Notes taken by Roman Kecher “The Ecology of Collective Behavior”, Deborah Gordon (15:30-16:05). Abstract: Ant colonies operate without central control, using networks of local interactions to regulate their behavior. Ants are a large and diverse taxon that have evolved in very diverse environments. Examples from ants provide a starting point for examining distributed solutions to environmental problems.I will discuss how environmental constraints, such as the patchiness of resources, the operating costs of maintaining the interaction network that regulates the system’s behavior, and the threat of rupture of the network, all shape the evolution of the algorithms used. Examples are the use of autocatalysis in harvester ants to regulate foraging, analogous to TCP-IP in data networks (“Anternet”), and the regulation of load balancing in the trail circuits of a tropical arboreal species. Notes: Essentially, the rate of interactions within the ant colony is in itself an important form of communication. For instance, interaction rate gives information on the size of the colony; only when the colony gets big enough, new queens which are able to reproduce will start appearing. Ants interact through a grease layer on their bodies which they are able to sense. Experiments have been made that show that ants react differently when that property changes. They also react differently when the interaction rate changes (for instance, when more ants are introduced). In collective search there’s a tradeoff between the thoroughness of the search and the speed at which more ground is covered. Argentinien ants actually operate differently according to the size of the area they’re covering: random walks are used in small spaces, while more straight stretched lines are used in bigger areas. Rate of interaction is used in order to estimate the size of the area. There’s also been an experiment conducted in a space station, with zero gravity, which was fairly close to that, just with a three dimensional search areas (as ants could then fly). In different environments, the actual information which is derived from the interaction rate is used differently: In a desert, the operating costs are high (you must spend water to get water). So ants only go out to forage if the interaction rate with returning foragers is high. Otherwise they stay in the nest. An experiment actually showed that if returning foragers are not allowed to get back to the nest for three minutes, the rate of outgoing foragers drops after some delay (to a certain minimum). This resembles the TCP/IP protocol which relies on ACKs in order to send the next packets. In a tropical forest, on the other hand, ants forage by default. They just stop foraging if the interaction rate is too low, since this might indicate a problem. It is actually better to forage less in dry days (in the desert for instance), and this is achieved through natural selection; offsprings tend to mimic their parents in this regard, and mapping ancestry relations in a specific area showed that colonies which forage less are indeed more successful (have more offsprings). “Aunt Clara, Francesca and her friends”, Laurent Keller (16:10-16:30). Title might be wrong. Notes: Ants are one of the most social animals. A tracking systems for ants was employed in order to track the interaction between all the ants for a few weeks, to provide insights about the social hierarchy of ants. Graphs have been constructed from the interaction information, and were analyzed in order to understand how many different groups are there. As it turns out, there are three distinct groups: nurses (ants closer to the queen), cleaners and foragers (closer to the exit). What is the criterion which decides on the group association? It seems that the age of the ants does not influence it, nor does their size. It also seems that ants transition from one role to the other, but the transitions are always in a specific order: nurses -> cleaners -> foragers. Also, most interactions are done within members of the same role. There are two possible reasons for that: either the ants selectively do that, or this is due to spatial proximity (ants locations). Careful analysis showed that ants interact randomly with other ants who are close to them, therefore the latter reason is the right one; the role decides the location of the ants, and once there the interaction is random. This also leads to the fact that information spreads fairly fast: first inside of members of the same role, but after 1-2 hours all the ants have it. The transitioning between the different roles matches a ‘social maturity’ property: as ants interact more with other ants and they are more socially mature, they will transition between the different roles. However, it is still unknown as to what actually triggers those changes. Poster session (16:30-18:00) “A binary Hopfield network with $1/\log(n)$ information rate and applications to grid cell decoding”, Ila Fiete, David Schwab and Ngoc Tran. A Hopfield network is an auto-associative, distributive model of neural memory storage and retrieval. A form of error-correcting code, the Hopfield network can learn a set of patterns as stable points of the network dynamic, and retrieve them from noisy inputs -- thus Hopfield networks are their own decoders. Unlike in coding theory, where the information rate of a good code (in the Shannon sense) is finite but the cost of decoding does not play a role in the rate, the information rate of Hopfield networks trained with state-of-the-art learning algorithms is of the order log(n)/n, a quantity that tends to zero asymptotically with n, the number of neurons in the network. For specially constructed networks, the best information rate currently achieved is of order 1/sqrt(n). In this work, the authors design simple binary Hopfield networks that have asymptotically vanishing error rates at an information rate of 1/log(n). These networks can be added as the decoders of any neural code with noisy neurons. As an example, they apply their network to a binary neural decoder of the grid cell code to attain information rate 1/log(n). “On Optimal Decision-Making in Ant Colonies”, Mahnush Movahedi and Mahdi Zamani. Colonies of ants can collectively choose the best of several nests, even when many of the active ants who organize the move visit only one site. Understanding such a behavior can help us design efficient distributed decision making algorithms. Marshall et al. propose a model for house-hunting in colonies of ant Temnothorax albipennis. Unfortunately, their model does not achieve optimal decision-making while laboratory experiments show that, in fact, colonies usually achieve optimality during the house-hunting process. In this paper, the authors argue that the model of Marshall et al. can achieve optimality by including nest size information in their mathematical model. They use lab results of Pratt et al. to re-define the differential equations of Marshall et al. Finally, they sketch a strategy for testing the optimality of the new model. “Know when you’re beaten: Efficient cooperative transport requires either a directional bias or that outnumbered individuals give up quickly”, Helen Mccreery and Nikolaus Correll. Ants are excellent examples of self-organized systems capable of remarkable coordination. Here, the authors model some proximate mechanisms that may lead to this coordination in the context of cooperative transport, which occurs when a group of ants work together to move a large object intact. Using a deterministic model in continuous time with one implicit spatial dimension, the authors examine the initial phase of transport, or how individuals agree on a direction to move the object. Key parameters of the model are the extent of bias in individuals’ preferred direction of movement, and persistence, or the reluctance of individuals to give up their preferred direction. In this deterministic system at least a small directional bias is necessary for movement. When directional bias is small, efficiency is maximized when individuals give up more readily if the transport is slow and they are outnumbered (more individuals pulling the opposite direction). In this situation high persistence reduces efficiency. However, high persistence increases efficiency if individuals either 1) give up more readily when transport is slow regardless of whether they are outnumbered or 2) give up at the same rate regardless of transport success. This model provides new insight into behavioral parameters that may modulate the cooperative transport efficiency of ant species in nature.