Soumya K. Ghosh

Prasanna Sattigeri, Soumya Ghosh, Inkit Padhi et al. · ibm-research

In consequential decision-making applications, mitigating unwanted biases in machine learning models that yield systematic disadvantage to members of groups delineated by sensitive attributes such as race and gender is one key intervention to strive for equity. Focusing on demographic parity and equality of opportunity, in this paper we propose an algorithm that improves the fairness of a pre-trained classifier by simply dropping carefully selected training data points. We select instances based on their influence on the fairness metric of interest, computed using an infinitesimal jackknife-based approach. The dropping of training points is done in principle, but in practice does not require the model to be refit. Crucially, we find that such an intervention does not substantially reduce the predictive performance of the model but drastically improves the fairness metric. Through careful experiments, we evaluate the effectiveness of the proposed approach on diverse tasks and find that it consistently improves upon existing alternatives.

Manish Nagireddy, Inkit Padhi, Soumya Ghosh et al.

Large language models (LLMs) have convincing performance in a variety of downstream tasks. However, these systems are prone to generating undesirable outputs such as harmful and biased text. In order to remedy such generations, the development of guardrail (or detector) models has gained traction. Motivated by findings from developing a detector for social bias, we adopt the notion of a use-mention distinction - which we identified as the primary source of under-performance in the preliminary versions of our social bias detector. Armed with this information, we describe a fully extensible and reproducible synthetic data generation pipeline which leverages taxonomy-driven instructions to create targeted and labeled data. Using this pipeline, we generate over 300K unique contrastive samples and provide extensive experiments to systematically evaluate performance on a suite of open source datasets. We show that our method achieves competitive performance with a fraction of the cost in compute and offers insight into iteratively developing efficient and capable guardrail models. Warning: This paper contains examples of text which are toxic, biased, and potentially harmful.

Maohao Shen, Yuheng Bu, Prasanna Sattigeri et al.

It is known that neural networks have the problem of being over-confident when directly using the output label distribution to generate uncertainty measures. Existing methods mainly resolve this issue by retraining the entire model to impose the uncertainty quantification capability so that the learned model can achieve desired performance in accuracy and uncertainty prediction simultaneously. However, training the model from scratch is computationally expensive and may not be feasible in many situations. In this work, we consider a more practical post-hoc uncertainty learning setting, where a well-trained base model is given, and we focus on the uncertainty quantification task at the second stage of training. We propose a novel Bayesian meta-model to augment pre-trained models with better uncertainty quantification abilities, which is effective and computationally efficient. Our proposed method requires no additional training data and is flexible enough to quantify different uncertainties and easily adapt to different application settings, including out-of-domain data detection, misclassification detection, and trustworthy transfer learning. We demonstrate our proposed meta-model approach's flexibility and superior empirical performance on these applications over multiple representative image classification benchmarks.

Maohao Shen, Soumya Ghosh, Prasanna Sattigeri et al.

Due to privacy or commercial constraints, large pre-trained language models (PLMs) are often offered as black-box APIs. Fine-tuning such models to downstream tasks is challenging because one can neither access the model's internal representations nor propagate gradients through it. This paper addresses these challenges by developing techniques for adapting PLMs with only API access. Building on recent work on soft prompt tuning, we develop methods to tune the soft prompts without requiring gradient computation. Further, we develop extensions that in addition to not requiring gradients also do not need to access any internal representation of the PLM beyond the input embeddings. Moreover, instead of learning a single prompt, our methods learn a distribution over prompts allowing us to quantify predictive uncertainty. Ours is the first work to consider uncertainty in prompts when only having API access to the PLM. Finally, through extensive experiments, we carefully vet the proposed methods and find them competitive with (and sometimes even improving on) gradient-based approaches with full access to the PLM.

Bidyut Saha, Riya Samanta, Soumya K. Ghosh et al.

