SmartMixed: Two-Phase Training Strategy Enables Neural Networks to Learn Optimal Per-Neuron Activation Functions
Researchers have introduced SmartMixed, a training strategy that enables individual neurons in a neural network to adaptively select their own activation functions from a candidate pool, then locks those choices in for efficient inference. Unlike conventional architectures that apply a single, uniform activation function across all neurons, SmartMixed uses a differentiable hard mixture mechanism in a first phase and fixes selections in a second phase. The approach offers a potential path toward more functionally diverse and computationally efficient neural networks without the overhead of dynamic activation at inference time.
SmartMixed is a two-phase training strategy proposed by Amin Omidvar and described in a preprint submitted to arXiv. In the first phase, each neuron independently learns to select from a pool of six candidate activation functions — ReLU, Sigmoid, Tanh, Leaky ReLU, ELU, and SELU — via a differentiable hard mixture mechanism that allows gradient-based optimization. In the second phase, each neuron's chosen activation function is frozen, enabling the network to use optimized vectorized operations during continued training and inference, preserving computational efficiency. Experiments were conducted on the MNIST handwritten digit dataset using feedforward neural networks of varying architectures. Results indicate that neurons in different layers develop distinct preferences for activation functions, suggesting that functional diversity across a network may be beneficial. SmartMixed was shown to compete favorably against models using a single fixed state-of-the-art activation function, though evaluations remain limited to a relatively simple benchmark dataset.
What's missing
The study evaluates SmartMixed exclusively on MNIST, a well-known but relatively simple benchmark; it is unclear how the method scales to more complex datasets (e.g., ImageNet), deeper architectures (e.g., transformers or convolutional networks), or tasks beyond image classification. The paper does not report wall-clock training time comparisons or memory overhead during Phase 1, nor does it address whether the learned activation-function distributions generalize across random seeds or dataset splits. The work is a preprint and has not yet undergone formal peer review.
What different sources said
- arXiv cs.AICenter
SmartMixed: A Two-Phase Training Strategy for Adaptive Activation Function Learning in Neural Networks
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.