{"id":498908,"date":"2018-08-17T10:26:03","date_gmt":"2018-08-17T17:26:03","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-event&p=498908"},"modified":"2018-08-17T10:26:03","modified_gmt":"2018-08-17T17:26:03","slug":"uw-summer-institute-2005","status":"publish","type":"msr-event","link":"https:\/\/www.microsoft.com\/en-us\/research\/event\/uw-summer-institute-2005\/","title":{"rendered":"UW Summer Institute 2005"},"content":{"rendered":"

Venue:<\/strong> University of Washington, and
\nFriday Harbor Laboratories
\nSeattle, Washington<\/p>\n

Related Events:<\/strong>
\n2018<\/a>
\n2017<\/a>
\n2016<\/a>
\n2015<\/a>
\n2014<\/a>
\n2013<\/a>
\n2012<\/a>
\n2011<\/a>
\n2010<\/a>
\n2009<\/a>
\n2008<\/a>
\n2007<\/a>
\n2006<\/a>
\n1999<\/a>
\n1998<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

The 2005 meeting comes 7 years after a predecessor symposium that we organized in August 1998 under the same title. Strong positive feedback from attendees and new research directions inspired by that meeting compelled us to put together another symposium. Numerous advances have come since the last meeting, some from attendees of the meeting and their teams. The intervening years have also seen some growing interest within neurobiology of the decision making under uncertainty perspective. We are organizing the forthcoming meeting to reflect on advances, and to share new directions and ongoing challenges.<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr_startdate":"2005-06-07","msr_enddate":"2005-06-10","msr_location":"San Juan Island, Washington, USA","msr_expirationdate":"","msr_event_recording_link":"","msr_event_link":"","msr_event_link_redirect":false,"msr_event_time":"","msr_hide_region":false,"msr_private_event":false,"footnotes":""},"research-area":[13556,13553],"msr-region":[197900],"msr-event-type":[197947],"msr-video-type":[],"msr-locale":[268875],"msr-program-audience":[],"msr-post-option":[],"msr-impact-theme":[],"class_list":["post-498908","msr-event","type-msr-event","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-medical-health-genomics","msr-region-north-america","msr-event-type-universities","msr-locale-en_us"],"msr_about":"Venue:<\/strong> University of Washington, and\r\nFriday Harbor Laboratories\r\nSeattle, Washington\r\n\r\nRelated Events:<\/strong>\r\n2018<\/a>\r\n2017<\/a>\r\n2016<\/a>\r\n2015<\/a>\r\n2014<\/a>\r\n2013<\/a>\r\n2012<\/a>\r\n2011<\/a>\r\n2010<\/a>\r\n2009<\/a>\r\n2008<\/a>\r\n2007<\/a>\r\n2006<\/a>\r\n1999<\/a>\r\n1998<\/a>","tab-content":[{"id":0,"name":"About","content":"

Biological and Computational Perspectives on Intelligent Systems<\/h1>\r\nWelcome to the home page for the 2005 Symposium on Biological and Computational Perspectives on Intelligent Systems. The invitation-only symposium will focus on presentations and discussions aimed at clarifying and addressing fundamental questions about intelligent systems, through a synthesis of insights from neurobiology, computer science, decision science, and control theory. Our overall goal is to catalyze connections and interdisciplinary thinking between neurobiology and the decision and computational sciences, taking invertebrate nervous systems as a compelling focus for making progress. With the motivating challenge of understanding neurobiological systems as machinery evolved for decisions under uncertainty, we seek a sharing of ideas between leading neurobiologists and researchers in decision science, computer science, and control theory.\r\n\r\nThe 2005 meeting comes 7 years after a predecessor symposium that we organized in August 1998 under the same title. Strong positive feedback from attendees and new research directions inspired by that meeting compelled us to put together another symposium. Numerous advances have come since the last meeting, some from attendees of the meeting and their teams. The intervening years have also seen some growing interest within neurobiology of the decision making under uncertainty perspective. We are organizing the forthcoming meeting to reflect on advances, and to share new directions and ongoing challenges.\r\n\r\nThe program will begin with a reception on Tuesday evening, June 7, and will continue through Friday afternoon, June 10, 2005. Although the technical program will end on Friday afternoon, we invite attendees to stay on for a Friday afternoon whale watching cruise and dinner, and plan to depart for free time on the island or to head back on Saturday morning, following breakfast.\r\n

Organizers<\/h2>\r\nEric Horvitz<\/a>, Microsoft Research\r\nDennis Willows, Friday Harbor Laboratories\r\n

Additional Background<\/h2>\r\nThere have been meetings and special interest groups on modeling among biologists and meetings, largely among computer scientists, e.g., under the herald of computational neuroscience<\/em>. However, we found great opportunities in bringing together, in an intimate workshop setting, leaders from neurobiology and researchers working more broadly within the computational, control, and decision sciences, including scientists who call computational neuroscience their home.\r\n\r\nThe forthcoming meeting, like its predecessor in 1998, is motivated at the high level by the challenges of understanding neurobiological systems as machinery evolved for making valuable decisions under uncertainty.\u00a0 We have worked\u00a0to bring together a\u00a0set of passionate people drawn from the biological and the computational sciences\u00a0to discuss questions\u00a0about systems that sense,\u00a0learn, perform inference, and make decisions under inescapable uncertainties\u2014whether the systems are built upon a biological substrate or\u00a0are based on computational representations and algorithmic procedures.\u00a0 We hope that the meeting will stimulate real-time discussions and\u00a0insights, as well as to catalyze\u00a0longer-term\u00a0syntheses and efforts\u00a0that bring together biological and computational perspectives on shared questions.\u00a0 The program overall takes invertebrate neurobiology as a valuable focus of attention\u2014a focus aimed at better understanding invertebrate intelligence, as well as at making progress on vertebrate intelligence.\u00a0 Our intuition is that vertebrate intelligence, including the capabilities we know as human intelligence,\u00a0likely leverages key innovations implemented within \u201colder,\u201d and potentially simpler and more transparent neurobiological fabric.\r\n\r\nWhen we organized a conference under the same title seven years ago, we\u00a0were uncertain but optimistic that valuable things might come out of an attempt to weave together the\u00a0brightest minds\u00a0in neurobiology, with scholars in computer science, decision science, statistics, and control theory.\u00a0\u00a0Given the\u00a0multiple influences that the 1998 meeting had, we\u00a0have learned that our optimism was well founded.\u00a0 We hope that this meeting will have similar positive interdisciplinary influences on addressing the challenges of understanding intelligent systems.\r\n\r\nThe meeting will also be featured as a Friday Harbor Laboratories Centennial Symposium, one of several key meetings being held during a year of festivities marking the centennial anniversary of Friday Harbor Laboratories."