In this work, we present TinyTNAS, a novel hardware-aware multi-objective Neural Architecture Search (NAS) tool specifically designed for TinyML time series classification. Unlike traditional NAS methods that rely on GPU capabilities, TinyTNAS operates efficiently on CPUs, making it accessible for a broader range of applications. Users can define constraints on RAM, FLASH, and MAC operations to discover optimal neural network architectures within these parameters. Additionally, the tool allows for time-bound searches, ensuring the best possible model is found within a user-specified duration. By experimenting with benchmark dataset UCI HAR, PAMAP2, WISDM, MIT BIH, and PTB Diagnostic ECG Databas TinyTNAS demonstrates state-of-the-art accuracy with significant reductions in RAM, FLASH, MAC usage, and latency. For example, on the UCI HAR dataset, TinyTNAS achieves a 12x reduction in RAM usage, a 144x reduction in MAC operations, and a 78x reduction in FLASH memory while maintaining superior accuracy and reducing latency by 149x. Similarly, on the PAMAP2 and WISDM datasets, it achieves a 6x reduction in RAM usage, a 40x reduction in MAC operations, an 83x reduction in FLASH, and a 67x reduction in latency, all while maintaining superior accuracy. Notably, the search process completes within 10 minutes in a CPU environment. These results highlight TinyTNAS's capability to optimize neural network architectures effectively for resource-constrained TinyML applications, ensuring both efficiency and high performance. The code for TinyTNAS is available at the GitHub repository and can be accessed at https://github.com/BidyutSaha/TinyTNAS.git.

Sameer K. Deshpande, Soumya Ghosh, Tin D. Nguyen et al.

Test log-likelihood is commonly used to compare different models of the same data or different approximate inference algorithms for fitting the same probabilistic model. We present simple examples demonstrating how comparisons based on test log-likelihood can contradict comparisons according to other objectives. Specifically, our examples show that (i) approximate Bayesian inference algorithms that attain higher test log-likelihoods need not also yield more accurate posterior approximations and (ii) conclusions about forecast accuracy based on test log-likelihood comparisons may not agree with conclusions based on root mean squared error.

Bidyut Saha, Riya Samanta, Soumya K Ghosh et al.

Human activity recognition (HAR) holds immense potential for transforming health and fitness monitoring, yet challenges persist in achieving personalized outcomes and sustainability for on-device continuous inferences. This work introduces a wrist-worn smart band designed to address these challenges through a novel combination of on-device TinyML-driven computing and cloud-enabled auto-deployment. Leveraging inertial measurement unit (IMU) sensors and a customized 1D Convolutional Neural Network (CNN) for personalized HAR, users can tailor activity classes to their unique movement styles with minimal calibration. By utilising TinyML for local computations, the smart band reduces the necessity for constant data transmission and radio communication, which in turn lowers power consumption and reduces carbon footprint. This method also enhances the privacy and security of user data by limiting its transmission. Through transfer learning and fine-tuning on user-specific data, the system achieves a 37\% increase in accuracy over generalized models in personalized settings. Evaluation using three benchmark datasets, WISDM, PAMAP2, and the BandX demonstrates its effectiveness across various activity domains. Additionally, this work presents a cloud-supported framework for the automatic deployment of TinyML models to remote wearables, enabling seamless customization and on-device inference, even with limited target data. By combining personalized HAR with sustainable strategies for on-device continuous inferences, this system represents a promising step towards fostering healthier and more sustainable societies worldwide.

Riya Samanta, Bidyut Saha, Soumya K. Ghosh et al.

Conditional Tabular Generative Adversarial Networks (CTGAN) and their various derivatives are attractive for their ability to efficiently and flexibly create synthetic tabular data, showcasing strong performance and adaptability. However, there are certain critical limitations to such models. The first is their inability to preserve the semantic integrity of contextually correlated words or phrases. For instance, skillset in freelancer profiles is one such attribute where individual skills are semantically interconnected and indicative of specific domain interests or qualifications. The second challenge of traditional approaches is that, when applied to generate contextually correlated tabular content, besides generating semantically shallow content, they consume huge memory resources and CPU time during the training stage. To address these problems, we introduce a novel framework, CTGKrEW (Conditional Tabular GAN with KMeans Clustering and Word Embedding), which is adept at generating realistic synthetic tabular data where attributes are collections of semantically and contextually coherent words. CTGKrEW is trained and evaluated using a dataset from Upwork, a realworld freelancing platform. Comprehensive experiments were conducted to analyze the variability, contextual similarity, frequency distribution, and associativity of the generated data, along with testing the framework's system feasibility. CTGKrEW also takes around 99\% less CPU time and 33\% less memory footprints than the conventional approach. Furthermore, we developed KrEW, a web application to facilitate the generation of realistic data containing skill-related information. This application, available at https://riyasamanta.github.io/krew.html, is freely accessible to both the general public and the research community.

Maohao Shen, J. Jon Ryu, Soumya Ghosh et al.

