IR is typically a sigmoidal function such as (1.2) but other choices than (1.2) are possible and of interest.

The algorithm presented performs gradient descent on the weight space of an Artificial Neural Network (ANN), using a finite difference to approximate the gradient The method is novel in that it achieves a computational complexity similar to that of Node Perturbation, O(N3), but does not require access to the activity of hidden or internal neurons. This is possible due to a stochastic relation between perturbations at the weights and the neurons of an ANN. The algorithm is also similar to Weight Perturbation in that it is optimal in terms of hardware requirements when used for the training ofVLSI implementations of ANN's.

Holographic Recurrent Networks (HRNs) are recurrent networks which incorporate associative memory techniques for storing sequential structure. HRNs can be easily and quickly trained using gradient descent techniques to generate sequences of discrete outputs and trajectories through continuous spaee. The performance of HRNs is found to be superior to that of ordinary recurrent networks on these sequence generation tasks.

Hidden units in multi-layer networks form a representation space in which each region can be identified with a class of equivalent outputs (Elman, 1989) or a logical state in a finite state machine (Cleeremans, Servan-Schreiber & McClelland, 1989; Giles, Sun, Chen, Lee, & Chen, 1990). We extend the analysis of the spatial structure of hidden unit space to a combinatorial task, based on binding features together in a visual scene. The logical structure requires a combinatorial number of states to represent all valid scenes. On analysing our networks, we find that the high dimensionality of hidden unit space is exploited by using the intersection of neighboring regions to represent conjunctions of features. These results show how combinatorial structure can be based on the spatial nature of networks, and not just on their emulation of logical structure.

The algorithm presented performs gradient descent on the weight space of an Artificial Neural Network (ANN), using a finite difference to approximate the gradient The method is novel in that it achieves a computational complexitysimilar to that of Node Perturbation, O(N3), but does not require access to the activity of hidden or internal neurons. This is possible due to a stochastic relation between perturbations at the weights and the neurons of an ANN. The algorithm is also similar to Weight Perturbation in that it is optimal in terms of hardware requirements whenused for the training ofVLSI implementations of ANN's.

IR is typically a sigmoidal function such as (1.2) but other choices than (1.2) are possible and of interest.

These results show how combinatorial structure can be based on the spatial nature of networks, and not just on their emulation of logical structure.

The methodological underpinnings of AI are slowly changing. Benchmarks, test beds, and controlled experimentation are becoming more common. Although we are optimistic that this change can solidify the science of AI, we also recognize a set of difficult issues concerning the appropriate use of this methodology. We discuss these issues as they relate to research on agent design. We survey existing test beds for agents and argue for appropriate caution in their use. We end with a debate on the proper role of experimental methodology in the design and validation of planning agents.

We report on the spring 1992 symposium on diagrammatic representations in reasoning and problem solving sponsored by the Association for the Advancement of Artificial Intelligence. The symposium brought together psychologists, computer scientists, and philosophers to discuss a range of issues covering both externally represented diagrams and mental images and both psychology -- and AI-related issues. In this article, we develop a framework for thinking about the issues that were the focus of the symposium as well as report on the discussions that took place. We anticipate that traditional symbolic representations will increasingly be combined with iconic representations in future AI research and technology and that this symposium is simply the first of many that will be devoted to this topic.

The field of AI is directed at the fundamental problem of how the mind works; its approach, among other things, is to try to simulate its working -- in bits and pieces. History shows us that mankind has been trying to do this for certainly hundreds of years, but the blooming of current computer technology has sparked an explosion in the research we can now do. The center of AI is the wonderful capacity we call learning, which the field is paying increasing attention to. Learning is difficult and easy, complicated and simple, and most research doesn't look at many aspects of its complexity. However, we in the AI field are starting. Let us now celebrate the efforts of our forebears and rejoice in our own efforts, so that our successors can thrive in their research. This article is the substance, edited and adapted, of the keynote address given at the 1992 annual meeting of the Association for the Advancement of Artificial Intelligence on 14 July in San Jose, California.