Chrisantha Fernando

Pattern Recognition in a Bucket

This paper demonstrates that the waves produced on the surface of water can be used as the medium for a “Liquid State Machine” that pre-processes inputs so allowing a simple perceptron to solve the XOR problem and undertake speech recognition. Interference between waves allows non-linear parallel computation upon simultaneous sensory inputs. Temporal patterns of stimulation are converted to spatial patterns of water waves upon which a linear discrimination can be made. Whereas Wolfgang Maass’ Liquid State Machine requires fine tuning of the spiking neural network parameters, water has inherent self-organising properties such as strong local interactions, time-dependent spread of activation to distant areas, inherent stability to a wide variety of inputs, and high complexity. Water achieves this “for free”, and does so without the time-consuming computation required by realistic neural models. An analogy is made between water molecules and neurons in a recurrent neural network.

Molecular circuits for associative learning in single-celled organisms

We demonstrate how a single-celled organism could undertake associative learning. Although to date only one previous study has found experimental evidence for such learning, there is no reason in principle why it should not occur. We propose a gene regulatory network that is capable of associative learning between any pre-specified set of chemical signals, in a Hebbian manner, within a single cell. A mathematical model is developed, and simulations show a clear learned response. A preliminary design for implementing this model using plasmids within Escherichia coli is presented, along with an alternative approach, based on double-phosphorylated protein kinases.

Selectionist and Evolutionary Approaches to Brain Function: A Critical Appraisal

We consider approaches to brain dynamics and function that have been claimed to be Darwinian. These include Edelman’s theory of neuronal group selection, Changeux’s theory of synaptic selection and selective stabilization of pre-representations, Seung’s Darwinian synapse, Loewenstein’s synaptic melioration, Adam’s selfish synapse, and Calvin’s replicating activity patterns. Except for the last two, the proposed mechanisms are selectionist but not truly Darwinian, because no replicators with information transfer to copies and hereditary variation can be identified in them. All of them fit, however, a generalized selectionist framework conforming to the picture of Price’s covariance formulation, which deliberately was not specific even to selection in biology, and therefore does not imply an algorithmic picture of biological evolution. Bayesian models and reinforcement learning are formally in agreement with selection dynamics. A classification of search algorithms is shown to include Darwinian replicators (evolutionary units with multiplication, heredity, and variability) as the most powerful mechanism for search in a sparsely occupied search space. Examples are given of cases where parallel competitive search with information transfer among the units is more efficient than search without information transfer between units. Finally, we review our recent attempts to construct and analyze simple models of true Darwinian evolutionary units in the brain in terms of connectivity and activity copying of neuronal groups. Although none of the proposed neuronal replicators include miraculous mechanisms, their identification remains a challenge but also a great promise.

The neuronal replicator hypothesis

We propose that replication (with mutation) of patterns of neuronal activity can occur within the brain using known neurophysiological processes. Thereby evolutionary algorithms implemented by neuro- nal circuits can play a role in cognition. Replication of structured neuronal representations is assumed in several cognitive architectures. Replicators overcome some limitations of selectionist models of neuronal search. Hebbian learning is combined with replication to structure exploration on the basis of associations learned in the past. Neuromodulatory gating of sets of bistable neurons allows patterns of activation to be copied with mutation. If the probability of copying a set is related to the utility of that set, then an evolutionary algorithm can be implemented at rapid timescales in the brain. Populations of neuronal replicators can undertake a more rapid and stable search than can be achieved by serial modification of a single solution. Hebbian learning added to neuronal replication allows a powerful structuring of variability capable of learning the location of a global optimum from multiple previously visited local optima. Replication of solutions can solve the problem of catastrophic forgetting in the stability-plasticity dilemma. In short, neuronal replication is essential to explain several features of flexible cognition. Predictions are made for the experimental validation of the neuronal replicator hypothesis.

A Stochastic Model of Nonenzymatic Nucleic Acid Replication: “Elongators” Sequester Replicators

The origin of nucleic acid template replication is a major unsolved problem in science. A novel stochastic model of nucleic acid chemistry was developed to allow rapid prototyping of chemical experiments designed to discover sufficient conditions for template replication. Experiments using the model brought to attention a robust property of nucleic acid template populations, the tendency for elongation to outcompete replication. Externally imposed denaturation-renaturation cycles did not reverse this tendency. For example, it has been proposed that fast tidal cycling could establish a TCR (tidal chain reaction) analogous to a PCR (polymerase chain reaction) acting on nucleic acid polymers, allowing their self-replication. However, elongating side-reactions that would have been prevented by the polymerase in the PCR still occurred in the simulation of the TCR. The same finding was found with temperature and monomer cycles. We propose that if cycling reactors are to allow template replication, oligonucleotide phenotypes that are capable of favorably altering the flux ratio between replication and elongation, for example, by facilitating sequence-specific cleavage within templates, are necessary; accordingly the minimal replicase ribozyme may have possessed restriction functionality.

