Dynamic Partition Models
نویسندگان
چکیده
We present a new approach for learning compact and intuitive distributed representations with binary encoding. Rather than summing up expert votes as in products of experts, we employ for each variable the opinion of the most reliable expert. Data points are hence explained through a partitioning of the variables into expert supports. The partitions are dynamically adapted based on which experts are active. During the learning phase we adopt a smoothed version of this model that uses separate mixtures for each data dimension. In our experiments we achieve accurate reconstructions of high-dimensional data points with at most a dozen experts.
منابع مشابه
Comparison of Geotextile Layers Effects on Static and Dynamic behavior of Pavement
One of the methods for the improvement of soil against the tensile stresses is using the artificial supplements in soil such as metal belts or nets, polymer materials and natural fibers. Technically this method is called the reinforced of soil. When a pavement is located on soft soil, great deformations can occur as a result of the crossing of traffic load in superstructure layers of the body....
متن کاملQSAR models to predict physico-chemical Properties of some barbiturate derivatives using molecular descriptors and genetic algorithm- multiple linear regressions
In this study the relationship between choosing appropriate descriptors by genetic algorithm to the Polarizability (POL), Molar Refractivity (MR) and Octanol/water Partition Coefficient (LogP) of barbiturates is studied. The chemical structures of the molecules were optimized using ab initio 6-31G basis set method and Polak-Ribiere algorithm with conjugated gradient within HyperChem 8.0 environ...
متن کاملPredictive Macro-Finance With Dynamic Partition Models
Dynamic partition models are used to predict movements in the term structure of interest rates. This allows one to understand historic cycles in the performance of how interest rates behave, and to offer policy makers guidance regarding future expectations on their evolution. Our approach allows for a random number of possible change points in the term structure of interest rates. We use partic...
متن کاملPerformance Evaluation of Dynamic Modulus Predictive Models for Asphalt Mixtures
Dynamic modulus characterizes the viscoelastic behavior of asphalt materials and is the most important input parameter for design and rehabilitation of flexible pavements using Mechanistic–Empirical Pavement Design Guide (MEPDG). Laboratory determination of dynamic modulus is very expensive and time consuming. To overcome this challenge, several predictive models were developed to determine dyn...
متن کاملAnalysis Efficieny of Graph Computation Based on Partitioning Techniques
Graph computation system performance is generally affected by the quality of graph partition. The balance factor and small edge cut ratio is used to measure the quality of graph partition. A balance graph partition with small edge cut ratio is generally preferred since it cut down the expensive network communication cost. Even so, according to the study of the Giraph, the execution over well pa...
متن کاملAn Eventual α Partition-Participant Detector for MANETs
With Mobile Ad hoc Networks (MANETs), communication between mobile users is possible without any infrastructure. MANETs are already a necessary part of wireless systems. Due to arbitrary node arrivals, departures, crashes and movements, network partitioning may arise resulting in a degradation of the service, but not necessarily in its interruption. In this paper, we propose a distributed syste...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
- CoRR
دوره abs/1702.04832 شماره
صفحات -
تاریخ انتشار 2017