},{"id":1,"name":"Abstracts","content":"[accordion]\r\n\r\n[panel header=\"Cell Signaling in Olfaction\"]\r\n\r\nBarry W. Ache, Whitney Laboratory for Marine Bioscience and Center for Smell and Taste, University of Florida\r\n\r\nOdorants in real-world situations are complex blends of chemicals. Odorants are recognized and discriminated by the brain using a combinatorial coding strategy in which the pattern of response across many neurons creates a unique \u2018signature\u2019 for the particular blend. Odorants inhibit as well as excite olfactory receptor neurons (ORNs) in an odorant-specific manner. This opponent input generates an integrated output from the ORNs that presumably enhances the combinatorial code by more finely matching the across-neuron pattern to the particular odorant blend. We showed that opponent input to lobster ORNs is mediated through separate intracellular signaling pathways, with phospholipid signaling mediating excitation and cyclic nucleotide signaling mediating inibition. Phospholipid signaling targets a lobster homolog of a transient receptor potential (TRP) non-selective cation channel, while cyclic nucleotide signaling targets one or more cyclic nucleotide activated ion channels capable of generating an opposing receptor current. We also showed that phospholipid signaling acts not only via the canonical phosphoinositol turnover pathway (i.e., via PLC) but also via 3-phosphoinositides (i.e., via PI3K). Odorants are well known to excite mammalian ORNs through cyclic nucleotide signaling, the target of which is the olfactory cyclic nucleotide gated non-selective cation channel. If and how odorants inhibit mammalian ORNs is less clear. Using the insight gained from studying lobsters ORNs, we showed that 3-phosphoinositides, alone and in combination with PLC-mediated signaling, decrease the sensitivity of the native olfactory cyclic nucleotide gated channel in mammalian ORNs to cAMP in an odorant-specific manner. This finding helps establish that odorants inhibit mammalian ORNs by revealing a potential phospholipid-dependent mechanism by which odorants could modulate the output of the cells. Collectively, these findings suggest that opponent input may be a general mechanism to enhance combinatorial based odorant coding, and underscore our ability to learn from the use of marine animal models in biomedicine.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Random Graph Theory Meets Biological Networks\"]\r\n\r\nDimitris Achlioptas, Microsoft Research\r\n\r\nA number of biological processes can be abstracted as networks (graphs). Autocatalytic networks, gene regulatory networks, and models of memory are some of the most well-known examples. The premise that such networks are the result of evolutionary processes, i.e., repeated random experiments with feedback, makes them the most natural and abundantly available examples of \"random\" networks. At the same time, the mathematical study of random networks has been an active field since the 1960s and by now a few central principles have emerged (while much more remains open). In this talk, I will review some connections between neurobiological networks and mathematical random networks hoping to elucidate how some of the mathematical principles might \"make sense\" in the context of neurobiology.\r\n\r\n \r\n\r\n[\/panel]\r\n\r\n[panel header=\"Biological Computing with Biological Parts\"]\r\n\r\nHenry Abarbanel, Department of Physics, University of California, San Diego\r\n\r\nSensory systems in animals encode analog environmental information in sequences of spikes. Many neural circuits which receive these spike trains are very sensitive to specific sequences which identify friend and foe or members of similar species. How can biological circuits which are made out of biological \"parts\" (neurons and synapses following biophysical rules) learn sequences of spikes and recognize them robustly and with the required sensitivity? We construct such a circuit and show how it can learn specific sequences of spikes.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Voltage-Sensitive Dye Imaging During Decision-Making in the Medicinal Leech\"]\r\n\r\nKevin Briggman, Computational Neurobiology Program, University of California, San Diego\r\n\r\nThe identification of neurons involved in behavioral choices is a first step towards understanding the neuronal mechanisms of decision-making. We use the nervous system of the medicinal leech as a model system to study decision-making. The isolated leech nervous system randomly chooses to swim or to crawl in response to the same stimulus. Is this choice made by individual command neurons or by a population of neurons interacting as a network?\r\n\r\nWe use FRET-based voltage-sensitive dyes to record the activity of populations of individual neurons during the choice. Our analysis of these high-dimensional datasets has shown that the activity of a network of neurons discriminates between the two behaviors earlier than any single neuron. By hyperpolarizing or depolarizing a specific neuron in this network, we are able to bias the choice towards swim or crawling, respectively. The ability to influence a choice by manipulating an individual neuron will allow us to explore the mechanisms of this form of decision-making.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Phase Changes in Locusts\"]\r\n\r\nMalcolm Burrows, Department of Zoology, University of Cambridge\r\n\r\nLocusts can change between two extreme forms called phases, which show striking differences in morphology, physiology and behaviour. These two extreme forms are the gregarious, swarming phase involving vast numbers of insects, and the solitarious phase in which individuals live alone and actively avoid each other. We can therefore investigate neuronal plasticity underlying behavioural differences in relatively simple neuronal networks.\r\n\r\nOur work has focused on four main areas. First, we are identifying the proximate stimuli that produce an initial rapid behavioural gregarization of solitarious locusts. Repeated tactile and proprioceptive stimulation of the hind legs can induce a behavioural phase change in 4 hours, and this gregarizing effect can be mimicked by electrical stimulation of a sensory nerve. Second, phase change is accompanied by significant changes in the levels of at least 11 different neurotransmitters and neuromodulators in the central nervous system that occur over different time scales. We are now investigating the possible causal effects of these substances in producing behavioural differences between phases. Third, we are analysing phase-related differences in the response properties of an identified visual interneurone, which has an important role in detecting approaching objects. We are currently analysing the effects of its changing outputs onto the motor systems. Fourth, we are analysing differences in walking and jumping behaviour and relating these to changes in the properties of the musculo-skeletal system and the networks of interneurones and motor neurones that control it.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Challenges of Understanding how Biological Transducers Make Sense of the World\"]\r\n\r\nShaun Cain, Friday Harbor Laboratories, University of Washington\r\n\r\nThere is something unique about the way that ciliated cells are organized that makes them ideally suited to serve as platforms for transducing almost any energy source into a neural signal. We know this because virtually every sensory transducer detecting any signal, in nearly every animal on the planet derives from ciliated cell precursors in the embryo of that animal. What's more, such living cells can apparently evolve to capture incredibly small electromagnetic and mechanical energy sources, and convert them into electrical signals that are meaningful in the lives of their owners. We will present behavioral, physiological, and cellular evidence for the way that many animals might detect and orient to one of the weakest, yet most pervasive, sensory sources available on earth, viz., the geomagnetic field. This environmental sensory signal, unlike all others, readily penetrates tissues and is present un-degraded everywhere in the animal. The energy inherent in the field is low; being present at a level where thermal noise is a confounding variable. We will show the behavioral and physiological evidence we have found for the existence of the geomagnetic sense in a sea slug. It is clear that they detect the Earth's field, orient to it on the ocean bottom, and have receptors distributed perhaps widely in the foot upon which they glide. Further, they have cells that contain what appear to be single domain magnets, electron dense spindles of iron oxide. We will discuss ways these might be connected to collect and make sense of directional magnetic information, and suggest ways this information is useful to an animal that must make appropriate decisions about directions in which to move, under circumstances of great uncertainty. (With Dennis Willows.)