This paper questions the effectiveness of a modern predictive uncertainty quantification approach, called \emph{evidential deep learning} (EDL), in which a single neural network model is trained to learn a meta distribution over the predictive distribution by minimizing a specific objective function. Despite their perceived strong empirical performance on downstream tasks, a line of recent studies by Bengs et al. identify limitations of the existing methods to conclude their learned epistemic uncertainties are unreliable, e.g., in that they are non-vanishing even with infinite data. Building on and sharpening such analysis, we 1) provide a sharper understanding of the asymptotic behavior of a wide class of EDL methods by unifying various objective functions; 2) reveal that the EDL methods can be better interpreted as an out-of-distribution detection algorithm based on energy-based-models; and 3) conduct extensive ablation studies to better assess their empirical effectiveness with real-world datasets. Through all these analyses, we conclude that even when EDL methods are empirically effective on downstream tasks, this occurs despite their poor uncertainty quantification capabilities. Our investigation suggests that incorporating model uncertainty can help EDL methods faithfully quantify uncertainties and further improve performance on representative downstream tasks, albeit at the cost of additional computational complexity.

Debojyoti Ghosh, Soumya K Ghosh, Adrijit Goswami

Efficient and accurate classification of waste and industrial surface defects is essential for ensuring sustainable waste management and maintaining high standards in quality control. This paper introduces the Neuroplastic Modular Classifier, a novel hybrid architecture designed for robust and adaptive image classification in dynamic environments. The model combines a ResNet-50 backbone for localized feature extraction with a Vision Transformer (ViT) to capture global semantic context. Additionally, FAISS-based similarity retrieval is incorporated to provide a memory-like reference to previously encountered data, enriching the model's feature space. A key innovation of our architecture is the neuroplastic modular design composed of expandable, learnable blocks that dynamically grow during training when performance plateaus. Inspired by biological learning systems, this mechanism allows the model to adapt to data complexity over time, improving generalization. Beyond garbage classification, we validate the model on the Kolektor Surface Defect Dataset 2 (KolektorSDD2), which involves industrial defect detection on metal surfaces. Experimental results across domains show that the proposed architecture outperforms traditional static models in both accuracy and adaptability. The Neuroplastic Modular Classifier offers a scalable, high-performance solution for real-world image classification, with strong applicability in both environmental and industrial domains.

Tejaswini Pedapati, Amit Dhurandhar, Soumya Ghosh et al.

Estimating uncertainty or confidence in the responses of a model can be significant in evaluating trust not only in the responses, but also in the model as a whole. In this paper, we explore the problem of estimating confidence for responses of large language models (LLMs) with simply black-box or query access to them. We propose a simple and extensible framework where, we engineer novel features and train a (interpretable) model (viz. logistic regression) on these features to estimate the confidence. We empirically demonstrate that our simple framework is effective in estimating confidence of Flan-ul2, Llama-13b, Mistral-7b and GPT-4 on four benchmark Q\&A tasks as well as of Pegasus-large and BART-large on two benchmark summarization tasks with it surpassing baselines by even over $10\%$ (on AUROC) in some cases. Additionally, our interpretable approach provides insight into features that are predictive of confidence, leading to the interesting and useful discovery that our confidence models built for one LLM generalize zero-shot across others on a given dataset.

Sukhpal Singh Gill, Minxian Xu, Panos Patros et al.

ChatGPT, an AI-based chatbot, was released to provide coherent and useful replies based on analysis of large volumes of data. In this article, leading scientists, researchers and engineers discuss the transformative effects of ChatGPT on modern education. This research seeks to improve our knowledge of ChatGPT capabilities and its use in the education sector, identifying potential concerns and challenges. Our preliminary evaluation concludes that ChatGPT performed differently in each subject area including finance, coding and maths. While ChatGPT has the ability to help educators by creating instructional content, offering suggestions and acting as an online educator to learners by answering questions and promoting group work, there are clear drawbacks in its use, such as the possibility of producing inaccurate or false data and circumventing duplicate content (plagiarism) detectors where originality is essential. The often reported hallucinations within Generative AI in general, and also relevant for ChatGPT, can render its use of limited benefit where accuracy is essential. What ChatGPT lacks is a stochastic measure to help provide sincere and sensitive communication with its users. Academic regulations and evaluation practices used in educational institutions need to be updated, should ChatGPT be used as a tool in education. To address the transformative effects of ChatGPT on the learning environment, educating teachers and students alike about its capabilities and limitations will be crucial.

William T. Stephenson, Soumya Ghosh, Tin D. Nguyen et al.

Gaussian processes (GPs) are used to make medical and scientific decisions, including in cardiac care and monitoring of atmospheric carbon dioxide levels. Notably, the choice of GP kernel is often somewhat arbitrary. In particular, uncountably many kernels typically align with qualitative prior knowledge (e.g.\ function smoothness or stationarity). But in practice, data analysts choose among a handful of convenient standard kernels (e.g.\ squared exponential). In the present work, we ask: Would decisions made with a GP differ under other, qualitatively interchangeable kernels? We show how to answer this question by solving a constrained optimization problem over a finite-dimensional space. We can then use standard optimizers to identify substantive changes in relevant decisions made with a GP. We demonstrate in both synthetic and real-world examples that decisions made with a GP can exhibit non-robustness to kernel choice, even when prior draws are qualitatively interchangeable to a user.

