Mridul Sahu | Searcharxiv

arXiv2026

Orbax: Distributed Checkpointing with JAX

In a landscape of high-performance distributed ML systems, JAX has emerged as a framework of choice. However, JAX's modular design philosophy leaves it without a standardized checkpointing solution. In this paper, we introduce Orbax, a modular, JAX-native checkpointing library that abstracts the complexities of distributed accelerator systems while also providing flexibility for user-friendly checkpoint manipulations throughout the ML model lifecycle. We demonstrate performance exceeding comparable PyTorch competitors by up to 3.5$\times$ for saving and 2$\times$ for loading. The library is available at https://github.com/google/orbax.

cs.DC↗

arXiv2018

EvoCut : A new Generalization of Albert-Barabási Model for Evolution of Complex Networks

With the evolution of social networks, the network structure shows dynamic nature in which nodes and edges appear as well as disappear for various reasons. The role of a node in the network is presented as the number of interactions it has with the other nodes. For this purpose a network is modeled as a graph where nodes represent network members and edges represent a relationship among them. Several models for evolution of social networks has been proposed till date, most widely accepted being the Barabási-Albert \cite{Network science} model that is based on \emph{preferential attachment} of nodes according to the degree distribution. This model leads to generation of graphs that are called \emph{Scale Free} and the degree distribution of such graphs follow the \emph{power law}. Several generalizations of this model has also been proposed. In this paper we present a new generalization of the model and attempt to bring out its implications in real life.

cs.SI↗