\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Synaptic Augmentation Contributes to the Temporal Sensitivity of Environmental Regulation in the Aplysia Siphon-Withdrawal Reflect\"]\r\n\r\nBob Calin-Jageman, Department of Biology, Georgia State University\r\n\r\nTemporal discriminations are critically important to the adaptive regulation of behavior. Here we investigate the synaptic and network mechanisms that contribute to a simple temporal discrimination in the Aplysia siphon withdrawal reflex (SWR). The duration of the siphon withdrawal reflex (SWR) is reduced during exposure to turbulence, an environmental stimulus. Recovery after turbulence is sensitive to exposure duration. Recovery takes more than 1 min following brief (10s -5 min) turbulence but less than 20s following long (10 min) turbulence.\r\n\r\nWe have proposed that the temporal sensitivity of SWR recovery is due, in part, to augmentation (AUG), an activity-dependent form of short-term synaptic plasticity expressed at the inhibitory synapses of L30 interneurons. To test this hypothesis, we measured the effects of turbulence on L30 activity and L30 plasticity in semi-intact preparations. We found that (1) turbulence produces L30 activity, leading to the induction of AUG; (2) that L30 activity and AUG decay over the course of a long exposure to turbulence, so that post-turbulence expression of AUG occurs only after brief turbulence, and (3) this pattern of L30 AUG directly contributes to the ability of the SWR circuit to discriminate between brief and long turbulence. Our results indicate that AUG, and other forms of short-term plasticity, may be of critical importance for enabling simple neural networks to processes temporal information during adaptive behavioral regulation. (With Thomas Fischer.)\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Neural Network Smarts: Lessons from the Hermissenda Eye\"]\r\n\r\nGreg Clark, Department of Bioengineering, University of Utah\r\n\r\nWhat makes neural networks \u201csmarter\u201d than single neurons? Further, why do biological neural networks often outperform their human-engineered counterparts? One important emergent property exhibited by biological neural networks is the use of contextual spike-timing relationships\u2014the timing of action potentials in one neuron, relative to the timing of action potentials in other neurons. Here we show that such contextual relationships can strengthen information flow within a system, and can help explain how biological nervous systems perform so splendidly in the presence of \u201cnoise\u201d, despite being inherently noisy themselves.\r\n\r\nWe have used the eye of the marine nudibranch mollusk Hermissenda crassicornis as a simple model system. Our physiological studies, in conjunction with biologically realistic, Hodgkin-Huxley level computational simulations of the Hermissenda eye, indicate two key findings. First, in both the biological eye and the fully connected simulated network, feedback inhibition from type-A photoreceptors onto type-B photoreceptors produces a striking absence of B-cell spikes shortly after A-cell spikes. Consequently, B cells fire later in the A-cell interspike interval, when the inhibitory B-to-A input is more potent in suppressing the next A-cell spike (Fost and Clark, 1996). Hence, by altering contextual spike timing, feedback inhibition of type-B cells in turn yields greater inhibition of the type-A cell output, thereby making the network with feedback connections \u201csmarter\u201d.\r\n\r\nSecond, in contrast with the traditional view that noise degrades system performance, we have found that random ionic noise, synaptic noise, and spike-timing noise improve, rather than impair, the ability of both simulated and real Hermissenda eyes to encode light intensity. In simulations, noise-induced improvements in light intensity encoding occurred across 8 light levels, were not confined to perithreshold stimulus intensities, and did not arise from simple stochastic resonance or DC-bias threshold effects (Butson & Clark, 2002; Clark & Butson, 2004). Noise-free conditions produced \u201czones of stability\u201d (Perkel et al., 1964), characterized by phase locking, in which an increase in the frequency of inhibitory postsynaptic potentials (IPSPs) could paradoxically increase, rather than decrease, the firing rate of postsynaptic photoreceptors, thus producing a non-monotonic relationship between light intensity and photoreceptor firing rate. In contrast, the addition of noise (random variations in ionic currents and IPSP amplitudes) disrupted the emergence of phase locking, yielding a more systematic relationship and improved light-intensity encoding.\r\n\r\nInitial physiological experiments support the major conclusions of these simulations. Noise-free conditions again produced phase locking within zones of stability. Consequently, increases in the frequency of artificial IPSPs (intracellular hyperpolarizing current injections) could increase as well as decrease the postsynaptic photoreceptor firing rate evoked by artificial light (depolarizing current) steps, thereby producing non-monotonic relationships that disrupted light-intensity encoding. Such anomalous effects were greatly reduced by the introduction of timing noise (IPSPs delivered at pseudo-random intervals), so photoreceptor firing rate was more closely related to light intensity.\r\n\r\nThese results indicate the importance of contextual spike timing relationships in the formation of emergent network properties, and help explain how biological neural networks, unlike most human-engineered devices, excel at accurate encoding of signals in noisy environments. They further suggest principles that may be advantageously incorporated into artificial intelligent systems, robotics, or neuroprostheses.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Inverse Problems in Motor Control: The Challenge of Multiple Time Scales\"]\r\n\r\nThomas Daniel, Department of Biology, University of Washington\r\n\r\nInsect flight is controlled by a variety of sensory modalities\u2014visual, chemical and mechanical\u2014each is processed with different temporal characteristics, and each directs an array of mechanical controls.\u00a0\u00a0 We have been looking at inverse solutions to the control of flight using genetic algorithms.\u00a0\u00a0 We ask what patterns of activation can give rise to a specified spatial trajectory and flight speed. We are particularly interested in the range of possible activation paradigms that yield a given behavior.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Novel and Adaptive Strategies in Chemical Defense and Chemical Signaling in Sea Hares\"]\r\n\r\nChuck Derby, Department of Biology, Georgia State University\r\n\r\nAplysia and other sea hares have a rich diversity of chemicals that mediate many important behaviors. They release ink when attacked by predators, and this ink contains antipredatory chemicals, alarm cues that warn conspecifics about attacks on neighboring sea hares, and antimicrobial and cytolytic compounds. We have been exploring mechanisms of action and molecular identity of these cues in the ink of Aplysia californica.