SPSameerchand Pudaruth

Sameerchand Pudaruth

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Sameerchand Pudaruth has a total of 38 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Annotator Disagreement as a Calibrated Language-Identity Signal for Code-Switched Speech Recognition in Kreol Morisien

    by Bibi Faatimah Shagufta Yarroo, Baby Gobin-Rahimbux, Sameerchand Pudaruth

    In code-switched speech between lexically related languages, the language of a word can be graded rather than categorical. We propose that disagreement between human annotators marks these positions and can serve as a calibration target. On a trilingual Kreol Morisien, French, and English broadcast corpus, a classifier trained to reproduce two annotators’ label distributions is well calibrated (expected calibration error 0.008) and, under cross-validation, is more uncertain at the 100 words they disputed than at agreed words (entropy ratio 1.68). We use this signal to gate language-specific low-rank adapters in the decoder of a fine-tuned Whisper large-v3 model, with controls that separate the effects of decoding, adaptation, and the signal itself. Standard decoding inserts an automatically detected language token the fine-tuned model was never trained with; decoding without it accounts for essentially the whole of the overall improvement over the baseline. Against a control decoded identically, language conditioning redistributes errors between the languages rather than reducing them: French word error rate (WER) falls by 1.5 to 2.0 points across three training seeds, Kreol Morisien WER rises by about 0.3 points, overall WER changes by less than 0.09 points, and switch-point error does not improve. Ablations locate the French effect in the decode-time signal: replacing it with a uniform, language-swapped, or shuffled signal returns the system to the level of the uniform control, a model trained without language information, and most of the French gain persists when the classifier’s input transcription is degraded to 56% WER. The annotators’ own labels do no better than the predicted signal, and a one-hot form performs comparably, so the recognition effect does not require the signal’s graded form. An error analysis traces the direction of the redistribution to the baseline’s dominant error class: French words written in Kreol Morisien spelling, which are five times more frequent than the reverse. An equivalence lexicon judged by two native speakers attributes 0.6 to 1.0 points of the baseline’s 11.58% WER to spelling convention rather than recognition. These results rest on one dependence that the design cannot remove: the evaluation annotations also train the classifier whose pseudo-labels train the adapters. The uniform control bounds what this dependence contributes. To our knowledge, this is the first use of annotator disagreement as a calibration target for language identification, and the first language-identity conditioning of a speech recogniser for a creole and its lexifier.

  • Open Access

    Curvature-Corrected Rotary Position Embeddings: An Entropy-Invariant Temperature for Mitigating Numerical-Rank Collapse in Long-Context Attention

    by Joseph Tafataona Mtetwa, Kingsley A. Ogudo, Sameerchand Pudaruth

    As large language models process ever longer contexts, the positional encoding—most commonly rotary position embeddings (RoPEs)—must remain numerically trustworthy. We give an information-geometric analysis of why the attention matrix becomes ill-conditioned as the sequence length L grows. We first show that the tempting frequency-aliasing explanation fails, because RoPE’s multi-frequency code keeps positions well separated. The mechanism is statistical: at fixed softmax temperature the attention entropy grows like logL, and the rows spread their mass over a collision support (inverse participation ratio) that grows polynomially as La (a≈0.9 measured), confining the matrix’s trailing singular values and producing a numerical-rank collapse. We prove, and confirm in 50-digit arithmetic, that the spectral condition number is intrinsically ill-posed here, so the numerical rank is the correct diagnostic. From the exact entropy–temperature identity dH/dβ=−βVarp(e) we derive the entropy-invariant schedule β★(L)=(L/Lref)c0, Curvature-Corrected RoPE (CC-RoPE), and prove that in the diffuse near-circulant regime it eliminates the collapse at a characterised conditioning–mixing cost, whereas no relative-position-preserving warp can. The two signatures are further confirmed in a trained RoPE model (Pythia-160M); the downstream perplexity effect is framed as a falsifiable prediction.

  • Open Access

    Adaptive Gradient Penalty for Wasserstein GANs: Theory and Applications

    by Joseph Tafataona Mtetwa, Kingsley A. Ogudo, Sameerchand Pudaruth

    Wasserstein Generative Adversarial Networks (WGANs) have gained significant attention due to their theoretical foundations and effectiveness in generative modeling. However, training stability remains a major challenge, typically addressed through fixed gradient penalty (GP) techniques. In this paper, we propose an Adaptive Gradient Penalty (AGP) framework that employs a Proportional–Integral (PI) controller to adjust the gradient penalty coefficient λt based on real-time training feedback. We provide a comprehensive theoretical analysis, including convergence guarantees, stability conditions, and optimal parameter selection. Experimental validation on MNIST and CIFAR-10 datasets demonstrates that AGP achieves an 11.4% improvement in FID scores on CIFAR-10 while maintaining comparable performance on MNIST. The adaptive mechanism automatically evolves penalty coefficients from 10.0 to 21.29 for CIFAR-10, appropriately responding to dataset complexity, and achieves superior gradient norm control with only 7.9% deviation from the target value compared to 18.3% for standard WGAN-GP. This work represents the first comprehensive investigation of adaptive gradient penalty mechanisms for WGANs, providing both theoretical foundations and empirical evidence for their advantages in achieving robust and efficient adversarial training.


