Compute-in-memory (CiM) is a promising approach to improving the computing speed and energy efficiency in dataintensive applications. Beyond existing CiM techniques of bitwise logic-in-memory operations and dot product operations, this paper extends the CiM paradigm with FAST, a new shift-based inmemory computation technique to handle high-concurrency operations on multiple rows in an SRAM. Such high-concurrency operations are widely seen in both conventional applications (e.g. the table update in a database), and emerging applications (e.g. the parallel weight update in neural network accelerators), in which low latency and low energy consumption are critical. The proposed shift-based CiM architecture is enabled by integrating the shifter function into each SRAM cell, and by creating a datapath that exploits the high-parallelism of shifting operations in multiple rows in the array. A 128-row 16-column shiftable SRAM in 65nm CMOS is designed to evaluate the proposed architecture. Postlayout SPICE simulations show average improvements of 4.4x energy efficiency and 96.0x speed over a conventional fully-digital memory-computing-separated scheme, when performing the 8-bit weight update task in a VGG-7 framework.
相關內容
During the last decade, hyperspectral images have attracted increasing interest from researchers worldwide. They provide more detailed information about an observed area and allow an accurate target detection and precise discrimination of objects compared to classical RGB and multispectral images. Despite the great potentialities of hyperspectral technology, the analysis and exploitation of the large volume data remain a challenging task. The existence of irrelevant redundant and noisy images decreases the classification accuracy. As a result, dimensionality reduction is a mandatory step in order to select a minimal and effective images subset. In this paper, a new filter approach normalized mutual synergy (NMS) is proposed in order to detect relevant bands that are complementary in the class prediction better than the original hyperspectral cube data. The algorithm consists of two steps: images selection through normalized synergy information and pixel classification. The proposed approach measures the discriminative power of the selected bands based on a combination of their maximal normalized synergic information, minimum redundancy and maximal mutual information with the ground truth. A comparative study using the support vector machine (SVM) and k-nearest neighbor (KNN) classifiers is conducted to evaluate the proposed approach compared to the state of art band selection methods. Experimental results on three benchmark hyperspectral images proposed by the NASA "Aviris Indiana Pine", "Salinas" and "Pavia University" demonstrated the robustness, effectiveness and the discriminative power of the proposed approach over the literature approaches. Keywords: Hyperspectral images; target detection; pixel classification; dimensionality reduction; band selection; information theory; mutual information; normalized synergy
Parametric BSDFs (Bidirectional Scattering Distribution Functions) are pervasively used because of their flexibility to represent a large variety of material appearances by simply tuning the parameters. While efficient evaluation of parametric BSDFs has been well-studied, high-quality importance sampling techniques for parametric BSDFs are still scarce. Existing sampling strategies either heavily rely on approximations and result in high variance, or solely perform sampling on a portion of the whole BSDF slice. Moreover, many of the sampling approaches are specifically paired with certain types of BSDFs. In this paper, we seek an efficient and general way for importance sampling parametric BSDFs. We notice that the nature of importance sampling is the mapping between a uniform distribution and the target distribution. Specifically, when BSDF parameters are given, the mapping that performs importance sampling on a BSDF slice can be simply recorded as a 2D image that we name as importance map. Following this observation, we accurately precompute the importance maps using a mathematical tool named optimal transport. Then we propose a lightweight neural network to efficiently compress the precomputed importance maps. In this way, we have completely brought parametric BSDF importance sampling to the precomputation stage, avoiding heavy runtime computation. Since this process is similar to light baking where a set of images are precomputed, we name our method importance baking. Together with a BSDF evaluation network and a PDF (probability density function) query network, our method enables full MIS without any revision to the rendering pipeline. Our method essentially performs perfect importance sampling. Compared with previous methods, we demonstrate reduced noise levels on rendering results with a rich set of appearances, including both conductors and dielectrics with anisotropic roughness.
