arXiv subjects

Manjot Singh

Publications and source records attributed to Manjot Singh.

9 recordsLinked to original sources

arXiv2026

Polyhedral Geometry of Time-to-First-Spike Neural Networks

We study the expressivity of spiking neural networks, which provide a natural framework for asynchronous, event-driven computation complementary to conventional feedforward neural networks. We consider the time-to-first-spike model in a setting for which the input-output map is continuous and piecewise linear, with affine pieces governed by causal feasibility constraints that determine which presynaptic spikes occur before a neuron fires. We first show that each neuron's firing time admits a maxout-like representation with exponentially many, highly constrained affine pieces. We then formalize causal regions as polyhedral regions with fixed causal sets and derive upper and lower bounds on the maximal number of causal regions in both shallow and multilayer feedforward spiking networks. Our theoretical and experimental results show that spiking networks can generate richer partitions of the input space than conventional feedforward ReLU networks.

cs.LG↗

arXiv2026

Radial Suppression Accelerates Algorithmic Generalization: A Geometric Analysis of Delayed Generalization

Why do neural networks memorize algorithmic training data long before they generalize? We present a geometric case study demonstrating that, on tasks where generalization requires discovering structured low-dimensional circuits, the memorization-generalization delay is driven by radial inflation of hidden representations under cross-entropy optimization. We formalize a radial-angular decomposition of activation-space dynamics and derive three testable propositions: (i) that penalizing radial inflation induces anisotropic, data-dependent weight regularization; (ii) that it suppresses radial gradient energy below the isotropic random baseline, forcing predominantly angular updates; and (iii) that it biases convergence toward flatter minima. To empirically validate these propositions, we study a single-hyperparameter norm penalty that softly constrains activations to a sqrt(d)-radius hypersphere. On modular arithmetic, this penalty accelerates grokking up to 6x across MLPs and Transformers, and halves training steps for a 10M-parameter nanoGPT on 3-digit addition.

cs.LG↗

arXiv2026

Are You Comfortable Sharing It?: Leveraging Image Obfuscation Techniques to Enhance Sharing Privacy for Blind and Visually Impaired Users

People with Blind Visual Impairments (BVI) face unique challenges when sharing images, as these may accidentally contain sensitive or inappropriate content. In many instances, they are unaware of the potential risks associated with sharing such content, which can compromise their privacy and interpersonal relationships. To address this issue, we investigated image filtering techniques that could help BVI users manage sensitive content before sharing with various audiences, including family, friends, or strangers. We conducted a study with 20 BVI participants, evaluating different filters applied to images varying in sensitivity, such as personal moments or embarrassing shots. Results indicated that pixelation was the least preferred method, while preferences for other filters varied depending on image type and sharing context. Additionally, participants reported greater comfort when sharing filtered versus unfiltered images across audiences. Based on the results, we offer a set of design guidelines to enhance the image-sharing experience for BVI individuals.

cs.HC↗

arXiv2025

Sustainable AI: Mathematical Foundations of Spiking Neural Networks

Deep learning's success comes with growing energy demands, raising concerns about the long-term sustainability of the field. Spiking neural networks, inspired by biological neurons, offer a promising alternative with potential computational and energy-efficiency gains. This article examines the computational properties of spiking networks through the lens of learning theory, focusing on expressivity, training, and generalization, as well as energy-efficient implementations while comparing them to artificial neural networks. By categorizing spiking models based on time representation and information encoding, we highlight their strengths, challenges, and potential as an alternative computational paradigm.

cs.NE↗

arXiv2024

Expressivity of Spiking Neural Networks

The synergy between spiking neural networks and neuromorphic hardware holds promise for the development of energy-efficient AI applications. Inspired by this potential, we revisit the foundational aspects to study the capabilities of spiking neural networks where information is encoded in the firing time of neurons. Under the Spike Response Model as a mathematical model of a spiking neuron with a linear response function, we compare the expressive power of artificial and spiking neural networks, where we initially show that they realize piecewise linear mappings. In contrast to ReLU networks, we prove that spiking neural networks can realize both continuous and discontinuous functions. Moreover, we provide complexity bounds on the size of spiking neural networks to emulate multi-layer (ReLU) neural networks. Restricting to the continuous setting, we also establish complexity bounds in the reverse direction for one-layer spiking neural networks.

cs.NE↗

arXiv2023

Incorporating Zero-Knowledge Succinct Non-interactive Argument of Knowledge for Blockchain-based Identity Management with off-chain computations

