Incremental learning aims to adapt to new sets of categories over time with minimal computational overhead. Prior work often addresses this task by training efficient task-specific adaptors that modify frozen layer weights or features to capture relevant information without affecting predictions on any previously learned categories. While these adaptors are generally more efficient than finetuning the entire network, they still can require tens to hundreds of thousands task-specific trainable parameters even for relatively small networks. This makes it challenging to operate in resource-constrained environments with high communication costs, such as edge devices or mobile phones. Thus, we propose Reparameterized, Compact weight Adaptation for Sequential Tasks (RECAST), a novel method that dramatically reduces the number of task-specific trainable parameters to fewer than 50 – several orders of magnitude less than competing methods like LoRA. RECAST accomplishes this efficiency by learning to decompose layer weights into a soft parameter-sharing framework consisting of a set of shared weight templates and very few module-specific scaling factors or coefficients. This soft parameter-sharing framework allows for effective task-wise reparameterization by tuning only these coefficients while keeping templates frozen. A key innovation of RECAST is the novel weight reconstruction pipeline called Neural Mimicry, which eliminates the need for pretraining from scratch. This allows for high-fidelity emulation of existing pretrained weights within our framework and provides quick adaptability to any model scale and architecture. Extensive experiments across six diverse datasets demonstrate RECAST outperforms the state-of-the-art by up to ∼ 1.5% and improves baselines by > 3% across various scales, architectures, and parameter spaces. Moreover, we show that RECAST’s architecture-agnostic nature allows for seamless integration with existing methods, further boosting performance.

BACKGROUND: Coronavirus Disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has become a global pandemic following its initial emergence in China. SARS-CoV-2 has a positive-sense single-stranded RNA virus genome of around 30Kb. Using next-generation sequencing technologies, a large number of SARS-CoV-2 genomes are being sequenced at an unprecedented rate and being deposited in public repositories. For the de novo assembly of the SARS-CoV-2 genomes, a myriad of assemblers is being used, although their impact on the assembly quality has not been characterized for this virus. In this study, we aim to understand the variabilities on assembly qualities due to the choice of the assemblers. RESULTS: We performed 6648 de novo assemblies of 416 SARS-CoV-2 samples using eight different assemblers with different k-mer lengths. We used Illumina paired-end sequencing reads and compared the assembly quality of those assemblers. We showed that the choice of assembler plays a significant role in reconstructing the SARS-CoV-2 genome. Two metagenomic assemblers, e.g. MEGAHIT and metaSPAdes, performed better compared with others in most of the assembly quality metrics including, recovery of a larger fraction of the genome, constructing larger contigs and higher N50, NA50 values, etc. We showed that at least 09% (259/2873) of the variants present in the assemblies between MEGAHIT and metaSPAdes are unique to one of the assembly methods. CONCLUSION: Our analyses indicate the critical role of assembly methods for assembling SARS-CoV-2 genome using short reads and their impact on variant characterization. This study could help guide future studies to determine the best-suited assembler for the de novo assembly of virus genomes.