Aditya Anand

Pirzada Suhail, Aditya Anand, Amit Sethi

Reliable and interpretable decision-making is essential in medical imaging, where diagnostic outcomes directly influence patient care. Despite advances in deep learning, most medical AI systems operate as opaque black boxes, providing little insight into why a particular diagnosis was reached. In this paper, we introduce Med-CAM, a framework for generating minimal and sharp maps as evidence-based explanations for Medical decision making via Classifier Activation Matching. Med-CAM trains a segmentation network from scratch to produce a mask that highlights the minimal evidence critical to model's decision for any seen or unseen image. This ensures that the explanation is both faithful to the network's behaviour and interpretable to clinicians. Experiments show, unlike prior spatial explanation methods, such as Grad-CAM and attention maps, which yield only fuzzy regions of relative importance, Med-CAM with its superior spatial awareness to shapes, textures, and boundaries, delivers conclusive, evidence-based explanations that faithfully replicate the model's prediction for any given image. By explicitly constraining explanations to be compact, consistent with model activations, and diagnostic alignment, Med-CAM advances transparent AI to foster clinician understanding and trust in high-stakes medical applications such as pathology and radiology.

Pirzada Suhail, Aditya Anand, Amit Sethi

In this paper we introduce an activation-matching--based approach to generate minimal, faithful explanations for the decision-making of a pretrained classifier on any given image. Given an input image $x$ and a frozen model $f$, we train a lightweight autoencoder to output a binary mask $m$ such that the explanation $e = m \odot x$ preserves both the model's prediction and the intermediate activations of \(x\). Our objective combines: (i) multi-layer activation matching with KL divergence to align distributions and cross-entropy to retain the top-1 label for both the image and the explanation; (ii) mask priors -- L1 area for minimality, a binarization penalty for crisp 0/1 masks, and total variation for compactness; and (iii) abductive constraints for faithfulness and necessity. Together, these objectives yield small, human-interpretable masks that retain classifier behavior while discarding irrelevant input regions, providing practical and faithful minimalist explanations for the decision making of the underlying model.

Vineetha Joy, Aditya Anand, Nidhi et al.

Metasurface-based radar absorbing structures (RAS) are highly preferred for applications like stealth technology, electromagnetic (EM) shielding, etc. due to their capability to achieve frequency selective absorption characteristics with minimal thickness and reduced weight penalty. However, the conventional approach for the EM design and optimization of these structures relies on forward simulations, using full wave simulation tools, to predict the electromagnetic (EM) response of candidate meta atoms. This process is computationally intensive, extremely time consuming and requires exploration of large design spaces. To overcome this challenge, we propose a surrogate model that significantly accelerates the prediction of EM responses of multi-layered metasurface-based RAS. A convolutional neural network (CNN) based architecture with Huber loss function has been employed to estimate the reflection characteristics of the RAS model. The proposed model achieved a cosine similarity of 99.9% and a mean square error of 0.001 within 1000 epochs of training. The efficiency of the model has been established via full wave simulations as well as experiment where it demonstrated significant reduction in computational time while maintaining high predictive accuracy.

Pirzada Suhail, Aditya Anand, Amit Sethi

Machine learning models, by virtue of training, learn a large repertoire of decision rules for any given input, and any one of these may suffice to justify a prediction. However, in high-dimensional input spaces, such rules are difficult to identify and interpret. In this paper, we introduce an activation-matching based approach to generate minimal and faithful explanations for the decisions of pre-trained image classifiers. We aim to identify minimal explanations that not only preserve the model's decision but are also concise and human-readable. To achieve this, we train a lightweight autoencoder to produce binary masks that learns to highlight the decision-wise critical regions of an image while discarding irrelevant background. The training objective integrates activation alignment across multiple layers, consistency at the output label, priors that encourage sparsity, and compactness, along with a robustness constraint that enforces faithfulness. The minimal explanations so generated also lead us to mechanistically interpreting the model internals. In this regard we also introduce a circuit readout procedure wherein using the explanation's forward pass and gradients, we identify active channels and construct a channel-level graph, scoring inter-layer edges by ingress weight magnitude times source activation and feature-to-class links by classifier weight magnitude times feature activation. Together, these contributions provide a practical bridge between minimal input-level explanations and a mechanistic understanding of the internal computations driving model decisions.

Ayush Chauhan, Aditya Anand, Shaddy Garg et al.

Often, data contains only composite events composed of multiple events, some observed and some unobserved. For example, search ad click is observed by a brand, whereas which customers were shown a search ad - an actionable variable - is often not observed. In such cases, inference is not possible on unobserved event. This occurs when a marketing action is taken over earned and paid digital channels. Similar setting arises in numerous datasets where multiple actors interact. One approach is to use the composite event as a proxy for the unobserved event of interest. However, this leads to invalid inference. This paper takes a direct approach whereby an event of interest is identified based on information on the composite event and aggregate data on composite events (e.g. total number of search ads shown). This work contributes to the literature by proving identification of the unobserved events' probabilities up to a scalar factor under mild condition. We propose an approach to identify the scalar factor by using aggregate data that is usually available from earned and paid channels. The factor is identified by adding a loss term to the usual cross-entropy loss. We validate the approach on three synthetic datasets. In addition, the approach is validated on a real marketing problem where some observed events are hidden from the algorithm for validation. The proposed modification to the cross-entropy loss function improves the average performance by 46%.

Aditya Anand, Palash Dey

We introduce the notion of {\em Distance Restricted Manipulation}, where colluding manipulator(s) need to compute if there exist votes which make their preferred alternative win the election when their knowledge about the others' votes is a little inaccurate. We use the Kendall-Tau distance to model the manipulators' confidence in the non-manipulators' votes. To this end, we study this problem in two settings - one where the manipulators need to compute a manipulating vote that succeeds irrespective of perturbations in others' votes ({\em Distance Restricted Strong Manipulation}), and the second where the manipulators need to compute a manipulating vote that succeeds for at least one possible vote profile of the others ({\em Distance Restricted Weak Manipulation}). We show that {\em Distance Restricted Strong Manipulation} admits polynomial-time algorithms for every scoring rule, maximin, Bucklin, and simplified Bucklin voting rules for a single manipulator, and for the $k$-approval rule for any number of manipulators, but becomes intractable for the Copeland$^α$ voting rule for every $α\in[0,1]$ even for a single manipulator. In contrast, {\em Distance Restricted Weak Manipulation} is intractable for almost all the common voting rules, with the exception of the plurality rule. For a constant number of alternatives, we show that both the problems are polynomial-time solvable for every anonymous and efficient voting rule.