\r\n\r\nWe have found that some antipredatory chemicals in ink operate via traditional mechanisms - unpalatable and aversive chemicals \u2013 but others function through novel mechanisms. For example, ink contains huge quantities of chemicals also found in the food of their predators, and these chemicals stimulate receptors in the spiny lobster's chemosensory pathway. Ink stimulates a variety of behaviors, including appetitive, feeding, grooming, and avoidance. The appetitive and feeding behaviors are caused by the chemicals mimicking food, which distract the lobsters and thus allow the sea slugs to escape: we call this novel chemical defense mechanism \u2018phagomimicry\u2019. These chemical stimulants, together with its sticky consistency, might also cause temporary disruption of the lobster\u2019s sensory systems (\u2018sensory disruption\u2019), adding to the sea slug\u2019s protection.\r\n\r\nCues in ink evoking alarm responses in conspecifics are different than those mediating antipredation. They include at least 7 active compounds, at least one of which is a nucleoside \u2013 uridine.\r\n\r\nAntimicrobial effects are mediated by a protein in the ink \u2013 an L-amino acid oxidases. In Aplysia californica, we call this molecule is called \u2018escapin\u2019. Escapin\u2019s antimicrobial activity is by two mechanisms \u2013 inhibition of growth, and killing. These effects are generated by a diversity of mechanisms that include but are not limited to the enzymatic activity of escapin. Escapin uses lysine and arginine as substrates to produce hydrogen peroxide, alpha-keto-acids, and carboxylic acids, and these products participate in different ways toward the antimicrobial effects. In addition, escapin itself, independent of its enzymatic activity, can mediate some antimicrobial effects.\r\n\r\nIn summary, the ink of Aplysia is a complex chemical m\u00e9lange that mediates a diversity of adaptive effects, including traditional and novel defenses against predators, alarm signaling to neighboring conspecifics, and preventing microbial infections that might result from predatory attacks.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Robustness and Biological Complexity: What I've learned about Biology in the Last 7 Years\"]\r\n\r\nJohn Doyle, Control and Dynamical Systems, Electrical Engineering, and Bioengineering, California Institute of Technology\r\n\r\nA surprisingly consistent view on the fundamental nature of complex systems can now be drawn from the convergence of three distinct research themes. First, molecular biology has provided a detailed description of much of the components of biological networks, and the organizational principles of these networks are becoming increasingly apparent. It is now clear that much of the complexity in biology is driven by its regulatory networks, however poorly understood the details remain. Second, advanced technology is creating engineering examples of networks where we do know all the details and that have complexity approaching that of biology. While the components are entirely different, there is striking convergence at the network level of the architecture and the role of protocols, layering, control, and feedback in structuring complex system modularity. Finally, there is a new mathematical framework for the study of complex networks that suggests that this apparent network-level evolutionary convergence both within biology and between biology and technology is not accidental, and follows necessarily from the requirements that both biology and technology be efficient, robust, adaptive, and evolvable. This talk will describe qualitatively in as much detail as time allows these features of biological systems and their parallels in technology, using hopefully familiar and concrete examples. The aim is to be accessible to biologists, and not to depend critically on the mathematical framework. A crucial insight is that both evolution and natural selection or engineering design must produce high robustness to uncertain environments and components in order for systems to persist. Yet this allows and even facilitates severe fragility to novel perturbations, particularly those that exploit the very mechanisms providing robustness, and this \"robust yet fragile'\" (RYF) feature must be exploited explicitly in any theory that hopes to scale to large systems. If time permits, we will briefly discuss how this view of \"organized complexity\" might influence neurosciences. It als contrasts sharply with the view of \"emergent complexity\" that is favored among researchers who draw their inspiration from models and concepts popular in physics, such as lattices, cellular automata, spin glasses, phase transitions, criticality, chaos, fractals, scale-free networks, self-organization, and so on.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Adaptation and Information Processing\"]\r\n\r\nAdrienne Fairhall, Department of Physiology and Biophysics, University of Washington\r\n\r\nThe idea that neurons encode information efficiently has been around since the 50\u2019s but until recently, there was little quantitative evidence to support this view. A specific prediction of the efficient coding hypothesis is that the statistics of neural outputs should be matched to the statistics of their inputs. Influential early work showed that the input\/output properties of some fly visual neurons appeared to match the contrast statistics of natural scenes. However, since many natural stimuli have statistical properties that can vary wildly in time and space, one might imagine that a better strategy is to continually adapt to these changing local statistics. We showed that for a motion-coding identified neuron in the housefly, Calliphora vicina, the neuron\u2019s coding properties do indeed constantly adapt to match local statistics, and do so in a way that optimizes information transmission through the system. While estimating the time required to perform this adaptation, we found that different adaptive processes were happening on different timescales. While information maximization occurs on the order of tens to hundreds of milliseconds, the firing rate shows slower dynamics on timescales of up to tens of seconds. These slow dynamics show interesting power-law-like properties. We discuss the mechanisms, generality and functional role of these different adaptive processes.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Data structures and decision making in computational olfaction\"]\r\n\r\nAlan Gelperin, Monell Chemical Senses Center\r\n\r\nThe olfactory system is built to make rapid decisions about stimulus identity and recall associations based on prior learning about olfactory stimuli. Recent advances in understanding the molecular genetics of wiring the olfactory system have clarified how information may be processed in early olfaction. Fundamental questions remain as to how odor stimuli are represented in the CNS and how associations between central odor representations and neural events previously associated with odor stimuli are linked. Recent network models of olfactory processing based on biological hardware and software are essential adjuncts to traditional reductionist approaches to dissecting mammalian olfaction experimentally. Robotic olfaction provides a testing ground for both algorithmic and neurobiological ideas for using olfactory information to guide actions of a mobile agent.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Building Animals: Cost-benefit Decision-Making in Simple Neural Networks\"]\r\n\r\nRhanor Gillette, Department of Molecular & Integrative Physiology, University of Illinois, Champagne-Urbana\r\n\r\nDecisions are the integration of sensation, internal state and experience by goal-directed neural networks. These processes are exemplified in the behavior of predators such as the sea-slug Pleurobranchaea, the octopus and sea-anemones. A simple general neural network model for decision derived from behavioral and neurophysiological studies of Pleurobranchaea is elaborated in a simulation of optimal foraging and made generally available.