arXiv subjects

Zan Chaudhry

Publications and source records attributed to Zan Chaudhry.

4 recordsLinked to original sources

arXiv2026

Adaptive Label Error Detection: A Bayesian Approach to Mislabeled Data Detection

Machine learning classification systems are susceptible to poor performance when trained with incorrect ground truth labels, even when data is well-curated by expert annotators. As machine learning becomes more widespread, it is increasingly imperative to identify and correct mislabeling to develop more powerful models. In this work, we motivate and describe Adaptive Label Error Detection (ALED), a novel method of detecting mislabeling. ALED extracts an intermediate feature space from a deep convolutional neural network, denoises the features, models the reduced manifold of each class with a multidimensional Gaussian distribution, and performs a simple likelihood ratio test to identify mislabeled samples. We show that ALED has markedly increased sensitivity, without compromising precision, compared to established label error detection methods, on multiple medical imaging datasets. We demonstrate an example where fine-tuning a neural network on corrected data results in a 33.8% decrease in test set errors, providing strong benefits to end users. The ALED detector is deployed in the Python package statlab.

cs.LG↗

arXiv2024

A Simple Sparse Matrix Vector Multiplication Approach to Padded Convolution

We introduce an algorithm for efficiently representing convolution with zero-padding and stride as a sparse transformation matrix, applied to a vectorized input through sparse matrix-vector multiplication (SpMV). We provide a theoretical contribution with an explicit expression for the number of non-zero multiplications in convolutions with stride and padding, offering insight into the potential for leveraging sparsity in convolution operations. A proof-of-concept implementation is presented in Python, demonstrating the performance of our method on both CPU and GPU architectures. This work contributes to the broader exploration of sparse matrix techniques in convolutional algorithms, with a particular focus on leveraging matrix multiplications for parallelization. Our findings lay the groundwork for future advancements in exploiting sparsity to improve the efficiency of convolution operations in fields such as machine learning and signal processing.

cs.LG↗

arXiv2024

Invariant neuromorphic representations of tactile stimuli improve robustness of a real-time texture classification system

Humans have an exquisite sense of touch which robotic and prosthetic systems aim to recreate. We developed algorithms to create neuron-like (neuromorphic) spiking representations of texture that are invariant to the scanning speed and contact force applied in the sensing process. The spiking representations are based on mimicking activity from mechanoreceptors in human skin and further processing up to the brain. The neuromorphic encoding process transforms analog sensor readings into speed and force invariant spiking representations in three sequential stages: the force invariance module (in the analog domain), the spiking activity encoding module (transforms from analog to spiking domain), and the speed invariance module (in the spiking domain). The algorithms were tested on a tactile texture dataset collected in 15 speed-force conditions. An offline texture classification system built on the invariant representations has higher classification accuracy, improved computational efficiency, and increased capability to identify textures explored in novel speed-force conditions. The speed invariance algorithm was adapted to a real-time human-operated texture classification system. Similarly, the invariant representations improved classification accuracy, computational efficiency, and capability to identify textures explored in novel conditions. The invariant representation is even more crucial in this context due to human imprecision which seems to the classification system as a novel condition. These results demonstrate that invariant neuromorphic representations enable better performing neurorobotic tactile sensing systems. Furthermore, because the neuromorphic representations are based on biological processing, this work can be used in the future as the basis for naturalistic sensory feedback for upper limb amputees.

cs.RO↗

arXiv2024

ExpTest: Loss-Curve Hypothesis Testing for Autonomous Learning-Rate Selection in Deep Neural Networks

Hyperparameter tuning remains a significant challenge in the training of deep neural networks (DNNs), requiring manual search or time-intensive grid searches that increase resource costs and limit the accessibility of machine learning. The global initial learning rate is among the most consequential of these hyperparameters. Adaptive and scheduling-based methods manage the learning rate during training but still require manual selection of an initial global value; learning-rate-free alternatives remove this selection at the cost of performance or stability on non-convex problems. We present ExpTest, an autonomous learning-rate controller that treats the training loss curve as an online signal and performs sequential statistical tests on theoretically motivated windows to detect convergent behavior and trigger learning-rate reductions. The framework combines a covariance-based initial learning-rate estimate, curvature-motivated window sizing, and two-phase test-driven decay, drawing on the approximately exponential decay behavior predicted under linearized network dynamics. We provide a mathematical motivation for ExpTest and evaluate it on regression, classification, forecasting, and natural-language tasks across fully-connected, convolutional, transformer-based, and pretrained architectures. Across these tasks, ExpTest achieves competitive performance relative to hand-tuned SGD-based baselines and recent learning-rate-free methods, without manual initial learning-rate selection or predefined scheduling.

cs.LG↗