Optimizing FORTRAN Programs for Hierarchical Memory Parallel Processing Systems

Parallel loops account for the greatest amount of parallelism in numerical programs.Executing nested loops in parallel with low run-time overhead is thus very important for achieving high perform- ance in parallel processing systems.However,in parallel processing systems with caches or local memo- ries in memory hierarchies,“thrashing problem”may arise whenever data move back and forth between the caches or local memories in different processors. Previous techniques can only deal with the rather simple cases with one linear function in the perfect- ly nested loop.In this paper,we present a parallel program optimizing technique called hybrid loop inter- change(HLI)for the cases with multiple linear functions and loop carried data dependences in the nested loop.With HLI we can easily eliminate or reduce the thrashing phenomena without reducing the program parallelism.

Three-Phase Unbalance Prediction of Electric Power Based on Hierarchical Temporal Memory

The difference in electricity and power usage time leads to an unbalanced current among the three phases in the power grid.The three-phase unbalanced is closely related to power planning and load distribution.When the unbalance occurs,the safe operation of the electrical equipment will be seriously jeopardized.This paper proposes a Hierarchical Temporal Memory(HTM)-based three-phase unbalance prediction model consisted by the encoder for binary coding,the spatial pooler for frequency pattern learning,the temporal pooler for pattern sequence learning,and the sparse distributed representations classifier for unbalance prediction.Following the feasibility of spatial-temporal streaming data analysis,we adopted this brain-liked neural network to a real-time prediction for power load.We applied the model in five cities(Tangshan,Langfang,Qinhuangdao,Chengde,Zhangjiakou)of north China.We experimented with the proposed model and Long Short-term Memory(LSTM)model and analyzed the predict results and real currents.The results show that the predictions conform to the reality;compared to LSTM,the HTM-based prediction model shows enhanced accuracy and stability.The prediction model could serve for the overload warning and the load planning to provide high-quality power grid operation.

Machine Recognition of Plan Typologies： Shotgun and Foursquare

The evolution of expert and knowledge-based systems in architecture requires the gradual population of building specific databases. Often these databases are slow to evolve due to the time consuming nature of effectively categorizing building features in a meaningful way that allows for retrieval and reuse. New advances in artificial intelligence such as Hierarchical Temporal Memory （HTM） have the potential to make the construction of these databases more realistic in the near future. Based on an emerging theory of human neurological function, HTMs excel at ambiguous pattern recognition. This paper includes a first experiment using HTMs for learning and recognizing patterns in the form of two distinct American house plan typologies, and further tests the relationship of HTM＇s recognition tendencies in alternate house plan types. Results from the experiment indicate that HTMs develop a similar storage of quality to humans and are therefore a promising option for capturing multi-modal information in future design automation efforts.

AN APPROACH TOWARDS HOLISM IN SCIENCE AND ENGINEERING

This paper posits the desirability of a shift towards a holistic approach over reductionist approaches in the understanding of complex phenomena encountered in science and engineering. An argument based on set theory is used to analyze three examples that illustrate the shortcomings of the reductionist approach. Using these cases as motivational points, a holistic approach to understand complex phenomena is proposed, whereby the human brain acts as a template to do so. Recognizing the need to maintain the transparency of the analysis provided by reductionism, a promising computational approach is offered by which the brain is used as a template for understanding complex phenomena. Some of the details of implementing this approach are also addressed.