Probabilistic Matrix Addition

ثبت نشده
چکیده

We introduce Probabilistic Matrix Addition (PMA) for modeling real-valued data matrices by simultaneously capturing covariance structure among rows and among columns. PMA additively combines two latent matrices drawn from two Gaussian Processes (GPs) respectively over rows and columns. The resulting joint distribution over the observed matrix does not factorize over entries, rows, or columns, and can thus capture intricate dependencies in the matrix. Exact inference in PMA is possible, but involves inversion of large matrices, and can be computationally prohibitive. Efficient approximate inference is possible due to the sparse dependency structure among latent variables. We propose two families of approximate inference algorithms for PMA based on Gibbs sampling and MAP inference. We demonstrate the effectiveness of PMA for matrix missing values problems and multi-label classification problems.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Performance Modeling of Power Generation System of a Thermal Plant

The present paper discusses the development of a performance model of power generation system of a thermal plant for performance evaluation using Markov technique and probabilistic approach. The study covers two areas: development of a predictive model and evaluation of performance with the help of developed model. The present system of thermal plant under study consists of four subsystems with...

متن کامل

Development of simulation model for performance evaluation of feed water system in a typical thermal power plant

The present paper deals with development of a simulation model for the performance evaluation of feed water system of a thermal power plant using Markov Birth-Death process and probabilistic approach. In present paper, the feed water system consists of four subsystems. After drawing transition diagram for feed water system, differential equations are developed and then solved recursively using ...

متن کامل

Correspondence between probabilistic norms and fuzzy norms

In this paper, the connection between Menger probabilistic norms and H"{o}hle probabilistic norms is discussed. In addition, the correspondence between probabilistic norms and Wu-Fang fuzzy (semi-) norms is established. It is shown that a probabilistic norm (with triangular norm $min$) can generate a Wu-Fang fuzzy semi-norm and conversely, a Wu-Fang fuzzy norm can generate a probabilistic norm.

متن کامل

Probabilistic Matrix Addition

We introduce Probabilistic Matrix Addition (PMA) for modeling real-valued data matrices by simultaneously capturing covariance structure among rows and among columns. PMA additively combines two latent matrices drawn from two Gaussian Processes respectively over rows and columns. The resulting joint distribution over the observed matrix does not factorize over entries, rows, or columns, and can...

متن کامل

Completeness in Probabilistic Metric Spaces

The idea of probabilistic metric space was introduced by Menger and he showed that probabilistic metric spaces are generalizations of metric spaces. Thus, in this paper, we prove some of the important features and theorems and conclusions that are found in metric spaces. At the beginning of this paper, the distance distribution functions are proposed. These functions are essential in defining p...

متن کامل

An Investigation and assessment of meteorological drought in Lake Urmia Basin using drought indices and probabilistic methods

Drought is one of the most important weather-induced phenomena which may have severe impacts on different areas such as agriculture, economy, energy production, and society. A number of drought indices have been introduced and used in various countries to date. In the current study, four meteorological drought indices including Percent of Normal Precipitation Index (PNPI), Standard Index of Ann...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2011