-Modeling Analog BEhaviour using Learning techniques (IE02028)
Acronym: MABEL
Luis Miguel Teixeira D Avila Pinto da Silveira
From 01-Nov-2007 to 31-Dec-2008
Prime Contractor: INESC-ID (Other)
Financed by: FCT (Other)
Members: Luis Miguel Teixeira D Avila Pinto da Silveira, Josť Carlos Alves Pereira Monteiro, Paulo Ferreira Godinho Flores

We propose to research the applicability of generic learning techniques to the macromodeling of functional behavior or other relevant characteristics of electronic systems that are essentially analog in nature. Specifically, to:

1. Develop an environment for testing and characterizing functional behavior in analog cells using standard learning techniques and to compare the results obtained with generic regression methods in terms of accuracy and generality.

2. Research kernel-based techniques for block macro-modeling of iinput-output cell behavior in mixed-signal systems and compare them to alternative techniques including manual abstraction and standard fitting.

3. Develop a framework for bottom-up automatic macromodeling of cell behavior by systematic composition of basic element models.

4. Develop and test algorithms for generating macromodels for power estimation of digital cell switching and compare them to existing techniques.