Modular neural networks outperform nonmodular neural networks on tasks ranging from visual question answering to robotics. These performance improvements are thought to be due to modular networks' superior ability to model the compositional and combinatorial structure of real-world problems. However, a theoretical explanation of how modularity improves generalizability, and how to leverage task modularity while training networks remains elusive. Using recent theoretical progress in explaining neural network generalization, we investigate how the amount of training data required to generalize on a task varies with the intrinsic dimensionality of a task's input. We show theoretically that when applied to modularly structured tasks, while nonmodular networks require an exponential number of samples with task dimensionality, modular networks' sample complexity is independent of task dimensionality: modular networks can generalize in high dimensions. We then develop a novel learning rule for modular networks to exploit this advantage and empirically show the improved generalization of the rule, both in- and out-of-distribution, on high-dimensional, modular tasks.

State estimation is crucial for the performance and safety of numerous robotic applications. Among the suite of estimation techniques, particle filters have been identified as a powerful solution due to their non-parametric nature. Yet, in high-dimensional state spaces, these filters face challenges such as ’particle deprivation’ which hinders accurate representation of the true posterior distribution. This paper introduces a novel resampling-free particle filter designed to mitigate particle deprivation by forgoing the traditional resampling step. This ensures a broader and more diverse particle set, especially vital in high-dimensional scenarios. Theoretically, our proposed filter is shown to offer a near-accurate representation of the desired posterior distribution in high-dimensional contexts. Empirically, the effectiveness of our approach is underscored through a high-dimensional synthetic state estimation task and a 6D pose estimation derived from videos. We posit that as robotic systems evolve with greater degrees of freedom, particle filters tailored for high-dimensional state spaces will be indispensable.

Efficient haplotype matching search is of great interest when large genotyped cohorts are becoming available. Positional Burrows-Wheeler Transform (PBWT) enables efficient searching for blocks of haplotype matches. However, existing efficient PBWT algorithms sweep across the haplotype panel from left to right, capturing all exact matches. As a result, PBWT does not account for mismatches. It is also not easy to investigate the patterns of changes between the matching blocks. Here, we present an extension to PBWT, called bi-directional PBWT that allows the information about the blocks of matches to be present at both sides of each site. We also present a set of algorithms to efficiently merge the matching blocks or examine the patterns of changes on both sides of each site. The time complexity of the algorithms to find and merge matching blocks using bi-directional PBWT is linear to the input size. Using real data from the UK Biobank, we demonstrate the run time and memory efficiency of our algorithms. More importantly, our algorithms can identify more blocks by enabling tolerance of mismatches. Moreover, by using mutual information (MI) between the forward and the reverse PBWT matching block sets as a measure of haplotype consistency, we found the MI derived from European samples in the 1000 Genomes Project is highly correlated (Spearman correlation r=0.87) with the deCODE recombination map.