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Modularity and Primitives in Tetrapod Spinal Motor Systems\"]\r\n\r\nSimon Giszter, Drexel University College of Medicine\r\n\r\nThe degrees of freedom problem is often considered a serious issue in biological motor control. Both vertebrates and invertebrates face this issue. It was perhaps first raised as a concern by Bernstein. Notions of spinal modularity date back to Sherrington and Brown and are one set of solutions to the degrees of freedom problem. Modularity, by coalescing degrees of freedom into useful simply controlled assemblies, may circumvent the problem. Pattern generators are one type of modularity found in the motor control of invertebrates and vetebrates. Our work in frogs and rats supports a further modular organization in spinal cord. Our data suggest that the spinal cord, in addition to an intrinsic capacity for organizing rhythmic timing of patterns, or central pattern generators, also possesses a modular \u2018motor basis set\u2019 or collection of \u2018primitives\u2019. These \u2018primitive\u2019 circuits organize modular premotor drives and patterns of feedback, coalescing degrees of freedom. The modular circuitry organizes multi-joint force-field patterns or force-field primitives at the biomechanical level. These can provide a useful set of basic motions and interactions with the environment. These force-field patterns can be combined in the limb. This combination provides a parsimonious method of adjusting, and correcting movement or constructing novel movements. We speculate that this functional modularity is intrinsic to the pattern shaping circuitry which may operate downstream from rhythm generation. These modules are used in motor pattern construction and pattern shaping and to \u2018bootstrap\u2019 motor behaviors and motor learning. The modules are likely to capture the statistics of the sensorimotor relationship and the lower dimensional descriptions possible for reflex and locomotor movements. We have demonstrated these elements in frogs in various ways, by biomechanical, physiological and signal decomposition techniques, and we have examined the competence of this organization to account for spinal motion repertoires. We are presently beginning to examine the neural underpinnings to assess to what extent a dedicated circuitry for primitives exists and to what extent the modularity is emergent from a broadly distributed representation. We are also beginning to consider how a modularity in mammalian spinal cords might impact on approaches to neural repair and therapy after spinal cord injury. Supported by NIH NS40412 and NS24707.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Uncertainty, Utility, and Architecture: A Decision-Theoretic Perspective on Intelligence\"]\r\n\r\nEric Horvitz, Microsoft Research\r\n\r\nNervous systems have been shaped by selective pressures to perform valuable sensory fusion, learning, inference, and action under the uncertainties associated with environmental niches.\u00a0 I will highlight key challenges and opportunities for research in neurobiology, by reviewing core concepts developed in attempts to engineer systems that sense, reason, and make decisions under uncertainty.\u00a0 I will touch on several challenges, including sensing, reasoning, and action under limited resources, ideal compilation of action into situation-action policies, continual computation, learning classifiers for urgency, and probabilistic models that predict surprise and anomaly.\u00a0 Finally, I will discuss questions and research directions highlighted by the concept of bounded optimality, in pursuit of optimizing the value of actions, conditioned on constraints of architecture and available sensing and reasoning machinery.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Learning an \u201cEpitome\u201d of Natural Signals\"]\r\n\r\nNebojsa Jojic, Microsoft Research\r\n\r\nI will present a novel representation of natural signals, which I dubbed an \u2018epitome\u2019 for its summarization quality. An \u2018epitome\u2019 of an ordered dataset, or a signal, is constructed from many overlapping data patches of variable size. The epitome\u2019s total size is limited and is typically much smaller than the original dataset. Such condensation of information is possible by exploiting both the overlaps and the repetitiveness of data fragments. By construction, epitome coordinates can be mapped to the data fragments and vice versa and the epitome can thus serve as an organization of visual memory in computer vision applications, or an organization of cellular memory in vaccine design. In the first case, epitome is built from overlapping image patches, wile in the other case it is built from overlapping short peptides taken from various strains of the viral protein chosen to be \u2018epitomized.\u2019 In addition to these applications, epitome has been used for video analysis, audio analysis and the analysis of motion capture data. As a simple way of achieving both fragment alignment and fragment size invariance, a representation similar to epitome may actually be used by the nervous system, as well.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Engineering Self-Organizing Systems, Using Inspiration from Developmental Biology\"]\r\n\r\nRadhika Nagpal, Computer Science Division, Engineering and Applied Sciences, Harvard University\r\n\r\nDuring embryogenesis, cells with identical DNA cooperate to form complex structures, such as ourselves, with incredible precision in the face of unreliable cells, variations in cell numbers, and changes in the environment. Emerging technologies have made it possible to build novel applications, from programmable materials and sensor networks, to self-reconfiguring modular robots. Acheiving similar complexity and reliability poses two challenges: (a) How do we achieve robust collective behavior from large numbers of unreliable agents? (b) How do we translate global goals into local interactions?\r\n\r\nIn this talk I will present some examples of an approach that combines local organization primitives inspired by studies of embryogenesis and developmental biology, with programming language techniques for managing complexity. This work demonstrates that not only is it possible to achieve global-to-local compilation but that one can also encode biologically-motivated principles such as scale-invariance and self-repair into the compilation process itself. The domain is mainly pattern formation and shape assembly\u2014one of the open questions is whether we can extend these ideas to other types of global goals.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Algorithms for Inverse Reinforcement Learning\"]\r\n\r\nAndrew Y. Ng, Computer Science Department, Stanford University\r\n\r\nIn the inverse reinforcement learning (IRL) problem, we observe the behavior of some agent, and wish to infer the reward function (or cost function) that the agent is trying to optimize. This problem is motivated by two applications: The first is in the use of reinforcement learning and related methods as computational models for animal and human learning. Here, IRL can be used to directly estimate the reward function that best explains observed behavior. The second motivation arises from the apprenticeship learning (also called imitation learning) setting, in which we have to learn to perform a task demonstrated by a teacher. In our approach, we apply IRL to infer the teacher's reward function, and then use the estimated reward function to learn to perform the same task ourselves. We prove that our algorithm learns to perform the task nearly as well as the teacher. We also illustrate the algorithm's behavior on a simulated car driving task.