In today's world, secure and efficient biometric authentication is of keen importance. Traditional authentication methods are no longer considered reliable due to their susceptibility to cyber-attacks. Biometric authentication, particularly fingerprint authentication, has emerged as a promising alternative, but it raises concerns about the storage and use of biometric data, as well as centralized storage, which could make it vulnerable to cyber-attacks. In this paper, a novel blockchain-based fingerprint authentication system is proposed that integrates zk-SNARKs, which are zero-knowledge proofs that enable secure and efficient authentication without revealing sensitive biometric information. A KNN-based approach on the FVC2002, FVC2004 and FVC2006 datasets is used to generate a cancelable template for secure, faster, and robust biometric registration and authentication which is stored using the Interplanetary File System. The proposed approach provides an average accuracy of 99.01%, 98.97% and 98.52% over the FVC2002, FVC2004 and FVC2006 datasets respectively for fingerprint authentication. Incorporation of zk-SNARK facilitates smaller proof size. Overall, the proposed method has the potential to provide a secure and efficient solution for blockchain-based identity management.

cs.CR↗

arXiv2023

An Efficient Ensemble Explainable AI (XAI) Approach for Morphed Face Detection

The extensive utilization of biometric authentication systems have emanated attackers / imposters to forge user identity based on morphed images. In this attack, a synthetic image is produced and merged with genuine. Next, the resultant image is user for authentication. Numerous deep neural convolutional architectures have been proposed in literature for face Morphing Attack Detection (MADs) to prevent such attacks and lessen the risks associated with them. Although, deep learning models achieved optimal results in terms of performance, it is difficult to understand and analyse these networks since they are black box/opaque in nature. As a consequence, incorrect judgments may be made. There is, however, a dearth of literature that explains decision-making methods of black box deep learning models for biometric Presentation Attack Detection (PADs) or MADs that can aid the biometric community to have trust in deep learning-based biometric systems for identification and authentication in various security applications such as border control, criminal database establishment etc. In this work, we present a novel visual explanation approach named Ensemble XAI integrating Saliency maps, Class Activation Maps (CAM) and Gradient-CAM (Grad-CAM) to provide a more comprehensive visual explanation for a deep learning prognostic model (EfficientNet-B1) that we have employed to predict whether the input presented to a biometric authentication system is morphed or genuine. The experimentations have been performed on three publicly available datasets namely Face Research Lab London Set, Wide Multi-Channel Presentation Attack (WMCA), and Makeup Induced Face Spoofing (MIFS). The experimental evaluations affirms that the resultant visual explanations highlight more fine-grained details of image features/areas focused by EfficientNet-B1 to reach decisions along with appropriate reasoning.

cs.CV↗

arXiv2021

A systematic study of the valence electronic structure of cyclo(Gly-Phe), cyclo(Trp-Tyr) and cyclo(Trp-Trp) dipeptides in gas phase

The electronic energy levels of cyclo(Glycine-Phenylalanine), cyclo(Tryptophan-Tyrosine) and cyclo(Tryptophan-Tryptophan) dipeptides are investigated with a joint experimental and theoretical approach. Experimentally, valence photoelectron spectra in the gas phase are measured using VUV radiation. Theoretically, we first obtain low-energy conformers through an automated conformer-rotamer ensemble sampling scheme based on tight-binding simulations. Then, different first principles computational schemes are considered to simulate the spectra: Hartree-Fock (HF), density functional theory (DFT) within the B3LYP approximation, the quasi--particle GW correction, and the quantum-chemistry CCSD method. Theory allows to assign the main features of the spectra. A discussion on the role of electronic correlation is provided, by comparing computationally cheaper DFT scheme (and GW) results with the accurate CCSD method.

physics.chem-ph↗

arXiv2020

Different approaches to unveil biomolecule configurations and their mutual interactions

A novel technique was demonstrated that overcome important drawbacks to crosslink cells by irradiation with ultrashort UV laser pulses (L-crosslinking). To use this technique coupled to Chromatin ImmunoPrecipitation (ChIP) in a high throughput context, a pre-screening fast method needs to be implemented to set up suitable irradiation conditions of the cell sample for efficient L-crosslinking with no final and long ChIP analysis. Here a fast method is reported where living human cells have been first transfected with a vector coding for Estrogen Receptor α (ERα), linked to Green Florescent protein (ERα-GFP), so that the well-known interaction between the Estrogen Receptor Elements (ERE) region of the cell DNA and the ERα protein can be detected by studying the fluorometric response of the irradiated cells. The damage induced to cells by UV irradiation is characterized by looking at DNA integrity, proteins stability and cellular viability. A second novel approach is presented to analyze or re-visit DNA and RNA sequences and their molecular configurations. This approach is based on methods derived from Chern-Simons super-gravity adapted to describe mutations in DNA/RNA strings, as well as interactions between nucleic acids. As a preliminary case we analyze the KRAS human gene sequence and some of its mutations. Interestingly, our model shows how the Chern-Simons current are capable to characterize the mutations within a sequence, in particular giving a quantitative indication of the mutation likelihood.

q-bio.BM↗