Compete to Compute
نویسندگان
چکیده
Local competition among neighboring neurons is common in biological neural networks (NNs). We apply the concept to gradient-based, backprop-trained artificial multilayer NNs. NNs with competing linear units tend to outperform those with non-competing nonlinear units, and avoid catastrophic forgetting when training sets change over time.
منابع مشابه
Using DEA to compute most favourable and least favourable sets of weights in ABC inventory classification
متن کامل
ارائه الگوریتم پویا برای تنظیم همروندی فرایندهای کسبوکار
Business process management systems (BPMS) are vital complex information systems to compete in the global market and to increase economic productivity. Workload balancing of resources in BPMS is one of the challenges have been long studied by researchers. Workload balancing of resources increases the system stability, improves the efficiency of the resources and enhances the quality of their pr...
متن کاملAn Algorithm to Compute the Complexity of a Static Production Planning (RESEARCH NOTE)
Complexity is one of the most important issues of any production planning. The increase in complexity of production planning can cause inconsistency between a production plan and an actual outcome. The complexity generally can be divided in two categories, the static complexity and the dynamic complexity, which can be computed using the ant ropy formula. The formula considers the probability of...
متن کاملApplication of semi-analytic method to compute the moments for solution of logistic model
The population growth, is increase in the number of individuals in population and it depends on some random environment effects. There are several different mathematical models for population growth. These models are suitable tool to predict future population growth. One of these models is logistic model. In this paper, by using Feynman-Kac formula, the Adomian decomposition method is applied to ...
متن کاملAn Efficient Schulz-type Method to Compute the Moore-Penrose Inverse
A new Schulz-type method to compute the Moore-Penrose inverse of a matrix is proposed. Every iteration of the method involves four matrix multiplications. It is proved that this method converge with fourth-order. A wide set of numerical comparisons shows that the average number of matrix multiplications and the average CPU time of our method are considerably less than those of other methods.
متن کامل