IEEE Journal on Emerging and Selected Topics in Circuits and Systems

In convolutional neural networks (CNN), convolutional layers consume dominant portion of computation energy due to large amount of multiply-accumulate operations (MACs). However, those MACs become meaningless (zeroes) after rectified linear unit (ReLU) when the convolution results become negative. In this paper, we present an efficient approach to predict and skip the convolutions generating zero ... View full abstract»

This paper devises a novel adaptive framework for energy-aware acquisition of spectrally-sparse signals. The adaptive quantized Compressive Sensing (CS) techniques, beyond-Complementary Metal Oxide Semiconductor (CMOS) hardware architecture, and corresponding algorithms which utilize them have been designed concomitantly to minimize the overall cost of signal acquisition. First, a spin-based Adapt... View full abstract»

Due to the distributed and asynchronous nature of neural computation through low-energy spikes, brain-inspired hardware systems offer high energy efficiency and massive parallelism. One such platform is the IBM TrueNorth Neurosynaptic System. Recently, TrueNorth compatible representation learning algorithms have emerged, achieving close to state-of-the-art performance in various datasets. However,... View full abstract»

The complexity of the computational problems is rising faster than the computational platforms’ capabilities which are also becoming increasingly costly to operation due to their increased need for energy. This forces researchers to find alternative paradigms and methods for efficient computing. One promising paradigm is accelerating compute intensive kernels using in-memory computing a... View full abstract»

Multicore architectures have been used to enhance computing capabilities, but the energy consumption is still an important concern. Embedded application domains usually require less accurate, but always in-time, results. Imprecise Computation (IC) can be used to divide a task into a mandatory subtask providing a baseline QoS and an optional subtask that further increases the baseline QoS. This wor... View full abstract»

This paper introduces a novel method for designing approximate circuits by fabricating and exploiting false timing paths, i.e. critical paths that cannot be logically activated. This allows to strongly relax timing constraints while guaranteeing minimal and controlled behavioral change. This technique is applied to an approximate adder architecture, called the Carry Cut-Back Adder (CCBA), in which... View full abstract»

Emerging portable applications, including Human Activity Recognition and Robot Navigation, require always-on sensing technologies to continuously monitor the environment. Continuous sensing, however, strongly dominates the hardware devices’ power consumption and consequently hampers the systems’ always-on functionality. In this paper we propose a circuitaware Machine Learning... View full abstract»

Mixed precision is a promising approach to save energy in iterative refinement algorithms since it obtains speedup without necessitating additional cores and parallelisation. However, conventional mixed precision methods utilise statically defined precision in a loop, thus hindering further speed-up and energy savings. We overcome this problem by proposing novel methods which allow iterative refin... View full abstract»

A common scheme to let a very large number of lowresources sensing units communicate their readings to a remote concentrator is to deploy intermediate hubs that collect subsets of readings by means of local communication and perform the needed long-range transmission of a compressed version of the data. We here propose to exploit Compressed Sensing as an extremely lightweight lossy compression sta... View full abstract»

A low-power scalable 3D face frontalization processor is proposed for accurate face recognition in mobile devices. In spite of recent improvement in face recognition accuracy mainly from convolutional neural networks (CNN), their performance is limited to face images with frontal view. For face recognition with human-level accuracy in real life environment, in which most of the face images are cap... View full abstract»

The intrinsic error tolerance of neural network (NN) presents opportunities for approximate computing techniques to improve the energy efficiency of NN inference. Conventional approximate computing focuses on exploiting the efficiencyaccuracy trade-off in existing pre-trained networks, which can lead to suboptimal solutions. In this paper, we first present AxTrain, a hardware-oriented training fra... View full abstract»

Biomedical applications often require classifiers that are both accurate and cheap to implement. Today, deep neural networks achieve state-of-the-art accuracy in most learning tasks that involve large datasets of unstructured data. However, the application of deep learning techniques may not be beneficial in problems with limited training sets and computational resources, or under domain-specific ... View full abstract»

Outstanding seizure detection algorithms have been developed over past two decades. Despite this success, their implementations as part of implantable or wearable devices are still limited. These works are mainly based on heavily handcrafted feature extraction, which is computationally expensive and is shown to be dataset specific. These issues greatly limit the applicability of such methods to ha... View full abstract»

Recently, there has been an increasing demand for advanced classification capabilities embedded on wearable battery constrained devices, such as smartphones or -watches. Achieving such functionality with a tight power and energy budget has proven a real challenge, specifically for large-scale Neural Network based applications. Previously, cascaded systems have been proposed to minimize energy cons... View full abstract»

This paper presents a novel technique to reduce energy consumption of a machine learning classifier based on incremental-precision feature computation and classification. Specifically, the algorithm starts with features computed using the lowest possible precision. Depending on the classification accuracy, the features of the previous level are combined with features of the incremental-precision t... View full abstract»

Deep Neural Networks (DNNs) have emerged as the state-of-the-art technique in a wide range of machine learning tasks for analytics and computer vision in the next generation of embedded (mobile, IoT, wearable) devices. Despite their success, they suffer from high energy requirements. In recent years, the inherent error resiliency of DNNs has been exploited by introducing approximations at either t... View full abstract»

Numerous approximate adders have been proposed in the literature in response to the languishing benefits of technology scaling. However, they have been obtained with an adhoc and non-systematic methodology which does not fully exploit the design space possibilities. This work provides a conceptual framework for the systematic design of approximate adders including hybrid and non-equally segmented ... View full abstract»

This paper presents a structure simplification procedure that allows efficient energy and accuracy tradeoffs in implementation of trained deep neural networks (DNNs). This structure simplification procedure identifies and eliminates redundant neurons in any layer of the DNN based on the output values of these neurons over the training samples. This procedure may be applied to all layers of a DNN. ... View full abstract»