Recent advances in camera designs and imaging pipelines allow us to capture high-quality images using smartphones. However, due to the small size and lens limitations of the smartphone cameras, we commonly find artifacts or degradation in the processed images. The most common unpleasant effects are noise artifacts, diffraction artifacts, blur, and HDR overexposure. Deep learning methods for image restoration can successfully remove these artifacts. However, most approaches are not suitable for real-time applications on mobile devices due to their heavy computation and memory requirements. In this paper, we propose LPIENet, a lightweight network for perceptual image enhancement, with the focus on deploying it on smartphones. Our experiments show that, with much fewer parameters and operations, our model can deal with the mentioned artifacts and achieve competitive performance compared with state-of-the-art methods on standard benchmarks. Moreover, to prove the efficiency and reliability of our approach, we deployed the model directly on commercial smartphones and evaluated its performance. Our model can process 2K resolution images under 1 second in mid-level commercial smartphones.
One of the major challenges with cell-free (CF) massive multiple-input multiple-output (MIMO) networks is providing backhaul links for a large number of distributed access points (APs). In general, providing fiber optics backhaul for these APs is not cost-effective and also reduces network scalability. Wireless backhauling can be a promising solution that can be integrated with wireless access links to increase spectrum efficiency. In this paper, the application of integrated access and backhaul (IAB) technique in millimeter-wave (mmWave) CF massive MIMO systems is investigated. The access and backhaul links share a frequency spectrum in the mmWave bands, and in both, hybrid beamforming techniques are adopted for signal transmission. The bandwidth allocation (division) parameter between the two link types as well as the beamforming matrices are optimized to maximize the end-to-end data-rate. This leads to a non-convex optimization problem for which an efficient solution method is proposed. The simulation results show the effectiveness of the IAB technique and our proposed scheme in CF massive MIMO systems. These simulations also compare the proposed hybrid beamforming method with a fully digital solution in terms of the number of radio frequency (RF) chains and the volume of backhaul traffic. Finally, the effect of increasing the number of APs on the users data rates in terms of wireless access and backhaul links constraints is also examined.
Spiking Neural Networks (SNNs) have emerged as an attractive spatio-temporal computing paradigm for complex vision tasks. However, most existing works yield models that require many time steps and do not leverage the inherent temporal dynamics of spiking neural networks, even for sequential tasks. Motivated by this observation, we propose an \rev{optimized spiking long short-term memory networks (LSTM) training framework that involves a novel ANN-to-SNN conversion framework, followed by SNN training}. In particular, we propose novel activation functions in the source LSTM architecture and judiciously select a subset of them for conversion to integrate-and-fire (IF) activations with optimal bias shifts. Additionally, we derive the leaky-integrate-and-fire (LIF) activation functions converted from their non-spiking LSTM counterparts which justifies the need to jointly optimize the weights, threshold, and leak parameter. We also propose a pipelined parallel processing scheme which hides the SNN time steps, significantly improving system latency, especially for long sequences. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), in contrast to expensive multiply-and-accumulates (MAC) needed for ANNs, except for the input layer when using direct encoding, yielding significant improvements in energy efficiency. We evaluate our framework on sequential learning tasks including temporal MNIST, Google Speech Commands (GSC), and UCI Smartphone datasets on different LSTM architectures. We obtain test accuracy of 94.75% with only 2 time steps with direct encoding on the GSC dataset with 4.1x lower energy than an iso-architecture standard LSTM.
Applications with low data reuse and frequent irregular memory accesses, such as graph or sparse linear algebra workloads, fail to scale well due to memory bottlenecks and poor core utilization. While prior work with prefetching, decoupling, or pipelining can mitigate memory latency and improve core utilization, memory bottlenecks persist due to limited off-chip bandwidth. Approaches doing processing in-memory (PIM) with Hybrid Memory Cube (HMC) overcome bandwidth limitations but fail to achieve high core utilization due to poor task scheduling and synchronization overheads. Moreover, the high memory-per-core ratio available with HMC limits strong scaling. We introduce Dalorex, a hardware-software co-design that achieves high parallelism and energy efficiency, demonstrating strong scaling with >16,000 cores when processing graph and sparse linear algebra workloads. Over the prior work in PIM, both using 256 cores, Dalorex improves performance and energy consumption by two orders of magnitude through (1) a tile-based distributed-memory architecture where each processing tile holds an equal amount of data, and all memory operations are local; (2) a task-based parallel programming model where tasks are executed by the processing unit that is co-located with the target data; (3) a network design optimized for irregular traffic, where all communication is one-way, and messages do not contain routing metadata; (4) novel traffic-aware task scheduling hardware that maintains high core utilization; and (5) a data placement strategy that improves work balance. This work proposes architectural and software innovations to provide the greatest scalability to date for running graph algorithms while still being programmable for other domains.
