3 July 2017 in IEEE Journal of Solid-State Circuits
An energy-efficient restricted Boltzmann machine (RBM) processor (RBM-P) supporting on-chip learning and inference is proposed for machine learning and Internet of Things (IoT) applications in this paper. To train a neural network (NN) model, the RBM structure is applied to supervised and unsupervised learning, and a multi-layer NN can be constructed and initialized by stacking multiple RBMs. Featuring NN model reduction for external memory bandwidth saving, low power neuron binarizer (LPNB) with dynamic clock gating and area-efficient NN-like activation function calculators for power reduction, user-defined connection map (UDCM) for both computation time and bandwidth saving, and early stopping (ES) mechanism for learning process, the proposed system integrates 32 RBM cores with maximal 4k neurons per layer and 128 candidates per sample for machine learning applications. Implemented in 65nm CMOS technology, the proposed RBM-P chip costs 2.2 M gates and 128 kB SRAM with 8.8 mm 2 area. Operated at 1.2 V and 210 MHz, this chip achieves 7.53G neuron weights (NWs) and 11.63G NWs per second with 41.3 and 26.7 pJ per NW for learning and inference, respectively.
19 March 2019 in IEEE Solid-State Circuits Letters
This paper proposes the world’s first deep learning (DL)-assisted video encoder LSI fabricated in a 10nm process with a core area of 0.76mm2 to integrate quad-core DL accelerators and 4K×2K H.264/H.265 video encoders. A visual-contact-field network (VCFNet) DL model is newly designed to predict human focus information with extraordinary reduction of encoding complexity, leading to 82.3% power reduction. Moreover, input channel reduction and layer merging approaches reduce VCFNet complexity by 69%. Operated at 0.9V and 504MHz, the proposed DL-assisted 4K video encoder LSI consumes 56.54mW to achieve 0.22nJ/pixel of energy efficiency, cutting 0.1-14nJ/pixel compared to conventional designs 1-3.