  • Open Access

    Bridging Modalities: An Analysis of Cross-Modal Wasserstein Adversarial Translation Networks and Their Theoretical Foundations

    by Joseph Tafataona Mtetwa, Kingsley A. Ogudo, Sameerchand Pudaruth

    What if machines could seamlessly translate between the visual richness of images and the semantic depth of language with mathematical precision? This paper presents a theoretical and empirical analysis of five novel cross-modal Wasserstein adversarial translation networks that challenge conventional approaches to cross-modal understanding. Unlike traditional generative models that rely on stochastic noise, our frameworks learn deterministic translation mappings that preserve semantic fidelity across modalities through rigorous mathematical foundations. We systematically examine: (1) cross-modality consistent dual-critical networks; (2) Wasserstein cycle consistency; (3) multi-scale Wasserstein distance; (4) regularization through modality invariance; and (5) Wasserstein information bottleneck. Each approach employs adversarial training with Wasserstein distances to establish theoretically grounded translation functions between heterogeneous data representations. Through mathematical analysis—including information-theoretic frameworks, differential geometry, and convergence guarantees—we establish the theoretical foundations underlying cross-modal translation. Our empirical evaluation across MS-COCO, Flickr30K, and Conceptual Captions datasets, including comparisons with transformer-based baselines, reveals that our proposed multi-scale Wasserstein cycle consistent (MS-WCC) framework achieves remarkable performance gains—12.1% average improvement in FID scores and 8.0% enhancement in cross-modal translation accuracy—compared to state-of-the-art methods, while maintaining superior computational efficiency. These results demonstrate that principled mathematical approaches to cross-modal translation can significantly advance machine understanding of multimodal data, opening new possibilities for applications requiring seamless communication between visual and textual domains.


  • Open Access

    Using Deep Learning to Identify Deepfakes Created Using Generative Adversarial Networks

    by Jhanvi Jheelan, Sameerchand Pudaruth

    Generative adversarial networks (GANs) have revolutionised various fields by creating highly realistic images, videos, and audio, thus enhancing applications such as video game development and data augmentation. However, this technology has also given rise to deepfakes, which pose serious challenges due to their potential to create deceptive content. Thousands of media reports have informed us of such occurrences, highlighting the urgent need for reliable detection methods. This study addresses the issue by developing a deep learning (DL) model capable of distinguishing between real and fake face images generated by StyleGAN. Using a subset of the 140K real and fake face dataset, we explored five different models: a custom CNN, ResNet50, DenseNet121, MobileNet, and InceptionV3. We leveraged the pre-trained models to utilise their robust feature extraction and computational efficiency, which are essential for distinguishing between real and fake features. Through extensive experimentation with various dataset sizes, preprocessing techniques, and split ratios, we identified the optimal ones. The 20k_gan_8_1_1 dataset produced the best results, with MobileNet achieving a test accuracy of 98.5%, followed by InceptionV3 at 98.0%, DenseNet121 at 97.3%, ResNet50 at 96.1%, and the custom CNN at 86.2%. All of these models were trained on only 16,000 images and validated and tested on 2000 images each. The custom CNN model was built with a simpler architecture of two convolutional layers and, hence, lagged in accuracy due to its limited feature extraction capabilities compared with deeper networks. This research work also included the development of a user-friendly web interface that allows deepfake detection by uploading images. The web interface backend was developed using Flask, enabling real-time deepfake detection, allowing users to upload images for analysis and demonstrating a practical use for platforms in need of quick, user-friendly verification. This application demonstrates significant potential for practical applications, such as on social media platforms, where the model can help prevent the spread of fake content by flagging suspicious images for review. This study makes important contributions by comparing different deep learning models, including a custom CNN, to understand the balance between model complexity and accuracy in deepfake detection. It also identifies the best dataset setup that improves detection while keeping computational costs low. Additionally, it introduces a user-friendly web tool that allows real-time deepfake detection, making the research useful for social media moderation, security, and content verification. Nevertheless, identifying specific features of GAN-generated deepfakes remains challenging due to their high realism. Future works will aim to expand the dataset by using all 140,000 images, refine the custom CNN model to increase its accuracy, and incorporate more advanced techniques, such as Vision Transformers and diffusion models. The outcomes of this study contribute to the ongoing efforts to counteract the negative impacts of GAN-generated images.