Pre-trained language models (LMs) obtain state-of-the-art performance when adapted to text classification tasks. However, when using such models in real-world applications, efficiency considerations are paramount. In this paper, we study how different training procedures that adapt LMs to text classification perform, as we vary model and train set size. More specifically, we compare standard fine-tuning, prompting, and knowledge distillation (KD) when the teacher was trained with either fine-tuning or prompting. Our findings suggest that even though fine-tuning and prompting work well to train large LMs on large train sets, there are more efficient alternatives that can reduce compute or data cost. Interestingly, we find that prompting combined with KD can reduce compute and data cost at the same time.
Simulation-based Bayesian inference (SBI) can be used to estimate the parameters of complex mechanistic models given observed model outputs without requiring access to explicit likelihood evaluations. A prime example for the application of SBI in neuroscience involves estimating the parameters governing the response dynamics of Hodgkin-Huxley (HH) models from electrophysiological measurements, by inferring a posterior over the parameters that is consistent with a set of observations. To this end, many SBI methods employ a set of summary statistics or scientifically interpretable features to estimate a surrogate likelihood or posterior. However, currently, there is no way to identify how much each summary statistic or feature contributes to reducing posterior uncertainty. To address this challenge, one could simply compare the posteriors with and without a given feature included in the inference process. However, for large or nested feature sets, this would necessitate repeatedly estimating the posterior, which is computationally expensive or even prohibitive. Here, we provide a more efficient approach based on the SBI method neural likelihood estimation (NLE): We show that one can marginalize the trained surrogate likelihood post-hoc before inferring the posterior to assess the contribution of a feature. We demonstrate the usefulness of our method by identifying the most important features for inferring parameters of an example HH neuron model. Beyond neuroscience, our method is generally applicable to SBI workflows that rely on data features for inference used in other scientific fields.
The essence of multivariate sequential learning is all about how to extract dependencies in data. These data sets, such as hourly medical records in intensive care units and multi-frequency phonetic time series, often time exhibit not only strong serial dependencies in the individual components (the "marginal" memory) but also non-negligible memories in the cross-sectional dependencies (the "joint" memory). Because of the multivariate complexity in the evolution of the joint distribution that underlies the data generating process, we take a data-driven approach and construct a novel recurrent network architecture, termed Memory-Gated Recurrent Networks (mGRN), with gates explicitly regulating two distinct types of memories: the marginal memory and the joint memory. Through a combination of comprehensive simulation studies and empirical experiments on a range of public datasets, we show that our proposed mGRN architecture consistently outperforms state-of-the-art architectures targeting multivariate time series.
Recent advances in 3D fully convolutional networks (FCN) have made it feasible to produce dense voxel-wise predictions of volumetric images. In this work, we show that a multi-class 3D FCN trained on manually labeled CT scans of several anatomical structures (ranging from the large organs to thin vessels) can achieve competitive segmentation results, while avoiding the need for handcrafting features or training class-specific models. To this end, we propose a two-stage, coarse-to-fine approach that will first use a 3D FCN to roughly define a candidate region, which will then be used as input to a second 3D FCN. This reduces the number of voxels the second FCN has to classify to ~10% and allows it to focus on more detailed segmentation of the organs and vessels. We utilize training and validation sets consisting of 331 clinical CT images and test our models on a completely unseen data collection acquired at a different hospital that includes 150 CT scans, targeting three anatomical organs (liver, spleen, and pancreas). In challenging organs such as the pancreas, our cascaded approach improves the mean Dice score from 68.5 to 82.2%, achieving the highest reported average score on this dataset. We compare with a 2D FCN method on a separate dataset of 240 CT scans with 18 classes and achieve a significantly higher performance in small organs and vessels. Furthermore, we explore fine-tuning our models to different datasets. Our experiments illustrate the promise and robustness of current 3D FCN based semantic segmentation of medical images, achieving state-of-the-art results. Our code and trained models are available for download: //github.com/holgerroth/3Dunet_abdomen_cascade.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191