Anupam Khan, Soumya K. Ghosh

Approval of credit card application is one of the censorious business decision the bankers are usually taking regularly. The growing number of new card applications and the enormous outstanding amount of credit card bills during the recent pandemic make this even more challenging nowadays. Some of the previous studies suggest the usage of machine intelligence for automating the approval process to mitigate this challenge. However, the effectiveness of such automation may depend on the richness of the training dataset and model efficiency. We have recently developed a novel classifier named random wheel which provides a more interpretable output. In this work, we have used an enhanced version of random wheel to facilitate a trustworthy recommendation for credit card approval process. It not only produces more accurate and precise recommendation but also provides an interpretable confidence measure. Besides, it explains the machine recommendation for each credit card application as well. The availability of recommendation confidence and explanation could bring more trust in the machine provided intelligence which in turn can enhance the efficiency of the credit card approval process.

Shreya Ghosh, Anwesha Mukherjee, Soumya K Ghosh et al.

Several researches and evidence show the increasing likelihood of pandemics (large-scale outbreaks of infectious disease) which has far reaching sequels in all aspects of human lives ranging from rapid mortality rates to economic and social disruption across the world. In the recent time, COVID-19 (Coronavirus Disease 2019) pandemic disrupted normal human lives, and motivated by the urgent need of combating COVID-19, researchers have put significant efforts in modelling and analysing the disease spread patterns for effective preventive measures (in addition to developing pharmaceutical solutions, like vaccine). In this regards, it is absolutely necessary to develop an analytics framework by extracting and incorporating the knowledge of heterogeneous datasources to deliver insights in improving administrative policy and enhance the preparedness to combat the pandemic. Specifically, human mobility, travel history and other transport statistics have significant impacts on the spread of any infectious disease. In this direction, this paper proposes a spatio-temporal knowledge mining framework, named STOPPAGE to model the impact of human mobility and other contextual information over large geographic area in different temporal scales. The framework has two major modules: (i) Spatio-temporal data and computing infrastructure using fog/edge based architecture; and (ii) Spatio-temporal data analytics module to efficiently extract knowledge from heterogeneous data sources. Typically, we develop a Pandemic-knowledge graph to discover correlations among mobility information and disease spread, a deep learning architecture to predict the next hot-spot zones; and provide necessary support in home-health monitoring utilizing Femtolet and fog/edge based solutions. The experimental evaluations on real-life datasets related to COVID-19 in India illustrate the efficacy of the proposed methods.

Suman Samui, Indrajit Chakrabarti, Soumya K. Ghosh

In recent years, Long Short-Term Memory (LSTM) has become a popular choice for speech separation and speech enhancement task. The capability of LSTM network can be enhanced by widening and adding more layers. However, this would introduce millions of parameters in the network and also increase the requirement of computational resources. These limitations hinders the efficient implementation of RNN models in low-end devices such as mobile phones and embedded systems with limited memory. To overcome these issues, we proposed to use an efficient alternative approach of reducing parameters by representing the weight matrix parameters of LSTM based on Tensor-Train (TT) format. We called this Tensor-Train factorized LSTM as TT-LSTM model. Based on this TT-LSTM units, we proposed a deep TensorNet model for single-channel speech enhancement task. Experimental results in various test conditions and in terms of standard speech quality and intelligibility metrics, demonstrated that the proposed deep TT-LSTM based speech enhancement framework can achieve competitive performances with the state-of-the-art uncompressed RNN model, even though the proposed model architecture is orders of magnitude less complex.

Melanie F. Pradier, Weiwei Pan, Jiayu Yao et al.

As machine learning systems get widely adopted for high-stake decisions, quantifying uncertainty over predictions becomes crucial. While modern neural networks are making remarkable gains in terms of predictive accuracy, characterizing uncertainty over the parameters of these models is challenging because of the high dimensionality and complex correlations of the network parameter space. This paper introduces a novel variational inference framework for Bayesian neural networks that (1) encodes complex distributions in high-dimensional parameter space with representations in a low-dimensional latent space, and (2) performs inference efficiently on the low-dimensional representations. Across a large array of synthetic and real-world datasets, we show that our method improves uncertainty characterization and model generalization when compared with methods that work directly in the parameter space.