Study Reveals Stability Challenges in Neural Network Growth During Training
Researchers have identified a fundamental asymmetry between growing and pruning neural networks during training, showing that newly inserted units are often 'forward-active but backward-starved.' While pruning selects from units trained from the start, growth inserts new units into an already specialized optimization trajectory, creating a gradient signal disadvantage. The findings suggest that neural network growth should be evaluated as a time-sensitive optimization process, not merely an architecture-search tool.
A preprint posted to arXiv investigates structural plasticity in deep learning, specifically comparing two approaches to modifying network architecture during training: pruning existing neurons and growing new ones. The authors demonstrate that these two operations are not symmetric — pruning selects among units that have been part of training from the beginning, while newly grown units enter an optimization landscape already shaped by incumbent neurons. This creates what the authors call a 'forward-active but backward-starved' condition, where new units participate in forward computation but receive substantially weaker gradient signals. The effect is minor in small MLP benchmarks but becomes pronounced in harder image-classification tasks with convolutional architectures. In continual-learning settings designed to stress plasticity loss, the Grow approach becomes competitive primarily when new units are given sufficient time to integrate into the network. The authors tested several interventions — targeting optimizer state, insertion strategy, selection, and trainability — finding that better integration of new units can improve adaptive performance but does not automatically yield better final subnetworks compared to pruning followed by retraining from scratch.
What's missing
As a preprint, this work has not yet undergone formal peer review. The study's own scope is limited to specific benchmark settings (MLPs and convolutional networks on image classification and continual learning); generalization to other architectures such as transformers or large language models is not established. The paper does not fully resolve which specific intervention most reliably closes the performance gap between Grow and Prune, leaving open the question of a practical best-practice recommendation.
What different sources said
- arXiv cs.LGCenter
On the Stability of Growth in Structural Plasticity
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.