Abstract Sensory systems are considered to be optimized for their ecological niche. In vision this means highly organised regular structure and function, where nearly identical photoreceptors have graded light responses in order to be able to handle as much information as possible. Instead, cockroach compound eyes show large amounts of irregularities in their optics and structure, and unusually long axons. In this thesis photoreceptors of the cockroach were studied with intracellular recordings of their light responses, biophysical systems analysis, and modelling of the relations between the light stimuli and responses. Cockroaches prefer living in dark or extremely dim environments. However, they have large and complex compound eyes. The aim of this study was to find out the functional properties by which the visual system and especially photoreceptors have adapted to cope with, i.e. to see in, dim light conditions. The function of photoreceptors was found to vary randomly in many respects, and the long axons seemed to utilise action potential coding of visual signals. Through model simulations it was shown that signals of a group of these functionally variable and spiking photoreceptors, when pooled, could provide more reliable coding than signals of identical cells of any experimentally characterised type. This naturally sacrifices spatial resolution. The filtering dynamics of the photoreceptors is matched to low light intensities and their temporal resolution does not markedly improve with increasing light adaptation. Adaptation processes in the photoreceptors saturated near an intensity of about 1000 effective photons/s. These are all both unexpected and novel features of photoreceptor function. Spatial summation of functionally different photoreceptors and reduced temporal resolution and contrast coding abilities can be considered to be permanent optimizations to a dim environment.

No description available.

Normal-hearing listeners can locate sound sources, using binaural cues for azimuth angle. These binaural differences in the timing and intensity of sound arriving at the two ears, interaural time differences (ITDs) and interaural intensity differences (IIDs), also support selective listening in multi-talker environments. Auditory-brainstem neurons of the medial superior olive (MSO) and lateral superior olive (LSO) encode ITD in the envelope of sound (ITDENV) and in the temporal fine structure of low-frequency sound (ITDTFS); LSO neurons encode IID. Bilateral-cochlear-implant (bCI) listeners generally receive only IID and ITDENV. Experimental bCI pulse-bursts overcome adaptation, and convey electrical ITDTFS. Improving the understanding of mechanisms for ITD sensitivity can help bCI developers convey acoustic ITDTFS. In this dissertation, models for auditory-brainstem neurons are developed that explain human ability to detect small differences in ITD, as neuronal and MSO population mechanisms. Promoting binaural-coincidence detection and limiting backpropagation, model MSO ion-channels set resting potentials that reproduce dendritic and somatic KLT activation, somatic Na+ inactivation, and a lower amount of axonal Na+ inactivation. Sensitivity to ITDTFS in moderately fast and very fast model MSO neurons collectively match physiological data from 150 to 2000 Hz. The best-ITD (the ITD of highest spike rate) can be made contralateral-leading, by contralateral inhibition of moderate speed, or by asymmetric axon location, leveraging dendritic filtering. Leveraging standard binaural-display models, neuronal populations based on these model MSO neurons match normal-hearing human discrimination thresholds for ITDTFS in sine tones from 39 to 1500 Hz. Adaptation before binaural interaction helps model MSO neurons glimpse the ITDTFS of sound direct from a source, before reflected sound arrives from different directions. With inputs from adapting model spherical bushy cells, a moderately fast model MSO neuron reproduces in vivo responses to amplitude-modulated binaural beats, with a frequency-dependent emphasis of rising vs. peak sound-pressure for ITDTFS encoding, which reflects human ITD detection and reverberation times in outdoor environments. Distinct populations of model LSO neurons, spanning the range of electrical membrane impedance as a function of frequency in LSO neurons, collectively reflect discrimination thresholds for ITDENV in transposed tones across carrier frequency (4-10 kHz) and modulation rate (32-800 Hz). / 2022-09-28T00:00:00Z

