We present an approach for learning the time- and parameter dependent solutions of electrical circuits [1]. Learning the solution becomes an interesting alternative to direct simulation when the number of parameters is large or many realizations are required, such as e.g. in design optimization, uncertainty quantification or when coupling the circuit simulator with other software in the context of multiphysics simulations. One of the most commonly used formulations for circuit simulation is the modified nodal analysis (MNA) [2]. This formulation leads to differential-algebraic equations (DAEs), which require extra care compared to their ordinary differential equation (ODE) counterpart [3]. In order to avoid DAE related difficulties, the approach makes use of the dissection index [4] to decouple the DAEs into ODEs and a set of algebraic equations. This guarantees that the learned solution fulfills the algebraic constraints up to the accuracy of the nonlinear system solver, reduces the number of solution components to learn, and also makes it possible to exploit the ODE structure to improve the learning process. The approach takes inspiration from previous work on index-aware model order reduction (MOR) [5], however there are two main differences. Firstly, the index-aware MOR approach from [5] uses numerically computed matrix kernels, whereas the dissection index allows for a topological description of these kernels in the case of MNA, and secondly, we transfer the idea from MOR to the context of machine learning. [1] I. Cortes Garcia, et. al, Index-aware learning of circuits. 2023. [2] C. Ho, A. Ruehli and P. Brennan, The modified nodal approach to network analysis. IEEE Trans. on Circuits and Syst., Vol. 22, pp. 504-509, 1975. [3] C. Tischendorf, Topological index calculation of differential-algebraic equations in circuit simulation. Surv. on Math. for Ind.. Vol. 8, pp. 187-199, 1999. [4] L. Jansen, A Dissection Concept for DAEs, Humboldt-Universität zu Berlin, 2014. [5] G. Alì, N. Banagaaya, W. Schilders and C. Tischendorf, Index-aware model order reduction for differential-algebraic equations. Math. and Comput. Model. of Dyn. Syst.. Vol. 20, pp. 345-373, 2014.