Publication | Open Access
A learning algorithm for boltzmann machines
709
Citations
0
References
1985
Year
EngineeringMachine LearningLearning AlgorithmComputational ComplexityData SciencePhysic Aware Machine LearningPattern RecognitionParallel Complexity TheoryParallel ComputingCombinatorial OptimizationCommunication BandwidthComputational Learning TheoryKnowledge DiscoveryComputer EngineeringLarge Scale OptimizationComputer ScienceStatistical Learning TheoryNeural Architecture SearchComputational ScienceParallel LearningParallel ProgrammingSimple Processing ElementsData-level Parallelism
Massively parallel networks of simple processing elements can solve large constraint‑satisfaction problems quickly by exploiting high‑bandwidth hardware connections, but efficient use requires a suitable parallel search technique and internal representations that encode domain constraints. The study seeks to devise a parallel search method that efficiently encodes domain constraints and informs learning of connection strengths for large constraint‑satisfaction problems. They propose a statistical‑mechanics–based parallel search algorithm that yields a learning rule to adjust connection strengths according to task knowledge. The algorithm produces internal representations that optimally exploit the existing connectivity, as demonstrated in simple examples.
The computational power of massively parallel networks of simple processing elements resides in the communication bandwidth provided by the hardware connections between elements. These connections can allow a significant fraction of the knowledge of the system to be applied to an instance of a problem in a very short time. One kind of computation for which massively parallel networks appear to be well suited is large constraint satisfaction searches, but to use the connections efficiently two conditions must be met: First, a search technique that is suitable for parallel networks must be found. Second, there must be some way of choosing internal representations which allow the preexisting hardware connections to be used efficiently for encoding the constraints in the domain being searched. We describe a general parallel search method, based on statistical mechanics, and we show how it leads to a general learning rule for modifying the connection strengths so as to incorporate knowledge about a task domain in an efficient way. We describe some simple examples in which the learning algorithm creates internal representations that are demonstrably the most efficient way of using the preexisting connectivity structure.