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Evolution of Brain and Behavior: Insights from Comparative Phylogenetic Studies\"]\r\n\r\nKiisa Nishikawa, Department of Biological Sciences, Northern Arizona University\r\n\r\nCladistic analyses have revolutionized our understanding of brain evolution by demonstrating that many neural structures have evolved numerous times independently. Examples can be found for nearly all sensory systems, major divisions of the brain, and animal phyla. Relatively few neuroethological studies have investigated the evolution of brain and behavior within an evolutionary framework. Three relatively well studied examples will be reviewed: electric communication in gymnotiform and mormyriform fishes, prey capture in frogs, and sound localization in owls. These three cases reveal similar patterns of brain evolution. First, novel abilities have appeared many times independently in species whose common ancestors lack these abilities. Second, relatively minor changes in neural pathways have led to dramatic changes in an organism's behavior. These evolutionary patterns imply that similar abilities may be conferred by convergent rather than homologous circuits, even among closely related species. Closely related species may use the same information in different ways, or they may use different means to obtain the same information. When novel sense organs evolve, little modification of existing neural circuits may be required for processing the new data. The evolutionary appearance of novel functions is associated with constraints, for example in the algorithms used to perform a given neural computation. Thus, convergence in functional organization may reveal basic design features of neural circuits in species that, despite their unique evolutionary histories, use similar algorithms to solve basic computational problems.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"How do Biological Cells Build Complex Structures and Move Purposefully Without Any Foreman Making Decisions or Giving Instructions?\"]\r\n\r\nGarry Odell, Friday Harbor Laboratories, University of Washington\r\n\r\nHow do biological cells accomplish complex purposeful motions and cell divisions that collectively build the multicellular body of an embryo starting with a single cell? What mechanisms motivate cells and cause them to act alive? At the NIH-funded Center for Cell Dynamics at Friday Harbor Laboratory, we are trying to figure out how the lifelike behavior of cells may emerge spontaneously from simple, haphazard, mechanochemical interactions among the myriad protein parts that genes encode. I will show movies, made through our computer-controlled microscopes, showing various parts interacting to build complex cytoskeletal structures that animate living cells.\r\n\r\nWe visualize the molecular parts using a technique born at Friday Harbor Laboratory. The Aquoria jelly fish, that swarm off our docks, evolved a gene that generates a fluorescent protein. We splice that jelly fish gene into other genes in our study organisms that code for the proteins we want to see. Each protein synthesized from such an engineered gene glows visibly under our microscopes. From the resulting movies of embryos developing with their parts aglow, we try to deduce how those parts interact. Then we write computer programs that represent tens of thousands of such parts, interacting as they bump into each other accidentally while diffusing inside cells, to find out whether complex lifelike action emerges spontaneously from those myriad simple interactions.\r\n\r\nI will show computer-animated movies demonstrating that it does. The mystery we are trying to unravel is how an army of numerous unintelligent parts, interacting chaotically, with no leader nor any 'plans' or blueprints available to specify what is supposed to happen, nevertheless accomplishes reliably life's essential tasks. Biological cells have evolved a scheme to replace the kind of centralized decision making processes that humans favor by the process sketched above. What most surprises us is that the self-organizing chaotic scheme that natural selection came up with is so astonishingly roust that it continues to yield the right outcomes in the face of extreme perturbations our experiments inflict. Cells seem untroubled by complete uncertainty about the conditions under which they must perform complicated jobs correctly else perish.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Vision as a Compensatory Mechanism for Disturbance Rejection in Upwind Flight\"]\r\n\r\nMichael Reiser, Computation and Neural Systems program, California Institute of Technology\r\n\r\nRecent experimental results demonstrate that flies posses a robust tendency to orient towards the frontally centered focus of the visual motion field that typically occurs during upwind flight. In this talk, I will present a closed-loop flight model, with a control algorithm based on feedback of the location of the visual focus of contraction, which is affected by changes in wind direction. The feasibility of visually guided upwind orientation is demonstrated with a model derived from current understanding of the biomechanics and sensorimotor computation of insects. The matched filter approach used to model the visual system computations compares extremely well with open-loop experimental data.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Postsynaptic Mechanisms of Synaptic Facilitation and Behavioral Sensitization in Aplysia\"]\r\n\r\nAdam Roberts, Physiological Science Program, UCLA\r\n\r\nUntil recently, nonassociative forms of learning in the marine snail Aplysia californica have been ascribed to simple, presynaptic cellular mechanisms. In particular, dishabituation and sensitization in this invertebrate organism have been ascribed to presynaptic facilitation. But recent evidence from our laboratory indicates learning in Aplysia depends critically upon postsynaptic mechanisms.\r\n\r\nWe have found that serotonin (5-HT), the endogenous monoamine that mediates dishabituation and sensitization of the siphon-withdrawal reflex in Aplysia, causes upregulation of AMPA receptor function in siphon motor neurons. This functional upregulation of AMPA receptors depends upon release of calcium from postsynaptic intracellular stores and postsynaptic exocytosis. We hypothesize that stimuli that induce dishabituation and sensitization in Aplysia modulate AMPA receptor trafficking in motor neurons that mediate that withdrawal reflex. Support for this hypothesis comes from experiments in which prior injection of botulinum toxin, an inhibitor of exocytosis, into identified siphon motor neurons blocks behavioral dishabituation of the siphon withdrawal reflex.\r\n\r\nWe propose that presynaptic facilitation is a short-lived form of plasticity that mediates only early stages of dishabituation\/sensitization in Aplysia. Persistent learning is mediated by postsynaptic mechanisms.