Inspired by natural biochemicals that perform complex information processing within living cells, we design and simulate a chemically implemented feedforward neural network, which learns by a novel chemical-reaction-based analogue of backpropagation. Our network is implemented in a simulated chemical system, where individual neurons are separated from each other by semipermeable cell-like membranes. Our compartmentalized, modular design allows a variety of network topologies to be constructed from the same building blocks. This brings us towards general-purpose, adaptive learning in chemico: wet machine learning in an embodied dynamical system.

It is obviously useful to think of evolved individuals in terms of their adaptations, yet the task of empirically classifying traits as adaptations has been claimed by some to be impossible in principle. I reject that claim by construction, introducing a formal method to empirically test whether a trait is an adaptation. The method presented is general, intuitive, and effective at identifying adaptations while remaining agnostic about their adaptive function. The test follows directly from the notion that adaptations arise from variation, heritability, and differential fitness in an evolving population: I operationalize these three concepts at the trait level, formally defining measures of individual traits. To test whether a trait is an adaptation, these measures are evaluated, locating the trait within a three-dimensional parameterized trait space. Within this space, I identify a region containing all adaptations; a trait's position relative to this adaptive region of trait space describes its status as an adaptation. The test can be applied in any evolving system where a few domain-specific statistical measures can be constructed; I demonstrate the construction of these measures, most notably a measure of an individual's hypothetical fitness if it were born with a different trait, in Packard's Bugs ALife model. The test is applied in Bugs, and shown to conform with our intuitive classification of adaptations.