassification rate. Second, pairwise Bhattacharyya bounds between different faults are summed up in the objective function, enabling it to address the fault diagnosis of multiple faults that may have large overlaps. The proposed method is validated through the Tennessee Eastman process and it shows advantageous performance than other methods such as Fisher discriminant analysis (FDA), dynamic FDA, and LP-DFDA.

This paper proposes a stochastic representation of maximized mutual information analysis (MIA) method for quality monitoring in which a manner of imposing prior probability distributions over projection parameters is employed and subsequently, a Bayesian estimation algorithm is put forward for projection learning. The proposed stochastic MIA (SMIA) based approach allows the enhanced performance of fault detection due to the following advantages over classic monitoring methods. First, the SMIA approach utilizes the mechanism of hierarchical priors and an individual prior over each projection direction, as a key feature of the proposed method, which enables SMIA to build a sparse model that can discard irrelevant components in the process data with respect to the prediction of quality variables. Second, the proposed SMIA method incorporates the advantage of maximizing mutual information on the minimum achievable error of model prediction as well as the advantage of describing the serial dynamics. Additionally, the optimal dimensionality of the latent space in an SMIA can be automatically determined during the procedure of Bayesian estimation by the utilization of these adaptive priors over projections. The effectiveness of the proposed approach for quality monitoring is demonstrated on the benchmark of Tennessee Eastman process.

In multivariate systems, the causality relationships between any two different data variables and the corresponding path models are often unknown. In this paper, the identification of bidirectional path models of a bivariate system is investigated by extending the augmented UD identification (AUDI) algorithm proposed by Niu(1992) which can simultaneously identify the order and parameters for open-loop systems with unclear physical meanings of the even columns in the data matrix. To extract more information than the AUDI algorithm for identification of bidirectional path models, we develop a novel approach based on construction of the interleave data vector and UD factorization of the data matrix. The odd and even columns of the resulting data matrix correspond to the parameters of the forward and backward path models, respectively. Moreover, the information contained in the data matrix can be evaluated to determine the causality between the two data variables. The ARMAX process with white noise is first considered. The results are then extended to the case with colored noise. Simulation results are presented to show the effectiveness of our proposed methods.