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Mechanisms of Memory in the Nematode Caenorhabditis elegans\"]\r\n\r\nJacqueline Rose, Department of Psychology and Brain Research Centre, University of British Columbia\r\n\r\nTo uncover biological mechanisms of memory we study the most basic form of learning (habituation) in one of the simplest invertebrate model systems, Caenorhabditis elegans. C. elegans is a soil-dwelling microscopic nematode with a nervous system comprised of 302 neurons. Under laboratory conditions, C. elegans exist in colonies on agar-filled Petri plates streaked with the food source E. coli. When the side of the Petri plate is tapped, this elicits the tap-withdrawal response (Rankin, Beck and Chiba, 1990). Worms respond 90% of the time with a reversal response (swimming backwards). The neural circuit of this mechanosensory response has been mapped and the connections between the specific neurons are known thus allowing us to predict possible locations for changes due to memory (Wicks and Rankin, 1995). As well, knowing the neurons involved in the tap-withdrawal response allows us to search for genes expressed in the circuit to examine their role in memory. Habituation of the tap-withdrawal response in C. elegans is described as responding with smaller reversals following repeated mechanosensory (tap) stimulations; a decrease not attributable to sensory adaptation or fatigue. Memory for habituation can be examined by training worms with repeated taps and measuring worms\u2019 responses to tap after some delay as a test of retention. This paradigm has resulted in the uncovering of several properties of memory. For instance, similar to memory in other organisms long-term memory for habituation (><\/u>24 hours) relies on; a distributed training protocol (training divided into blocks separated by rest periods), de novo protein synthesis and intact transmission of the glutamate neurotransmitter (Rose et al 2003). As well, long-term memory is correlated with a decrease in expression in a type of glutamate receptor. We have also tested memory at shorter delays (i.e., 12 hours) and found different properties of memory. Finally, we have also examined the permanence of consolidated memory by examining long-term memory to determine whether it can be altered by experience. Our results show that memory processes in C. elegans are very similar to memory processes in other, more complex organisms. This leads to the hypothesis that memory is critical for the survival of multicellular organisms, and that the mechanisms of memory are highly conserved across evolution. (With Catharine H. Rankin.)\r\n\r\n[\/panel]\r\n\r\n[panel header=\"How the Brain Decides\"]\r\n\r\nMike Shadlen, Department of Physiology and Biophysics, University of Washington\r\n\r\nNeurobiology is beginning to furnish an understanding of the brain mechanisms that give rise to such higher cognitive functions as planning, remembering, and deciding. Progress has come about mainly by measuring the electrical activity from parts of the brain that lie between the sensory and motor areas. The neurons in these association areas operate on a time scale that is not controlled by external events: their electrical activity can outlast sensory input for many seconds, and they do not cause any overt change in behavior. Put simply, these neurons play neither a purely sensory nor a purely motor role but appear instead to give rise to mental states. My lecture will focus on neurons in the parietal lobe that underlie a simple kind of decision-making\u2014forming a commitment to one of two competing hypotheses about a visual scene. We have discovered that these neurons work by accumulating \u201cevidence\u201d from the sensory cortex as a function of time. The brain makes a decision when the accumulated evidence represented by the electrical discharge from these neurons reaches a criterion level. These neurons therefore explain both what is decided and when a decision is reached. Interestingly, the neural computations that underlie such a decision process were anticipated during WWII by Alan Turing and Abraham Wald. Turing applied this tool to break the German Navy\u2019s enigma cipher, while Wald invented the field of sequential analysis. In addition to mathematical elegance and winning wars, our experiments suggest that this computational strategy may lie at the core of higher brain function.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"On Building Agents that Maximize Long-Term Reward\"]\r\n\r\nSatinder Singh, Computer Science & Engineering, University of Michigan\r\n\r\nOver the last decade and more, there has been rapid progress in various subfields of artificial intelligence on building agents that maximize long-term reward. I will review this substantial progress that has come about by exploiting the well-established formalism of Markov decision processes (MDPs). At the core of the MDP-formalism are particular formulations of the elemental notions of state, action and reward. I will describe recent progress on rethinking these basic elements and argue that these new results together point to a new formalism and representation simultaneously more suited to biological modeling as well as to engineering advances.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Adaptive Sensory Plasticity for the Encoding of Communication Signals\"]\r\n\r\nJoseph Sisneros, Department of Psychology, University of Washington, Seattle\r\n\r\nMany seasonally breeding vertebrates are faced with the same fundamental problem of locating a mate during the reproductive season. Sensory receiver systems used for the detection and localization of conspecific mates were previously thought to be fixed throughout the adult life history of an animal. Two recently demonstrated examples of adaptive sensory plasticity for the detection and localization of mates by fishes will be presented. Both the electrosensory system of the Atlantic stingray (Dasyatis sabina) and the auditory system of the plainfin midshipman fish (Porichthys notatus) have undergone evolutionary adaptations for the enhancement of encoding social and reproductive communication signals by means of a common steroid-dependent mechanism. Implications of this steroid-dependent mechanism for information processing and its adaptive coupling of sender and receiver will be discussed.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Visual detection of motion in flies: Invariance in coding of velocity of naturalistic scenes\"]\r\n\r\nAndrew Straw, Department of Bioengineering, California Institute of Technology\r\n\r\nFrom a low contrast foggy morning to a high contrast dark forest pierced by shafts of light, the wide range of contrasts under which animals or machines must operate challenges their ability to see. When viewing simple sine-wave grating patterns, perception of velocity is confounded with luminance contrast of the pattern. At a neural level, this contrast-dependence is present in motion-sensitive interneurons in flies and mammalian visual cortex. To investigate the significance of this ambiguity under naturalistic conditions, we recorded intracellularly from such cells in the hoverfly while displaying moving photographic images of natural environments. Contrary to results obtained with grating patterns, we show that fly motion detectors encode the velocity of natural images independently of the particular image used, despite a threefold range of contrast. We suggest that the fly visual system may be matched to natural scenes, enabling accurate estimates of velocity largely independent of the particular scene.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Desert Ant Navigation: Mini Brains\u2014Mega Tasks\u2014Smart Solutions\"]\r\n\r\nRudiger Wehner, Institute of Zoology, University of Zuerich, Switzerland\r\n\r\nAnts of the Sahara desert, Cataglyphis by name, are skillful navigators. While foraging and homing over distances of several thousand times their body lengths, they accomplish truly formidable tasks. They use a pattern in the sky that is invisible to man to steer their compass courses, and then they integrate all angles steered and all distances covered by remarkable acumen. This system of path integration works even in completely featureless terrain. In addition, Cataglyphis can use landmarks by employing photographic skyline memories. Finally, they rely on search strategies that are much more efficient than a random walk would let one assume.