Does Adversarial Oversampling Help us?

Dam, Tanmoy; Ferdaus, Meftahul; Anavatti, Sreenatha G.; Senthilnath, J.; Abbass, Hussein A.

doi:10.1145/3459637.3482091

Cited by 5 publications

(5 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These gains over state-of-the-art techniques are attributed to X-Fuzz's adaptive architecture and lower dependence on user-defined parameters. The empirical results thus highlight X-Fuzz's capabilities in classifying challenging non-stationary and imbalanced data streams [50], [51], demonstrating its viability for real-world applications.…”

Section: Results and Discussion Under Prequential Analysismentioning

confidence: 77%

X-Fuzz: An Evolving and Interpretable Neuro-Fuzzy Learner for Data Streams

Ferdaus,

Dam,

Alam

et al. 2024

IEEE Trans. Artif. Intell.

View full text Add to dashboard Cite

While evolving neuro-fuzzy systems have shown promise for learning from non-stationary streaming data with concept drift, most existing models lack transparency due to the limited interpretability of Takagi-Sugeno fuzzy architecture's linear rule consequents. The lack of transparency limits the reliability of crucial applications. To address this limitation, this paper proposes a new evolving neuro-fuzzy system called X-Fuzz that enhances interpretability by integrating the LIME technique to provide local explanations and evaluates them using faithfulness and monotonicity metrics. X-Fuzz is rigorously tested on streaming datasets with diverse concept drifts via prequential analysis. Experiments demonstrate X-Fuzz's capabilities in mining insights from large and dynamic data streams exhibiting diverse concept drifts including abrupt, gradual, recurring contextual, and cyclical drifts. In addition, for online runway exit prediction using real aviation data, X-Fuzz achieved 98.04% accuracy, significantly exceeding recent methods. With its balance of efficiency and transparency, X-Fuzz represents a promising approach for trustworthy evolving artificial intelligence that can handle complex, non-stationary data streams in critical realworld settings. We have made the X-Fuzz source code available in https://github.com/m-ferdaus/X Fuzz for reproducibility and facilitating future research.Impact Statement-The transparency of current evolving neuro-fuzzy systems restricts trustworthiness, limiting adoption in critical real-world applications. X-Fuzz tackles this barrier by transparently learning from non-stationary data via integrated local explanations and rigorous explanation evaluation. In online runway exit prediction, X-Fuzz achieves 98.04% accuracy, exceeding recent models by 6.7%. With its balanced efficiency and transparency, X-Fuzz significantly advances reliable and interpretable evolving AI. By making model behaviors traceable, X-Fuzz expands the safe deployment of adaptive intelligent systems in mission-critical settings across domains. Overall, X-Fuzz enables trustworthy neuro-fuzzy learners that incrementally adapt to dynamic data, overcoming reliability barriers and driving real-world impact.

show abstract

Section: Results and Discussion Under Prequential Analysismentioning

confidence: 77%

X-Fuzz: An Evolving and Interpretable Neuro-Fuzzy Learner for Data Streams

Ferdaus,

Dam,

Alam

et al. 2024

IEEE Trans. Artif. Intell.

View full text Add to dashboard Cite

show abstract

“…• For handling class imbalance problem, a latent preserving GAN [21,22] can be used to generate minority class samples in dynamic tests environments.…”

Section: Discussionmentioning

confidence: 99%

Improving Self-Supervised Learning for Out-Of-Distribution Task via Auxiliary Classifier

Boonlia

Dam

Ferdaus

et al. 2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

In real world scenarios, out-of-distribution (OOD) datasets may have a large distributional shift from training datasets. This phenomena generally occurs when a trained classifier is deployed on varying dynamic environments, which causes a significant drop in performance. To tackle this issue, we are proposing an end-to-end deep multi-task network in this work. Observing a strong relationship between rotation prediction (self-supervised) accuracy and semantic classification accuracy on OOD tasks, we introduce an additional auxiliary classification head in our multi-task network along with semantic classification and rotation prediction head. To observe the influence of this addition classifier in improving the rotation prediction head, our proposed learning method is framed into bi-level optimisation problem where the upper-level is trained to update the parameters for semantic classification and rotation prediction head. In the lower-level optimisation, only the auxiliary classification head is updated through semantic classification head by fixing the parameters of the semantic classification head. The proposed method has been validated through three unseen OOD datasets where it exhibits a clear improvement in semantic classification accuracy than other two baseline methods. Our code is available on GitHub https://github.com/harshita-555/OSSL

show abstract

“…Without loss of generalisation, the class distribution follows p maj ≥ p 2 ≥ ...p l ≥ ... ≥ p min , N = C l=c p l , in which p maj and p min denote the majority class and the minority class, respectively. The relationship between the majority class and minority class(es) is set as p maj ≥ 50 * p min following the problem setting of [4,6,7]. The main objective is to design a deep neural network estimating the underlying data distribution of C classes, thereby producing robust decision boundaries.…”

Section: Problem Formulationsmentioning

confidence: 99%

“…The proposed SLPPL is able to preserve stable network parameters (ϕ, θ), and to maintain stable class distributions in the low-dimensional manifold where (3) is solved using the ADAM optimiser [7]. The i-th sample is easily encoded in the latent space afterward as z i ∈ Z = Enc θ (x i ) and preserves a multivariate normal distribution (MND) [11] for the i-th class.…”

Section: Slpplmentioning

confidence: 99%

See 1 more Smart Citation

Latent Preserving Generative Adversarial Network for Imbalance Classification

Dam

Ferdaus

Pratama

et al. 2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Many real-world classification problems have imbalanced frequency of class labels; a well-known issue known as the "class imbalance" problem. Classic classification algorithms tend to be biased towards the majority class, leaving the classifier vulnerable to misclassification of the minority class. While the literature is rich with methods to fix this problem, as the dimensionality of the problem increases, many of these methods do not scale-up and the cost of running them become prohibitive. In this paper, we present an end-to-end deep generative classifier. We propose a domain-constraint autoencoder to preserve the latent-space as prior for a generator, which is then used to play an adversarial game with two other deep networks, a discriminator and a classifier. Extensive experiments are carried out on three different multi-class imbalanced problems and a comparison with state-of-the-art methods. Experimental results confirmed the superiority of our method over popular algorithms in handling high-dimensional imbalanced classification problems. Our code is available on https://github.com/TanmDL/SLPPL-GAN

show abstract

Does Adversarial Oversampling Help us?

Cited by 5 publications

References 9 publications

X-Fuzz: An Evolving and Interpretable Neuro-Fuzzy Learner for Data Streams

X-Fuzz: An Evolving and Interpretable Neuro-Fuzzy Learner for Data Streams

Improving Self-Supervised Learning for Out-Of-Distribution Task via Auxiliary Classifier

Latent Preserving Generative Adversarial Network for Imbalance Classification

Contact Info

Product

Resources

About