arXiv subjects

Meenakshi D'Souza

Publications and source records attributed to Meenakshi D'Souza.

3 recordsLinked to original sources

arXiv2026

SETA: Statistical Fault Attribution for Compound AI Systems

Modern AI systems increasingly comprise multiple interconnected neural networks to tackle complex inference tasks. Testing such systems for robustness and safety entails significant challenges. Current state-of-the-art robustness testing techniques, whether black-box or white-box, have been proposed and implemented for single-network models and do not scale well to multi-network pipelines. We propose a modular robustness testing framework that applies a given set of perturbations to test data. Our testing framework supports (1) a component-wise system analysis to isolate errors and (2) reasoning about error propagation across the neural network modules. The testing framework is architecture and modality agnostic and can be applied across domains. We apply the framework to a real-world autonomous rail inspection system composed of multiple deep networks and successfully demonstrate how our approach enables fine-grained robustness analysis beyond conventional end-to-end metrics.

cs.AI↗

arXiv2021

LipBaB: Computing exact Lipschitz constant of ReLU networks

The Lipschitz constant of neural networks plays an important role in several contexts of deep learning ranging from robustness certification and regularization to stability analysis of systems with neural network controllers. Obtaining tight bounds of the Lipschitz constant is therefore important. We introduce LipBaB, a branch and bound framework to compute certified bounds of the local Lipschitz constant of deep neural networks with ReLU activation functions up to any desired precision. We achieve this by bounding the norm of the Jacobians, corresponding to different activation patterns of the network caused within the input domain. Our algorithm can provide provably exact computation of the Lipschitz constant for any p-norm.

cs.LG↗

arXiv2021

Knowledge Value Stream Framework For Complex Product Design Decisions

Product Development value stream includes all the activities, value added and non value added, for designing and developing a product. It is characterized by flow of knowledge that drives the decisions, and deals with how the product is conceptualized, architected and designed by the design teams. The intangible flow of knowledge is determined by knowledge value stream, which shapes how the raw concepts and ideas flow into mature knowledge, how the knowledge is socialized, internalized, and how the knowledge impels the decisions in the product development value stream. For complex products, the design teams encounter tough challenges, such as uncertainty and variability, while making design decisions. This paper proposes a framework for knowledge value stream for complex product design decisions. The framework encompasses knowledge cadence and learning cycles as its core elements and incorporates rudiments for complex product design such as uncertainty, variability and perceptions. It helps in managing uncertainty and variability and evolving the required knowledge to drive optimal decisions during the design of complex products. It advocates a phased model for framework deployment towards establishing and progressively maturing the knowledge value stream.

cs.SI↗