7.5
C499-L1 regularization
C499-TUSC
C880-L1 regularization
C880-TUSC
0.9 1 1.1
#measurements/M
Figure 6.3: Variation estimation error vs. number of power measurements.
power), we used the HSPICE simulator on 65nm CMOS transistor technology.
Figure 6.3 presents variations estimation error for the C499 and the C880
benchmark circuits. The horizontal axis is the power measurement noise and
the vertical axis is the variations estimation error. The variation estimation is
calculated in a 7V∕3-dimensional subspace, where N is the number of gates. Note
that by construction the estimation space is orthogonal to the null space of the
measurement matrix. Thus, for low noise measurements the Z⅛-regularization and
TUSC are very similar. As the noise level increases, TUSC performs better than
the ^ɪ-norm regularization. Note that ^-minimization performs much worse than
^ɪ-regularization and TUSC; it is not shown on the figure, please refer to Table
6.2 for this comparison.
70
More intriguing information
1. Evolutionary Clustering in Indonesian Ethnic Textile Motifs2. Regional specialisation in a transition country - Hungary
3. The name is absent
4. Constrained School Choice
5. The use of formal education in Denmark 1980-1992
6. EMU: some unanswered questions
7. Inflation Targeting and Nonlinear Policy Rules: The Case of Asymmetric Preferences (new title: The Fed's monetary policy rule and U.S. inflation: The case of asymmetric preferences)
8. The storage and use of newborn babies’ blood spot cards: a public consultation
9. NEW DEVELOPMENTS IN FARM PRICE AND INCOME POLICY PROGRAMS: PART I. SITUATION AND PROBLEM
10. Optimal Vehicle Size, Haulage Length, and the Structure of Transport Costs