\r\n\r\nThe talk focuses on the behavioural performances as well as on the sensory and neural mechanisms that are involved in mediating this behaviour. How can a 0.1-mg brain equipped with a panoramic compound-eye system accomplish these awe-inspiring modes of behaviour? The presentation will focus on the general sensory stratagems employed by Cataglyphis, and will show that this small-brain navigator uses simpler tricks than meets the human designer\u2019s eye. Cataglyphoid robots are used to test the hypotheses derived from neurophysiological analyses.\r\n\r\nThe general message is that a high-level task can be solved by the co-operation of a number of low-level systems. These low-level systems are adapted to the particular ecological niche, within which the desert navigator operates.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Challenges of Understanding how Biological Transducers Make Sense of the World\"]\r\n\r\nDennis Willows, Friday Harbor Laboratories, University of Washington\r\n\r\nThere is something unique about the way that ciliated cells are organized that makes them ideally suited to serve as platforms for transducing almost any energy source into a neural signal. We know this because virtually every sensory transducer detecting any signal, in nearly every animal on the planet derives from ciliated cell precursors in the embryo of that animal. What's more, such living cells can apparently evolve to capture incredibly small electromagnetic and mechanical energy sources, and convert them into electrical signals that are meaningful in the lives of their owners. We will present behavioral, physiological, and cellular evidence for the way that many animals might detect and orient to one of the weakest, yet most pervasive, sensory sources available on earth, viz., the geomagnetic field. This environmental sensory signal, unlike all others, readily penetrates tissues and is present un-degraded everywhere in the animal. The energy inherent in the field is low; being present at a level where thermal noise is a confounding variable. We will show the behavioral and physiological evidence we have found for the existence of the geomagnetic sense in a sea slug. It is clear that they detect the Earth's field, orient to it on the ocean bottom, and have receptors distributed perhaps widely in the foot upon which they glide. Further, they have cells that contain what appear to be single domain magnets, electron dense spindles of iron oxide. We will discuss ways these might be connected to collect and make sense of directional magnetic information, and suggest ways this information is useful to an animal that must make appropriate decisions about directions in which to move, under circumstances of great uncertainty. (With Shaun Cain.)\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Intracellular recordings with implantable silicon microelectrodes?\"]\r\n\r\nRussell Wyeth, Friday Harbor Laboratories, University of Washington\r\n\r\nBehavioral and environmental context is critical in efforts to monitor and understand underlying nervous system function.\u00a0 Unfortunately recording neural activity in freely behaving animals in their natural environment is largely prohibited by current technology.\u00a0 In particular, intracellular recordings from neurons and real time manipulations of their electrical signals requires fragile glass electrodes and large, cumbersome electronic devices.\u00a0 Electronics can be miniaturized, however conventional glass microelectrodes present a greater challenge.\u00a0 We are working to build intracellular electrode arrays from the same silicon substrate used in the fabrication of integrated circuits.\u00a0 Our goals are to integrate the microelectrodes with the circuits used to amplify, process and store the electrical signals and to shrink the entire package to the point where it can become implantable.\u00a0 Construction of silicon electrodes adequately sharp for penetration of neuron cell membranes has proven possible and several prototypes have recorded intracellular electrical signals from neurons in the isolated brain of the sea slug Tritonia diomedea. Indeed, thus far we have encountered no insurmountable problems of a biological nature, and we suggest that using such devices to record the neural activity of behaving animals in their natural habitat holds potential for the future. (With Dennis Willows.)\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Online Probabilistic Inference in Neural Populations\"]\r\n\r\nRichard Zemel, Computer Science Department, University of Toronto\r\n\r\nAs animals interact with their environments, they must constantly update estimates about relevant states of the world. For example, a batter must rapidly re-estimate the velocity of a baseball as he decides whether and when to swing at a pitch. Bayesian models provide a description of statistically optimal updating based on prior probabilities, a dynamical model, and sensory evidence, and have proved to be consistent with the results of many diverse psychophysical studies. In this talk I will review various schemes that have been proposed for how populations of neurons can represent the uncertainty that underlies this probabilistic formulation. I will also propose a particular formalism for forming a spike train in a population of neurons to effectively maintain a proper probabilistic representation of the dynamic state. A focus of the talk will be on how models based on standard, simple neural architecture and activations can effectively implement, or at least approximate, this optimal computation, which should make the model applicable to a range of biological systems.\r\n\r\n[\/panel]\r\n\r\n[\/accordion]"}],"msr_startdate":"2005-06-07","msr_enddate":"2005-06-10","msr_event_time":"","msr_location":"San Juan Island, Washington, USA","msr_event_link":"","msr_event_recording_link":"","msr_startdate_formatted":"June 7, 2005","msr_register_text":"Watch now","msr_cta_link":"","msr_cta_text":"","msr_cta_bi_name":"","featured_image_thumbnail":null,"event_excerpt":"The 2005 meeting comes 7 years after a predecessor symposium that we organized in August 1998 under the same title. Strong positive feedback from attendees and new research directions inspired by that meeting compelled us to put together another symposium. Numerous advances have come since the last meeting, some from attendees of the meeting and their teams. The intervening years have also seen some growing interest within neurobiology of the decision making under uncertainty perspective.…","msr_research_lab":[199565],"related-researchers":[],"msr_impact_theme":[],"related-academic-programs":[],"related-groups":[],"related-projects":[],"related-opportunities":[],"related-publications":[],"related-videos":[],"related-posts":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/498908"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-event"}],"version-history":[{"count":10,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/498908\/revisions"}],"predecessor-version":[{"id":501794,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/498908\/revisions\/501794"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=498908"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=498908"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=498908"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=498908"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=498908"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=498908"},{"taxonomy":"msr-program-audience","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-program-audience?post=498908"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=498908"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=498908"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}