Cellular automata as convolutional neural networks.

Deep-learning techniques have recently demonstrated broad success in predicting complex dynamical systems ranging from turbulence to human speech, motivating broader questions about how neural networks encode and represent dynamical rules. We explore this problem in the context of cellular automata (CA), simple dynamical systems that are intrinsically discrete and thus difficult to analyze using standard tools from dynamical systems theory. We show that any CA may readily be represented using a convolutional neural network with a network-in-network architecture. This motivates the development of a general convolutional multilayer perceptron architecture, which we find can learn the dynamical rules for arbitrary CA when given videos of the CA as training data. In the limit of large network widths, we find that training dynamics are nearly identical across replicates, and that common patterns emerge in the structure of networks trained on different CA rulesets. We train ensembles of networks on randomly sampled CA, and we probe how the trained networks internally represent the CA rules using an information-theoretic technique based on distributions of layer activation patterns. We find that CA with simpler rule tables produce trained networks with hierarchical structure and layer specialization, while more complex CA produce shallower representations-illustrating how the underlying complexity of the CA's rules influences the specificity of these internal representations. Our results suggest how the entropy of a physical process can affect its representation when learned by neural networks.

Cryptic selection forces and dynamic heritability in generalized phenotypic evolution.

Individuals with different phenotypes can have widely-varying responses to natural selection, yet many classical approaches to evolutionary dynamics emphasize only how a population's average phenotype increases in fitness over time. However, recent experimental results have produced examples of populations that have multiple fitness peaks, or that experience frequency-dependence that affects the direction and strength of selection on certain individuals. Here, we extend classical fitness gradient formulations of natural selection in order to describe the dynamics of a phenotype distribution in terms of its moments-such as the mean, variance, and skewness. The number of governing equations in our model can be adjusted in order to capture different degrees of detail about the population. We compare our simplified model to direct Wright-Fisher simulations of evolution in several canonical fitness landscapes, and we find that our model provides a low-dimensional description of complex dynamics not typically explained by classical theory, such as cryptic selection forces due to selection on trait ranges, time-variation of the heritability, and nonlinear responses to stabilizing or disruptive selection due to asymmetric trait distributions. In addition to providing a framework for extending general understanding of common qualitative concepts in phenotypic evolution - such as fitness gradients, selection pressures, and heritability - our approach has practical importance for studying evolution in contexts in which genetic analysis is infeasible.

Cryptographic hashing using chaotic hydrodynamics.

Fluids may store and manipulate information, enabling complex applications ranging from digital logic gates to algorithmic self-assembly. While controllable hydrodynamic chaos has previously been observed in viscous fluids and harnessed for efficient mixing, its application to the manipulation of digital information has been sparsely investigated. We show that chaotic stirring of a viscous fluid naturally produces a characteristic signature of the stirring process in the arrangement of particles in the fluid, and that this signature directly satisfies the requirements for a cryptographic hash function. This includes strong divergence between similar stirring protocols' hashes and avoidance of collisions (identical hashes from distinct stirs), which are facilitated by noninvertibility and a broad chaotic attractor that samples many points in the fluid domain. The hashing ability of the chaotic fluidic map implicates several unexpected mechanisms, including incomplete mixing at short time scales that produces a hyperuniform hash distribution. We investigate the dynamics of hashing using interparticle winding statistics, and find that hashing starts with large-scale winding of kinetically disjoint regions of the chaotic attractor, which gradually gives way to smaller scale braiding of single-particle trajectories. In addition to providing a physically motivated approach to implementing and analyzing deterministic chaotic maps for cryptographic applications, we anticipate that our approach has applications in microfluidic proof-of-work systems and characterizing large-scale turbulent flows from sparse tracer data.

Ecocultural range-expansion scenarios for the replacement or assimilation of Neanderthals by modern humans.

Recent archaeological records no longer support a simple dichotomous characterization of the cultures/behaviors of Neanderthals and modern humans, but indicate much cultural/behavioral variability over time and space. Thus, in modeling the replacement or assimilation of Neanderthals by modern humans, it is of interest to consider cultural dynamics and their relation to demographic change. The ecocultural framework for the competition between hominid species allows their carrying capacities to depend on some measure of the levels of culture they possess. In the present study both population densities and the densities of skilled individuals in Neanderthals and modern humans are spatially distributed and subject to change by spatial diffusion, ecological competition, and cultural transmission within each species. We analyze the resulting range expansions in terms of the demographic, ecological and cultural parameters that determine how the carrying capacities relate to the local densities of skilled individuals in each species. Of special interest is the case of cognitive and intrinsic-demographic equivalence of the two species. The range expansion dynamics may consist of multiple wave fronts of different speeds, each of which originates from a traveling wave solution. Properties of these traveling wave solutions are mathematically derived. Depending on the parameters, these traveling waves can result in replacement of Neanderthals by modern humans, or assimilation of the former by the latter. In both the replacement and assimilation scenarios, the first wave of intrusive modern humans is characterized by a low population density and a low density of skilled individuals, with implications for archaeological visibility. The first invasion is due to weak interspecific competition. A second wave of invasion may be induced by cultural differences between moderns and Neanderthals. Spatially and temporally extended coexistence of the two species, which would have facilitated the transfer of genes from Neanderthal into modern humans and vice versa, is observed in the traveling waves, except when niche overlap between the two species is extremely high. Archaeological findings on the spatial and temporal distributions of the Initial Upper Palaeolithic and the Early Upper Palaeolithic and of the coexistence of Neanderthals and modern humans are discussed.

