We present a fast, robust algorithm for applying a matrix transfer function of a linear time invariant system (LTI) in time domain. Computing $L$ states of a multiple-input multiple-output (MIMO) LTI appears to require $L$ matrix-vector multiplications. We demonstrate that, for any finite user-selected accuracy, the number of matrix-vector multiplications can be reduced to $\mathcal{O}\left(\log_{2}L\right)$ (within an $\mathcal{O}\left(L\right)$ algorithm). The algorithm uses an approximation of the rational transfer function in the z-domain by a matrix polynomial of degree $2^{N+1}-1$, where $N$ is chosen to achieve any user-selected accuracy. Importantly, using a cascade implementation in time domain, applying the transfer function requires only $N+1$ matrix-vector multiplications. We note that LTI systems are used in state space models (SSMs) for modeling long range dependencies where $L$ is large. In applications where the state matrix of LTI system is approximated by a structured matrix, the computational cost is further reduced. We briefly describe several structured approximations of matrices that can be used for such purpose.

Fast convolution algorithms, including Winograd and FFT, can efficiently accelerate convolution operations in deep models. However, these algorithms depend on high-precision arithmetic to maintain inference accuracy, which conflicts with the model quantization. To resolve this conflict and further improve the efficiency of quantized convolution, we proposes SFC, a new algebra transform for fast convolution by extending the Discrete Fourier Transform (DFT) with symbolic computing, in which only additions are required to perform the transformation at specific transform points, avoiding the calculation of irrational number and reducing the requirement for precision. Additionally, we enhance convolution efficiency by introducing correction terms to convert invalid circular convolution outputs of the Fourier method into effective ones. The numerical error analysis is presented for the first time in this type of work and proves that our algorithms can provide a 3.68x multiplication reduction for 3x3 convolution, while the Winograd algorithm only achieves a 2.25x reduction with similarly low numerical errors. Experiments carried out on benchmarks and FPGA show that our new algorithms can further improve the computation efficiency of quantized models while maintaining accuracy, surpassing both the quantization-alone method and existing works on fast convolution quantization.

Signal processing and pattern recognition algorithms make exten(cid:173) sive use of convolution. In many cases, computational accuracy is not as important as computational speed. This form of noise justifies some level of quantization in order to achieve faster feature extraction . Our approach consists of approximating regions of the signal with low degree polynomi(cid:173) als, and then differentiating the resulting signals in order to obtain impulse functions (or derivatives of impulse functions). With this representation, convolution becomes extremely simple and can be implemented quite effectively.

Feature extraction is a typical example: The distance between a small pattern (i.e.

Feature extraction is a typical example: The distance between a small pattern (i.e.

Feature extraction is a typical example: The distance between a small pattern (i.e.