  • Open Access

    Automatic Translation Between Kreol Morisien and English Using the Marian Machine Translation Framework

    by Zaheenah Beebee Jameela Boodeea, Sameerchand Pudaruth, Nitish Chooramun, Aneerav Sukhoo

    Kreol Morisien is a vibrant and expressive language that reflects the multicultural heritage of Mauritius. There are different versions of Kreol languages. While Kreol Morisien is spoken in Mauritius, Kreol Rodrige is spoken only in Rodrigues, and they are distinct languages. Being spoken by only about 1.5 million speakers in the world, Kreol Morisien falls in the category of under-resourced languages. Initially, Kreol Morisien lacked a formalised writing system, with many people using different spellings for the same words. The first step towards standardisation of writing Kreol Morisien was after the publication of the Kreol Morisien orthography in 2011 and Kreol Morisien grammar in 2012 by the Kreol Morisien Academy. Kreol Morisien obtained a national position in the year 2012 when it was introduced in educational organisations. This was a major breakthrough for Kreol Morisien to be recognised as a national language on the same level as English, French, and other oriental languages. By providing a means for Kreol Morisien speakers to connect with others, a translation system will help to preserve and strengthen the identity of the language and its speakers in an increasingly globalized world. The aim of this paper is to develop a translation system for Kreol Morisien and English. Thus, a dataset consisting of 50,000 parallel Kreol Morisien and English sentences was created, where 48,000 sentence pairs were used to train the models, while 1000 sentences were used for evaluation and another 1000 sentences were used for testing. Several machine translation systems such as statistical machine translation, open-source neural machine translation, a Transformer model with attention mechanism, and Marian machine translation are trained and evaluated. Our best model, using MarianMT, achieved a BLEU score of 0.62 for the translation of English to Kreol Morisien and a BLEU score of 0.58 for the translation of Kreol Morisien into English. To our knowledge, these are the highest BLEU scores that are available in the literature for this language pair. A high-quality translation tool for Kreol Morisien will facilitate its integration into digital platforms. This will make previously inaccessible knowledge more accessible, as the information can now be translated into the mother tongue of most Mauritians with reasonable accuracy.


  • Open Access

    A neural machine translation system for Kreol Repiblik Moris and English

    by Sameerchand Pudaruth, Sheeba Armoogum, Nirmal Kumar Betchoo, Aneerav Sukhoo, Vandanah Gooria, Abdallah Peerally, Mohammad Zafar Khodabocus

    Although Google Translate is a widely used machine translation service that supports 133 languages, it does not incorporate support for the Kreol Repiblik Moris (KRM) language. Addressing this limitation, the current research focuses on enhancing the accuracy and fluency of machine translation between KRM and English through natural language processing and deep neural machine translation techniques. In this study, a machine translation system using a transformer model trained with a dataset of 50,000 parallel corpora has been developed. The model was evaluated using manual translations and the bilingual evaluation understudy (BLEU) score. A score of 31.46 for translating from KRM to English and 28.15 for translating from English to KRM was achieved. To our knowledge, these are the highest BLEU scores for translation between these two languages. This is due to utilising the largest dataset and extensive atomic words from the KRM dictionary. This successful interdisciplinary funded project led to the setting up of a free online translation service and a smartphone app for Mauritian citizens and tourists.


  • Open Access

    Enhancing Stock Price Forecasting with Generative Adversarial Networks and Conformal Prediction: A Novel Approach for Quantifying Uncertainty

    by Joseph Tafataona Mtetwa, Kingsley Ogudo, Sameerchand Pudaruth

    This study introduces a novel method that combines Generative Adversarial Networks (GANs) with conformal prediction to revolutionize the domain of stock ticker forecasting. Our approach enhances prediction accuracy and offers dependable prediction intervals to gauge uncertainty, surpassing current GAN-based techniques in unpredictable market scenarios. We made a validation using AAPL stock data, showing the ability of our model to generate accurate forecasts and reliable assessments of prediction uncertainty. The prediction intervals generated provide investors and risk managers with a powerful tool to make informed decisions. Our approach represents a notable advancement in financial prediction as it improves the clarity and reliability of forecasts. By incorporating conformal prediction into the GAN framework, the reliability of forecasts is improved, providing a thorough evaluation of uncertainty that is crucial for successful risk management. This framework has the potential to transform the field of predictive analytics in finance. It shows a solution that balances accuracy and uncertainty quantification. The results clearly show the effectiveness of this method, paving the way for its potential application in a wide range of financial prediction scenarios.

  • Open Access

    Front Cover

    by Sameerchand Pudaruth, Upasana Singh