Search for: All records

Abstract Various biological and synthetic media out of equilibrium can be viewed as many-ratchet systems that rectify environmental noise through local measurements and information processing, like Maxwell’s prototypical demon. These systems pose a challenge to standard coarse-graining approaches because they are described in terms of decision-making protocols similar to computer programs rather than force laws. Here, we study a many-body generalization of the Maxwell demon problem: a fluid composed of adaptive particles that achieve collective behavior by biasing noise-driven scattering events subject to measurements. Using a combination of information-theoretic, kinetic, and hydrodynamic tools, we elucidate how microscopic decision-making protocols, rather than microscopic forces, generate macroscopic active states sustained by continuous measurements. These include an informational version of flocking whose order parameter is bounded by the information measured, and the onset of which may be viewed as a measurement-induced phase transition. We find that the signature of such microscopic choices is an “informational activity” that selectively compresses phase space, without work, and causes deviations from equilibrium scaling with the magnitude of environmental noise. We envision applications to noise-induced patterning performed by collections of microrobots guided by reinforcement learning or programmable phoretic colloids in turbulent flows that exploit local measurements and control actions to counteract the scrambling of information by chaos. 

Abstract Multiple SARS-CoV-2 variants have arisen during the first years of the pandemic, often bearing many new mutations. Several explanations have been offered for the surprisingly sudden emergence of multiple mutations that enhance viral fitness, including cryptic transmission, spillover from animal reservoirs, epistasis between mutations, and chronic infections. Here, we simulated pathogen evolution combining within-host replication and between-host transmission. We found that, under certain conditions, chronic infections can lead to SARS-CoV-2-like bursts of mutations even without epistasis. Chronic infections can also increase the global evolutionary rate of a pathogen even in the absence of clear mutational bursts. Overall, our study supports chronic infections as a plausible origin for highly mutated SARS-CoV-2 variants. More generally, we also describe how chronic infections can influence pathogen evolution under different scenarios. 

Abstract Cortical populations are in a broadly asynchronous state that is sporadically interrupted by brief epochs of coordinated population activity. Inhibitory stabilized networks reproduce a low activity asynchronous regime but cannot generate population events. In contrast, synaptic depression stabilized excitatory networks create transient surges of activity, yet give inhibition only a perfunctory role. We analyzed spontaneously active in vitro mouse auditory cortex slices that show both regimes, including slow (2–12 Hz) oscillations within some events. We built firing rate and biophysically realistic spiking models in which excitation is balanced by recurrent inhibition, yet all excitatory synapses undergo short term depression. In our model a depression of synaptic excitation onto inhibition neurons initiates events, while depression of excitation onto excitatory neurons shapes rhythmicity of the events, reproducing the full repertoire observed experimentally. Our work unifies balanced and depression stabilized theories and provides a mechanistic framework for nonlinear, population wide correlations in cortex. 

Abstract The backpropagation method has enabled transformative uses of neural networks. Alternatively, for energy-based models, local learning methods involving only nearby neurons offer benefits in terms of decentralized training, and allow for the possibility of learning in computationally-constrained substrates. One class of local learning methodscontraststhe desired, clamped behavior with spontaneous, free behavior. However, directly contrasting free and clamped behaviors requires explicit memory. Here, we introduce ‘Temporal Contrastive Learning’, an approach that uses integral feedback in each learning degree of freedom to provide a simple form of implicit non-equilibrium memory. During training, free and clamped behaviors are shown in a sawtooth-like protocol over time. When combined with integral feedback dynamics, these alternating temporal protocols generate an implicit memory necessary for comparing free and clamped behaviors, broadening the range of physical and biological systems capable of contrastive learning. Finally, we show that non-equilibrium dissipation improves learning quality and determine a Landauer-like energy cost of contrastive learning through physical dynamics. 

Abstract Fully developed turbulence is a universal and scale-invariant chaotic state characterized by an energy cascade from large to small scales at which the cascade is eventually arrested by dissipation^1–6. Here we show how to harness these seemingly structureless turbulent cascades to generate patterns. Pattern formation entails a process of wavelength selection, which can usually be traced to the linear instability of a homogeneous state⁷. By contrast, the mechanism we propose here is fully nonlinear. It is triggered by the non-dissipative arrest of turbulent cascades: energy piles up at an intermediate scale, which is neither the system size nor the smallest scales at which energy is usually dissipated. Using a combination of theory and large-scale simulations, we show that the tunable wavelength of these cascade-induced patterns can be set by a non-dissipative transport coefficient called odd viscosity, ubiquitous in chiral fluids ranging from bioactive to quantum systems^8–12. Odd viscosity, which acts as a scale-dependent Coriolis-like force, leads to a two-dimensionalization of the flow at small scales, in contrast with rotating fluids in which a two-dimensionalization occurs at large scales⁴. Apart from odd viscosity fluids, we discuss how cascade-induced patterns can arise in natural systems, including atmospheric flows^13–19, stellar plasma such as the solar wind^20–22, or the pulverization and coagulation of objects or droplets in which mass rather than energy cascades^23–25. 

Biological systems, with many interacting components, face high-dimensional environmental fluctuations, ranging from diverse nutrient deprivations to toxins, drugs, and physical stresses. Yet, many biological control mechanisms are “simple,” i.e., restoring homeostasis through low-dimensional representations of the system’s high-dimensional state. How do low-dimensional controllers maintain homeostasis in high-dimensional systems? We develop an analytically tractable model of integral feedback for complex systems in fluctuating environments. We find that selection for homeostasis leads to the emergence of a soft mode that provides the dimensionality reduction required for the functioning of simple controllers. Our theory predicts that simple controllers that buffer environmental perturbations (e.g., stress response pathways) will also buffer mutational perturbation, an equivalence we test using experimental data across 5,000 strains in the yeast knockout collection. We also predict, counterintuitively, that knocking out a simple controller will decrease the dimensionality of the response to environmental change; we outline transcriptomics tests to validate this. Our work suggests an evolutionary origin of soft modes, with implications ranging from cryptic genetic variation to global epistasis.