arXiv Science⌕ Search

arXiv subjects

Soham Mitra

Publications and source records attributed to Soham Mitra.

4 recordsLinked to original sources

When Cell Crops Share a Parent: Measuring and Isolating Acquisition Leakage in Cervical Cytology Benchmarks

Public cervical cytology benchmarks release many cell crops cut from far fewer microscope fields or slides and are often split at the crop level, so test cells share acquisition parents with training cells. In a preregistered study, crop-level splitting raised SIPaKMeD accuracy by 0.97 percentage points across twelve architectures and twenty seeds (95% CI 0.59-1.39), and CRIC accuracy by 12.08 points (11.16-12.99). On SIPaKMeD, a factorial that fixes training crop and parent counts attributes the gain to sibling exposure (1.73 points, 0.95-2.51), not to the extra parent diversity of crop-level splits (0.00). Parent identity is decodable at about 100 times chance, yet erasing 82% of its above-floor linear recoverability left the gap unchanged, and Grad-CAM did not distinguish leaking from clean models; matching test cells' color statistics to an unseen same-class field removed most of the advantage. Decodability and saliency cannot substitute for a parent-disjoint split.

cs.CV↗

Resisting Drying Stress in Laponite Bentonite Clay Composites, Two Step Yielding and Crack Peel Selection

Clay films that dry on a rigid substrate crack, peel or stay intact, and which of these happens matters for coatings, barrier layers and ceramic processing. Composites of clay with polymers have been studied widely. Clay composites, whose components differ in size and charge, are much less explored. We show that mixing the nano scale clay Laponite with the micron scale clay Bentonite reduces desiccation damage on both a hydrophobic and a hydrophilic substrate, Laponite films stay essentially intact, Bentonite films fracture extensively, and on glass most intermediate compositions are less damaged than a simple average of the two would suggest. Rheology of the gelled suspensions reveals a two step yielding in every composition, and the cohesive energy derived from it depends by orders of magnitude on which step is used, a point that any rheology cracking correlation must specify. A dimensionless cohesion adhesion index, whose sign we derive from an energy balance, is consistent with compositions that peel on polypropylene and crack on glass. Replicate experiments were performed for robustness of results. The results identify composition as a simple handle on the crack resistance of clay films and open the search for the critical crack-free thickness as a function of clay composition.

cond-mat.soft↗

Prospect of Constraining the EoS of Neutron Stars Using Post-Merger Signals

Post-merger gravitational-wave emission from a binary neutron star merger carries crucial information about the equation of state (EoS) of matter at high temperatures. Current gravitational wave detectors have limited sensitivities at post-merger frequencies in the range [1.5, 4] kHz. Therefore, valuable inferences can only be made after combining information from multiple (BNS) events. Criswell et al. [Phys. Rev. D 107, 043021 (2023)] carries out an injection study to infer the radius posterior for a $1.6 M_{\odot}$ NS by combining the information from injected BNS events via the hierarchical Bayesian inference (HBI) formulation. This formulation utilizes empirical relations that connect the peak frequency of the post-merger remnant with the chirp mass of the system, and the EoS proxy parameter $R_{1.6}$. In this work, we extend the HBI formulation to other EoS proxy parameters (i.e., the radius of (NS), $R_{1.X}$, with masses 1.2, 1.4, and 1.8 $M_{\odot}$) and combine the four $R_{1.X}$ posteriors through the piecewise polytropic EoS model to obtain measurable constraints on the EoS of the NS. We show that the NS radii can be constrained to within $\sim$1 km ( $\sim$0.55 km) assuming uniform (astrophysical) prior on $R_{1.X}$ for injections in the A+ noise. We also study systematic biases in the analysis coming from the limitations of empirical relations.

astro-ph.HE↗

Reliable or Deceptive? Investigating Gated Features for Smooth Visual Explanations in CNNs

Deep learning models have achieved remarkable success across diverse domains. However, the intricate nature of these models often impedes a clear understanding of their decision-making processes. This is where Explainable AI (XAI) becomes indispensable, offering intuitive explanations for model decisions. In this work, we propose a simple yet highly effective approach, ScoreCAM++, which introduces modifications to enhance the promising ScoreCAM method for visual explainability. Our proposed approach involves altering the normalization function within the activation layer utilized in ScoreCAM, resulting in significantly improved results compared to previous efforts. Additionally, we apply an activation function to the upsampled activation layers to enhance interpretability. This improvement is achieved by selectively gating lower-priority values within the activation layer. Through extensive experiments and qualitative comparisons, we demonstrate that ScoreCAM++ consistently achieves notably superior performance and fairness in interpreting the decision-making process compared to both ScoreCAM and previous methods.

cs.CV↗