Efficient Message Computation in Sigma’s Graphical Architecture
نویسندگان
چکیده
Human cognition runs at ~50 msec per cognitive cycle, implying that any biologically inspired cognitive architecture that strives for real-time performance needs to be able to run at this speed. Sigma is a cognitive architecture built upon graphical models – a broadly applicable state-of-the-art formalism for implementing cognitive capabilities – that are solved via message passing (with complex messages based on n-dimensional piecewise-linear functions). Earlier work explored optimizations to Sigma that reduced by an order of magnitude the number of messages sent per cycle. Here, optimizations are introduced that reduce by an order of magnitude the average time required per message sent.
منابع مشابه
Rethinking Sigma's Graphical Architecture: An Extension to Neural Networks
The status of Sigma’s grounding in graphical models is challenged by the ways in which their semantics has been violated while incorporating rule-based reasoning into them. This has led to a rethinking of what goes on in its graphical architecture, with results that include a straightforward extension to feedforward neural networks (although not yet with learning).
متن کاملToward a Neural-Symbolic Sigma: Introducing Neural Network Learning
Building on earlier work extending Sigma’s mixed (symbols + probabilities) graphical band to inference in feedforward neural networks, two forms of neural network learning – target propagation and backpropagation – are introduced, bringing Sigma closer to a full neural-symbolic architecture. Adapting Sigma’s reinforcement learning (RL) capability to use neural networks in policy learning then y...
متن کاملTowards Emotion in Sigma: From Appraisal to Attention
A first step is taken towards incorporating emotional processing into Sigma, a cognitive architecture that is grounded in graphical models, with the addition of appraisal variables for expectedness and desirability plus their initial implications for attention at two levels of the control hierarchy. The results leverage many of Sigma’s existing capabilities but with a few key additions.
متن کاملEfficient graphical models for sequence segmentation
Segmentation of sequences is an important modeling primitive with several applications. Training and inference of segmentation models involves dynamic programming computations that in the worst case can be cubic in the length of a sequence. In contrast, typical sequence labeling models require linear time. We propose an alternative graphical model for efficient sharing of potentials across over...
متن کاملThesis Proposal Parallel Learning and Inference in Probabilistic Graphical Models
Probabilistic graphical models are one of the most influential and widely used techniques in machine learning. Powered by exponential gains in processor technology, graphical models have been successfully applied to a wide range of increasingly large and complex real-world problems. However, recent developments in computer architecture, large-scale computing, and data-storage have shifted the f...
متن کامل