Landmarking is a novel approach to describing tasks in meta-learning. Previous approaches to meta-learning mostly considered only statistics-inspired measures of the data as a source for the definition of metaattributes. Contrary to such approaches, landmarking tries to determine the location of a specific learning problem in the space of all learning problems by directly measuring the performance of some simple and efficient learning algorithms themselves. In the experiments reported we show how such a use of landmark values can help to distinguish between areas of the learning space favouring different learners. Experiments, both with artificial and real-world databases, show that landmarking selects, with moderate but reasonable level of success, the best performing of a set of learning algorithms. 1.

Humans routinely generalize universal relationships to unfamiliar instances. If we are told ‘‘if glork then frum,’ ’ and ‘‘glork,’ ’ we can infer ‘‘frum’’; any name that serves as the subject of a sentence can appear as the object of a sentence. These universals are pervasive in language and reasoning. One account of how they are generalized holds that humans possess mechanisms that manipulate symbols and variables; an alternative account holds that symbol-manipulation can be eliminated from scientific theories in favor of descriptions couched in terms of networks of interconnected nodes. Can these ‘‘eliminative’ ’ connectionist models offer a genuine alternative? This article shows that eliminative connectionist models cannot account for how we extend universals to arbitrary items. The argument runs as follows. First, if these models, as currently conceived, were to extend universals to arbitrary instances, they would have to generalize outside the space of training examples. Next, it is shown that the class of eliminative connectionist models that is currently popular cannot learn to extend universals outside the training space. This limitation might be avoided through the use of an architecture that implements symbol manipulation.

We show that complex visual tasks, such as position and size invariant shape recognition and navigation in the environment, can be tackled with simple architectures generated by a co-evolutionary process of active vision and feature selection. Behavioral machines equipped with primitive vision systems and direct pathways between visual and motor neurons are evolved while freely interacting with their environments. We describe the application of this methodology in three sets of experiments, namely shape discrimination, car driving, and robot navigation. We show that these systems develop sensitivity to a number of oriented, retinotopic, visual features oriented edges, corners, height – and a behavioral repertoire to locate, bring, and keep these features in sensitive regions of the vision system, resembling strategies observed in simple insects.

We describe an evolutionary approach to active vision systems for dynamic feature selection. After summarizing recent work on evolution of a simulated active retina for complex shape discrimination, we describe in detail experiments that extend this approach to an all-terrain mobile robot equipped with a mobile camera. We show that evolved robots are capable of selecting simple visual features and actively maintaining them on the same retinal position, which largely simplifies the “recognition ” task, in order to generate efficient navigation trajectories with an extremely simple neural control system. Analysis of evolved solutions indicates that robots develop a simple

In this paper I will show how reactive agents can solve relatively complex tasks without requiring any internal state and I will demonstrate that this is due to their ability to coordinate perception and action. By acting (i.e. by modifying their position with respect to the external environment and/or the external environment itself), agents partially determine the sensory patterns they receive from the environment. As I will show, agents can take advantage of this ability to: (1) select sensory patterns that are not affected by the aliasing problem and avoiding those that are; (2) select sensory patterns in which groups of patterns requiring different answers do not strongly overlap; (3) exploit the fact that, given a certain behavior, sensory states might indirectly encode information about useful features of the environment; (4) exploit emergent behaviors resulting from a sequence of sensory-motor loops and from the interaction between the robot and the environment. Final...

Constructivism is the Piagetian notion that learning leads the child to develop new types of representations. For example, on the Piagetian view, a child is born without knowing that objects persist in time even when they are occluded; through a process of learning, the child comes to know that objects persist in time. The trouble with this view has always been the lack of a concrete, computational account of how a learning mechanism could lead to such a change. Recently, however, in a book entitled Rethinking Innateness, Elman et al. (Elman,

We introduce a genetic algorithm (GA) with a new representation method which we call the proportional GA (PGA). The PGA is a multi-character GA that relies on the existence or non-existence of genes to determine the information that is expressed. The information represented by a PGA individual depends only on what is present on the individual and not on the order in which it is present. As a result, the order of the encoded information is free to evolve in response factors other than the value of the solution, for example, in response to the identification and formation of building blocks. The PGA is also able to dynamically evolve the resolution of encoded information. In this paper, we describe our motivations for developing this representation and provide a detailed description of a PGA along with discussion of its benefits and drawbacks. We compare the behavior of a PGA with that of a canonical GA (CGA) and discuss conclusions and future work based on these preliminary studies.

We explore incremental assimilation of new knowledge by sequential learning. Of particular interest is how a network of many knowledge layers can be constructed in an on-line manner, such that the learned units represent building blocks of knowledge that serve to compress the overall representation and facilitate transfer. We motivate the need for many layers of knowledge, and we advocate sequential learning as an avenue for promoting construction of layered knowledge structures. Finally, our novel STL algorithm demonstrates an efficient method for simultaneously acquiring and organizing a collection of concepts and functions from a stream of rich but otherwise unstructured information. 1

This thesis combines an epistemological concern about induction with a computational exploration of inductive mechanisms. It aims to investigate how inductive performance could be improved by using induction to select appropriate generalisation procedures. The thesis revolves around a meta-learning system, called designed to investigate how inductive performances could be improved by using induction to select appropriate generalisation procedures. The performance of is discussed against the background of epistemological issues concerning induction, such as the role of theoretical vocabularies and the value of simplicity.

. One important implication of embodiment is that, by acting, agents partially determine the sensory patterns they receive from the environment. The motor actions performed by an agent, by modifying the agent's position with respect to the external environment and/or the external environment itself, partially determine the type of sensory patterns received from the environment. In this paper we investigate how agents can take advantage of this ability. In particular, we discuss how agents coordinate sensory and motor processes in order to (1) select sensory patterns which are not affected by the aliasing problem and avoid those which are; (2) select sensory patterns such that groups of patterns which require different responses do not strongly overlap; (3) exploit emergent behaviors that result from the interaction between the agent and the environment. 1. Introduction Recently, a new research paradigm has challenged the traditional view according to which intelligence is ...