Core Concepts
The authors explore distributed estimation by two agents using different feature spaces, presenting a framework for fusion and knowledge retrieval.
Abstract
The content discusses the problem of estimating functions by two agents with different features. It covers the construction of individual knowledge spaces, fusion space, uploading and downloading operators, regression problems, and function fusion problems. The example provided demonstrates the application of the theory developed in the previous sections to estimate a real-valued cubic polynomial. The results show how different kernels used by agents can impact estimation accuracy.
Stats
Data comprises samples of an independent variable and corresponding values of a dependent variable.
Agent 1 was provided with 20 samples of input-output data on the interval [-5, 5].
Agent 2 was provided with 20 samples of input-output data on the interval [-10, -5] ∪ [5, 10].
Fusion problem involves finding a linear combination to minimize dissimilarity between functions.
Centralized estimation method processes data from both agents simultaneously.
Quotes
"The fused function is considered as the function estimated by the system."
"By choosing different kernels, agents tune their estimation procedure to specific data collected."
"The fused function is the best estimate among all estimates considered so far."