Minerva: Enabling Low-Power, Highly-Accurate Deep Neural Network Accelerators

By Brandon Reagen, Paul N. Whatmough, Robert Adolf, Saketh Rama, Hyunkwang Lee, Sae Kyu Lee, José Miguel Hernández-Lobato, Gu-Yeon Wei, David M. Brooks. 2016

613 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Electrical and Electronic Engineering, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Electrical and Electronic Engineering

In Efficiency & Compression

Citations in the corpus

Chasing Carbon: The Elusive Environmental Footprint of Computing (2020, background): “For example, in the case of AI, co-designing neural-network fault tolerance and compression for specialized circuits and memories can allow hardware consolidation and, therefore, smaller carbon footprints [4], [5], [7], [118], [119].”
EIE: Efficient Inference Engine on Compressed Deep Neural Network (2016, methodology): “Minerva [39] also uses data quantization to save memory energy.”
Designing Energy-Efficient Convolutional Neural Networks Using Energy-Aware Pruning (2016, methodology): “The proposed energy-efficient pruning algorithm can be combined with other techniques to further reduce the energy consumption, such as bitwidth reduction of weights or feature maps [15, 25, 26], weight sharing and Huffman coding [8], student-teacher learning [27], filter decomposition [28, 29] and ”
SCNN: An accelerator for compressed-sparse convolutional neural networks (2017, background): “Unlike previous convolutional dataflows [6, 8, 12, 28], the SCNN dataflow only delivers to the multiplier array weights and activations that can all be multiplied with one another in the manner of a Cartesian product.”
Efficient Processing of Deep Neural Networks: A Tutorial and Survey (2017, background): “For instance, if all activations with small values are pruned, this can be translated into an additional 11% speed up [138] or 2× power reduction [139] with little impact on accuracy.”
Compressing DMA Engine: Leveraging Activation Sparsity for Training Deep Neural Networks (2017, background): “A series of accelerator designs have also been proposed for CNNs recently [24], [25], [26], [27], [28], [29], [30], [31], [15], [16], [17].”
Edge Intelligence: Paving the Last Mile of Artificial Intelligence With Edge Computing (2019, neutral): “[74] B. Reagen et al., “Minerva: Enabling low-power, highly-accurate deep neural network accelerators,” ACM SIGARCH Comput.”
Edge AI: On-Demand Accelerating Deep Neural Network Inference via Edge Computing (2019, background): “Hardware acceleration generally embraces basic DNN computation operations in hardware level design [28]–[30] while some works aim at optimizing the utilization of the existing hardware resources [31]–[33].”
Channel Gating Neural Networks (2018, methodology): “In this paper, we propose a novel approach to reduce CNN computation, called channel gating, which dynamically prunes the unnecessary computation specific to a particular image.”
Timeloop: A Systematic Approach to DNN Accelerator Evaluation (2019, background): “To provide high performance and energy efficiency to this class of compute-intensive algorithms, many DNN hardware accelerators have been proposed [1], [2], [5], [6], [12], [13], [15], [18], [25], [29], [33], [35], [39].”
Cognitive computation and communication: A complement solution to cloud for IoT (2016, background): “low-voltage storage [41] is also an attractive option for aggressively lowering energy consumption.”
What Do Compressed Deep Neural Networks Forget (2019, background): “in top-1 test set accuracy. These more compact networks are frequently favored in resource constrained settings; pruned models require less memory, energy consumption and have lower inference latency [59, 6, 65, 39, 67]. The ability to prune networks with seemingly so little degradation to generaliz”
Google Workloads for Consumer Devices: Mitigating Data Movement Bottlenecks (2018, background): “Several works [19, 36, 61, 81, 119] focus on accelerating inference.”
RecNMP: Accelerating Personalized Recommendation with Near-Memory Processing (2019, background): “highly-regular computational patterns found in fully-connected (FC), convolution (CNN), and recurrent (RNN) neural networks [5], [6], [10], [15], [18], [24], [27], [34], [39], [42], [48], [56], [57], [59]–[61], [63]–[68], [70], [72], [73], [75]. Unlike CNNs and RNNs, Washington University in St. Lou”
Scalpel: Customizing DNN pruning to the underlying hardware parallelism (2017, neutral): “For hardware-based methods, different ASIC designs [2, 5, 7, 8, 10, 14, 18, 25, 27, 33, 36, 37] and tools for accelerating DNN on existing platform have been proposed.”
YodaNN: An Architecture for Ultralow Power Binary-Weight CNN Acceleration (2016, methodology): “In this work we focus on implementing a CNN inference accelerator for neural networks supporting per-channel scaling and biasing, and implementing binary weights and fixedpoint activation.”
Gist: Efficient Data Encoding for Deep Neural Network Training (2018, methodology): “Many HW accelerators have been designed employing limited precision and leveraging sparsity to reduce computational and memory requirements [17], [5], [6], [35], [29], [38], [47], [2].”