Bats are a paragon of evolutionary success. They rely on parsimonious sensory inputs provided by echolocation, yet are able to master lives in complex environments. The outer ears (pinnae) of bats are intricately shaped receiver baffles that encode sensory information through a diffraction process. In some bat species with particularly sophisticated biosonar systems, such as horseshoe bats (Rhinolophidae), the pinnae are characterized by static as well as dynamic geometrical features. Furthermore, bats from these species can deform their pinnae while the returning ultrasonic waves impinge on them. Hence, these dynamic pinna geometries could be a substrate for novel, dynamic sensory encoding paradigms. In this dissertation, two aspects of this dynamic sensing process were investigated: (i) Do local shape features impact the acoustic effects during dynamic deformation of the bat pinna? and (ii) do these shape deformations provide a substrate for the dynamic encoding of sensory information? For this, a family of simplified biomimetic prototypes has been designed based on obliquely truncated cones manufactured from sheets of isobutyl rubber. These prototypes were augmented with biomimetic local shape features as well as with a parsimonious deformation mechanism based on a single linear actuator. An automated setup for the acoustic characterization of the time-variant prototype shapes has been devised and used to characterize the acoustic responses of the prototypes as a function of direction. It was found that the effects of local shape features did interact with each other and with the deformation of the overall shape. The impact of the local features was larger for bent than for upright shape configurations. Although the tested devices were much simpler than actual bat pinnae, they were able to reproduce numerical beampattern predictions that have been obtained for deforming horseshoe bat pinnae in a qualitative fashion. The dynamically deformable biomimetic pinna shapes were estimated to increase the sensory encoding capacity of the device by unit[80]{%} information when compared to static baffles. To arrive at this estimate, spectral clustering was used to break up the direction- and deformation-depended device transfer function into a discrete signal alphabet. For this alphabet, we could estimate the joint signal entropy across a bending cycle as a measure for sensory coding capacity. The results presented in this thesis suggest that bat biosonar posses unique dynamic sensing abilities which have no equivalent in man-made technologies. Sensing paradigms derived from bat biosonar could hence inspire new deformable wave-diffracting structures for the advancement in sensor technology. / Ph. D.

Neurons are sensitive to specific aspects of natural stimuli, which are according to different statistical criteria an optimal representation of the natural sensory input. Since these representations are purely sensory, it is still an open question whether they are suited to generate meaningful behavior. Here we introduce an optimization scheme that applies a statistical criterion to an agent s sensory input while taking its motor behavior into account. We first introduce a general cognitive model, and second develop an optimization scheme that increases the predictability of the sensory outcome of the agent s motor actions and apply this to a navigational paradigm.In the cognitive model, place cells divide the environment into discrete states, similar to hippocampal place cells. The agents learned the sensory outcome of its action by the state-to-state transition probabilities and the extent to which these motor actions are caused by sensory-driven reflexive behavior (obstacle avoidance). Navigational decision making integrates both learned components to derive the actions that are most likely to lead to a navigational goal. Next we introduced an optimization process that modified the state distributions to increase the predictability of the sensory outcome of the agent s actions.The cognitive model successfully performs the navigational task, and the differentiation between transitions and reflexive processing increases both behavioral accuracy, as well as behavioral adaptation to changes in the environment. Further, the optimized sensory states are similar to place fields found in behaving animals. The spatial distribution of states depends on the agent s motor capabilities as well as on the environment. We proofed the generality of predictability as a coding principle by comparing it to the existing ones. Our results suggest that the agent s motor apparatus can play a profound role in the formation of place fields and thus in higher sensory representations.

Establishing and maintaining successful communication stands as a critical prerequisite for achieving the goals of inducing and studying advanced computation in small-scale living neuronal networks. The following work establishes a novel and effective method for communicating arbitrary "sensory" input information to cultures of living neurons, living neuronal networks (LNNs), consisting of approximately 20 000 rat cortical neurons plated on microelectrode arrays (MEAs) containing 60 electrodes. The sensory coding algorithm determines a set of effective codes (symbols), comprised of different spatio-temporal patterns of electrical stimulation, to which the LNN consistently produces unique responses to each individual symbol. The algorithm evaluates random sequences of candidate electrical stimulation patterns for evoked-response separability and reliability via a support vector machine (SVM)-based method, and employing the separability results as a fitness metric, a genetic algorithm subsequently constructs subsets of highly separable symbols (input patterns). Sustainable input/output (I/O) bit rates of 16-20 bits per second with a 10% symbol error rate resulted for time periods of approximately ten minutes to over ten hours. To further evaluate the resulting code sets' performance, I used the system to encode approximately ten hours of sinusoidal input into stimulation patterns that the algorithm selected and was able to recover the original signal with a normalized root-mean-square error of 20-30% using only the recorded LNN responses and trained SVM classifiers. Response variations over the course of several hours observed in the results of the sine wave I/O experiment suggest that the LNNs may retain some short-term memory of the previous input sample and undergo neuroplastic changes in the context of repeated stimulation with sensory coding patterns identified by the algorithm.