Flowtrace: simple visualization of coherent structures in biological fluid flows.

We present a simple, intuitive algorithm for visualizing time-varying flow fields that can reveal complex flow structures with minimal user intervention. We apply this technique to a variety of biological systems, including the swimming currents of invertebrates and the collective motion of swarms of insects. We compare our results with more experimentally difficult and mathematically sophisticated techniques for identifying patterns in fluid flows, and suggest that our tool represents an essential 'middle ground' allowing experimentalists to easily determine whether a system exhibits interesting flow patterns and coherent structures without resorting to more intensive techniques. In addition to being informative, the visualizations generated by our tool are often striking and elegant, illustrating coherent structures directly from videos without the need for computational overlays. Our tool is available as fully documented open-source code for MATLAB, Python or ImageJ at www.flowtrace.org.

A phase transition induces chaos in a predator-prey ecosystem with a dynamic fitness landscape.

In many ecosystems, natural selection can occur quickly enough to influence the population dynamics and thus future selection. This suggests the importance of extending classical population dynamics models to include such eco-evolutionary processes. Here, we describe a predator-prey model in which the prey population growth depends on a prey density-dependent fitness landscape. We show that this two-species ecosystem is capable of exhibiting chaos even in the absence of external environmental variation or noise, and that the onset of chaotic dynamics is the result of the fitness landscape reversibly alternating between epochs of stabilizing and disruptive selection. We draw an analogy between the fitness function and the free energy in statistical mechanics, allowing us to use the physical theory of first-order phase transitions to understand the onset of rapid cycling in the chaotic predator-prey dynamics. We use quantitative techniques to study the relevance of our model to observational studies of complex ecosystems, finding that the evolution-driven chaotic dynamics confer community stability at the "edge of chaos" while creating a wide distribution of opportunities for speciation during epochs of disruptive selection-a potential observable signature of chaotic eco-evolutionary dynamics in experimental studies.

An ecocultural model predicts Neanderthal extinction through competition with modern humans.

Archaeologists argue that the replacement of Neanderthals by modern humans was driven by interspecific competition due to a difference in culture level. To assess the cogency of this argument, we construct and analyze an interspecific cultural competition model based on the Lotka-Volterra model, which is widely used in ecology, but which incorporates the culture level of a species as a variable interacting with population size. We investigate the conditions under which a difference in culture level between cognitively equivalent species, or alternatively a difference in underlying learning ability, may produce competitive exclusion of a comparatively (although not absolutely) large local Neanderthal population by an initially smaller modern human population. We find, in particular, that this competitive exclusion is more likely to occur when population growth occurs on a shorter timescale than cultural change, or when the competition coefficients of the Lotka-Volterra model depend on the difference in the culture levels of the interacting species.

PyPDB: a Python API for the Protein Data Bank.

SUMMARY: We have created a Python programming interface for the RCSB Protein Data Bank (PDB) that allows search and data retrieval for a wide range of result types, including BLAST and sequence motif queries. The API relies on the existing XML-based API and operates by creating custom XML requests from native Python types, allowing extensibility and straightforward modification. The package has the ability to perform many types of advanced search of the PDB that are otherwise only available through the PDB website. AVAILABILITY AND IMPLEMENTATION: PyPDB is implemented exclusively in Python 3 using standard libraries for maximal compatibility. The most up-to-date version, including iPython notebooks containing usage tutorials, is available free-of-charge under an open-source MIT license via GitHub at https://github.com/williamgilpin/pypdb, and the full API reference is at http://williamgilpin.github.io/pypdb_docs/html/. The latest stable release is also available on PyPI. CONTACT: wgilpin@stanford.edu.

Worms under Pressure: Bulk Mechanical Properties of C. elegans Are Independent of the Cuticle.

The mechanical properties of cells and tissues play a well-known role in physiology and disease. The model organism Caenorhabditis elegans exhibits mechanical properties that are still poorly understood, but are thought to be dominated by its collagen-rich outer cuticle. To our knowledge, we use a novel microfluidic technique to reveal that the worm responds linearly to low applied hydrostatic stress, exhibiting a volumetric compression with a bulk modulus, κ = 140 ± 20 kPa; applying negative pressures leads to volumetric expansion of the worm, with a similar bulk modulus. Surprisingly, however, we find that a variety of collagen mutants and pharmacological perturbations targeting the cuticle do not impact the bulk modulus. Moreover, the worm exhibits dramatic stiffening at higher stresses-behavior that is also independent of the cuticle. The stress-strain curves for all conditions can be scaled onto a master equation, suggesting that C. elegans exhibits a universal elastic response dominated by the mechanics of pressurized internal organs.