A method to generate predictor circuits for the classification of tabular data

Differences between current approaches of AutoML, NAS, NAIS and our auto tiny classifier circuits. a,b, AutoML (a) and NAS (b) generate an ML model and a neural architecture model, respectively, with maximized prediction performance. However, the ML model must be translated into RTL and verified. c, NAIS selects a specific neural network (NN) and a known neural network accelerator to iterate over the space, identifying the best parameters from the hardware (HW) pool to maximize the prediction accuracy. d, Our proposed methodology automatically searches the classifier circuit space using an evolutionary algorithm. During circuit evolution, the generated circuit does …

Be the first to comment

Leave a Reply

Your email address will not be published.