What is comparable in comparative cognition

To understand how complex, or ‘advanced’ various forms of cognition are, and to compare them between species for evolutionary studies, we need to understand the diversity of neural–computational mechanisms that may be involved, and to identify the genetic changes that are necessary to mediate changes in cognitive functions. The same overt cognitive capacity might be mediated by entirely different neural circuitries in different species, with a many-to-one mapping between behavioural routines, computations and their neural implementations. Comparative behavioural research needs to be complemented with a bottom-up approach in which neurobiological and molecular-genetic analyses allow pinpointing of underlying neural and genetic bases that constrain cognitive variation. Often, only very minor differences in circuitry might be needed to generate major shifts in cognitive functions and the possibility that cognitive traits arise by convergence or parallel evolution needs to be taken seriously. Hereditary variation in cognitive traits between individuals of a species might be extensive, and selection experiments on cognitive traits might be a useful avenue to explore how rapidly changes in cognitive abilities occur in the face of pertinent selection pressures.

Copying and evolution of neuronal topology.

We propose a mechanism for copying of neuronal networks that is of considerable interest for neuroscience for it suggests a neuronal basis for causal inference, function copying, and natural selection within the human brain. To date, no model of neuronal topology copying exists. We present three increasingly sophisticated mechanisms to demonstrate how topographic map formation coupled with Spike-Time Dependent Plasticity (STDP) can copy neuronal topology motifs. Fidelity is improved by error correction and activity-reverberation limitation. The high-fidelity topology-copying operator is used to evolve neuronal topologies. Possible roles for neuronal natural selection are discussed.

The origin of autonomous agents by natural selection.

We propose conditions in which an autonomous agent could arise, and increase in complexity. It is assumed that on the primitive Earth there arose a recycling flow-reactor containing spontaneously formed oil droplets or lipid aggregates. These droplets grew at a basal rate by simple incorporation of lipid phase material, and divided by external agitation. This type of system was able to implement a natural selection algorithm once heredity was added. Macroevolution became possible by selection for rarely occurring chemical reactions that produced holistic autocatalytic molecular replicators (contained within the aggregate) capable of doubling at least as fast as the lipid aggregate, and which were also capable of benefiting the growth of its lipid aggregate container. No nucleotides or monomers capable of modular heredity were required at the outset. To explicitly state this hypothesis, a computer model was developed that employed an artificial chemistry, exhibiting conservation of mass and energy, incorporated within each individual of a population of lipid aggregates. This model evolved increasingly complex self-sustaining processes of constitution, a result that is also expected in real chemistry.

Is there a Liquid State Machine in the Bacterium Escherichia Coli

The bacterium Escherichia coli has the capacity to respond to a wide range of environmental inputs, which have the potential to change suddenly and rapidly. Although the functions of many of its signal transduction and gene regulation networks have been identified, E.Colis capacity for perceptual categorization, especially for discrimination between complex temporal patterns of chemical inputs, has been experimentally neglected. Real-time computations on time-varying inputs can be undertaken by a system possessing a high dimensional analog fading memory, i.e. a liquid-state machine (LSM). For example, the cortical microcolumn is hypothesized to be a LSM. A model of the gene regulation network (GRN) of E.Coli was assessed for its LSM properties for a range of increasingly complex stimuli. Cooperativity between transcription factors (TFs) is necessary for complex temporal discriminations. However, the low recurrence within the GRNs autonomous dynamics decreases its capacity for a rich fading memory, and hence for integrating temporal sequence information. We conclude that coupling of the GRN with signal transduction networks possessing cross-talk, and with metabolic networks is expected to increase the extent of non-autonomous recurrence and hence to facilitate enhanced LSM properties.

Natural Selection in the Brain

This chapter explores the possibility that natural selection takes place in the brain. We review the theoretical and experimental evidence for selectionist and competitive dynamics within the brain. We propose that in order to explain human problem-solving, selectionist mechanisms demand extension to encompass the full Darwinian dynamic that arises from introducing replication of neuronal units of selection. The algorithmic advantages of replication of representations that occur in natural selection are not immediately obvious to the neuroscientist when compared with the kind of search normally proposed by instrumental learning models, i.e. stochastic hill-climbing. Indeed, the notion of replicator dynamics in the brain remains controversial and unproven. It starts from early thoughts on the evolution of ideas, and extends to behaviourist notions of selection of state-action pairs, memetics, synaptic replicators, and hexagonal cortical replicators. Related but distinct concepts include neural selectionism, and synfire chains. Our contribution here is to introduce three possible neuronal units of selection and show how they relate to each other. First, we introduce the Adams synapse that can replicate (by quantal budding) and mutate by attaching to nearby postsynaptic neurons rather than to the current postsynaptic neuron. More generally, we show that Oja’s formulation of Hebbian learning is isomorphic to Eigen’s replicator equations, meaning that Hebbian learning can be thought of as a special case of natural selection. Second, we introduce a synaptic group replicator, a pattern of synaptic connectivity that can be copied to other neuronal groups. Third, we introduce an activity replicator that is a pattern of bistable neuronal activities that can be copied between vectors of neurons. This last type of replicator is not composed of the first two kinds, but may be dependent upon them. We suggest how these replicators may take part in diverse aspects of cognition such as causal inference, human problem solving, and memory.