Research Lines

Decision trees are a light and efficient machine learning technique that has proved their effectiveness in several classification problems. In the context of embedded systems, energy efficiency is as important as accuracy, so it is necessary to search for efficient algorithms liable to be accelerated. This makes the decision trees a perfect target for developing an FPGA accelerator. A single decision tree is frequently not very accurate for complicated tasks but, thanks to ensemble methods, it is possible to combine several trees in order to deal with complex problems…

Bayesian neural networks (BNNs) allow us to obtain uncertainty metrics related to the data processed, and the uncertainty generated by the model selected. We have identified how these metrics can be used for several practical applications such as identifying predictions that do not reach the required level of accuracy, or identifying when the predictions are affected by the increase of the level of noise in the input data. However, it is important to consider…

This work has explored the possibility of drastically reducing the supply voltage of on­chip activation memories in CNN inference accelerators to save energy consumption. To address the impact on CNN accuracy of bitcell permanent faults as a consequence of supply voltage underscaling, this work has proposed a couple of low­cost microarchitectural mechanisms based on flipping and patching approaches. These mechanisms are a consequence of a characterization study…