Searcharxiv⌕ Search

arXiv subjects

Akash Kumar

Publications and source records attributed to Akash Kumar.

At least 109 records · Page 6Linked to original sources

dCSR: A Memory-Efficient Sparse Matrix Representation for Parallel Neural Network Inference

Reducing the memory footprint of neural networks is a crucial prerequisite for deploying them in small and low-cost embedded devices. Network parameters can often be reduced significantly through pruning. We discuss how to best represent the indexing overhead of sparse networks for the coming generation of Single Instruction, Multiple Data (SIMD)-capable microcontrollers. From this, we develop Delta-Compressed Storage Row (dCSR), a storage format for sparse matrices that allows for both low overhead storage and fast inference on embedded systems with wide SIMD units. We demonstrate our method on an ARM Cortex-M55 MCU prototype with M-Profile Vector Extension(MVE). A comparison of memory consumption and throughput shows that our method achieves competitive compression ratios and increases throughput over dense methods by up to $2.9 \times$ for sparse matrix-vector multiplication (SpMV)-based kernels and $1.06 \times$ for sparse matrix-matrix multiplication (SpMM). This is accomplished through handling the generation of index information directly in the SIMD unit, leading to an increase in effective memory bandwidth.

cs.DS↗

Ultrathin ferrimagnetic GdFeCo films with very low damping

Ferromagnetic materials dominate as the magnetically active element in spintronic devices, but come with drawbacks such as large stray fields, and low operational frequencies. Compensated ferrimagnets provide an alternative as they combine the ultrafast magnetization dynamics of antiferromagnets with a ferromagnet-like spin-orbit-torque (SOT) behavior. However to use ferrimagnets in spintronic devices their advantageous properties must be retained also in ultrathin films (t < 10 nm). In this study, ferrimagnetic Gdx(Fe87.5Co12.5)1-x thin films in the thickness range t = 2-20 nm were grown on high resistance Si(100) substrates and studied using broadband ferromagnetic resonance measurements at room temperature. By tuning their stoichiometry, a nearly compensated behavior is observed in 2 nm Gdx(Fe87.5Co12.5)1-x ultrathin films for the first time, with an effective magnetization of Meff = 0.02 T and a low effective Gilbert damping constant of α = 0.0078, comparable to the lowest values reported so far in 30 nm films. These results show great promise for the development of ultrafast and energy efficient ferrimagnetic spintronic devices.

cond-mat.mtrl-sci↗

Energy-efficient W$_{\text{100-x}}$Ta$_{\text{x}}$/CoFeB/MgO spin Hall nano-oscillators

We investigate a W-Ta alloying route to reduce the auto-oscillation current densities and the power consumption of nano-constriction based spin Hall nano oscillators. Using spin-torque ferromagnetic resonance (ST-FMR) measurements on microbars of W$_{\text{100-x}}$Ta$_{\text{x}}$(5 nm)/CoFeB(t)/MgO stacks with t = 1.4, 1.8, and 2.0 nm, we measure a substantial improvement in both the spin-orbit torque efficiency and the spin Hall conductivity. We demonstrate a 34\% reduction in threshold auto-oscillation current density, which translates into a 64\% reduction in power consumption as compared to pure W-based SHNOs. Our work demonstrates the promising aspects of W-Ta alloying for the energy-efficient operation of emerging spintronic devices.

cond-mat.mes-hall↗

Fabrication of voltage gated spin Hall nano-oscillators

We demonstrate an optimized fabrication process for electric field (voltage gate) controlled nano-constriction spin Hall nano-oscillators (SHNOs), achieving feature sizes of <30 nm with easy to handle ma-N 2401 e-beam lithography negative tone resist. For the nanoscopic voltage gates, we utilize a two-step tilted ion beam etching approach and through-hole encapsulation using 30 nm HfO x . The optimized tilted etching process reduces sidewalls by 75% compared to no tilting. Moreover, the HfO x encapsulation avoids any sidewall shunting and improves gate breakdown. Our experimental results on W/CoFeB/MgO/SiO 2 SHNOs show significant frequency tunability (6 MHz/V) even for moderate perpendicular magnetic anisotropy. Circular patterns with diameter of 45 nm are achieved with an aspect ratio better than 0.85 for 80% of the population. The optimized fabrication process allows incorporating a large number of individual gates to interface to SHNO arrays for unconventional computing and densely packed spintronic neural networks.

cond-mat.mes-hall↗

The complexity of testing all properties of planar graphs, and the role of isomorphism

Consider property testing on bounded degree graphs and let $\varepsilon>0$ denote the proximity parameter. A remarkable theorem of Newman-Sohler (SICOMP 2013) asserts that all properties of planar graphs (more generally hyperfinite) are testable with query complexity only depending on $\varepsilon$. Recent advances in testing minor-freeness have proven that all additive and monotone properties of planar graphs can be tested in $poly(\varepsilon^{-1})$ queries. Some properties falling outside this class, such as Hamiltonicity, also have a similar complexity for planar graphs. Motivated by these results, we ask: can all properties of planar graphs can be tested in $poly(\varepsilon^{-1})$ queries? Is there a uniform query complexity upper bound for all planar properties, and what is the "hardest" such property to test? We discover a surprisingly clean and optimal answer. Any property of bounded degree planar graphs can be tested in $\exp(O(\varepsilon^{-2}))$ queries. Moreover, there is a matching lower bound, up to constant factors in the exponent. The natural property of testing isomorphism to a fixed graph needs $\exp(Ω(\varepsilon^{-2}))$ queries, thereby showing that (up to polynomial dependencies) isomorphism to an explicit fixed graph is the hardest property of planar graphs. The upper bound is a straightforward adapation of the Newman-Sohler analysis that tracks dependencies on $\varepsilon$ carefully. The main technical contribution is the lower bound construction, which is achieved by a special family of planar graphs that are all mutually far from each other. We can also apply our techniques to get analogous results for bounded treewidth graphs. We prove that all properties of bounded treewidth graphs can be tested in $\exp(O(\varepsilon^{-1}\log \varepsilon^{-1}))$ queries. Moreover, testing isomorphism to a fixed forest requires $\exp(Ω(\varepsilon^{-1}))$ queries.

cs.DS↗

NMPO: Near-Memory Computing Profiling and Offloading

Real-world applications are now processing big-data sets, often bottlenecked by the data movement between the compute units and the main memory. Near-memory computing (NMC), a modern data-centric computational paradigm, can alleviate these bottlenecks, thereby improving the performance of applications. The lack of NMC system availability makes simulators the primary evaluation tool for performance estimation. However, simulators are usually time-consuming, and methods that can reduce this overhead would accelerate the early-stage design process of NMC systems. This work proposes Near-Memory computing Profiling and Offloading (NMPO), a high-level framework capable of predicting NMC offloading suitability employing an ensemble machine learning model. NMPO predicts NMC suitability with an accuracy of 85.6% and, compared to prior works, can reduce the prediction time by using hardware-dependent applications features by up to 3 order of magnitude.

cs.AR↗

Toward the Design of Fault-Tolerance- and Peak- Power-Aware Multi-Core Mixed-Criticality Systems

Mixed-Criticality (MC) systems have recently been devised to address the requirements of real-time systems in industrial applications, where the system runs tasks with different criticality levels on a single platform. In some workloads, a high-critically task might overrun and overload the system, or a fault can occur during the execution. However, these systems must be fault-tolerant and guarantee the correct execution of all high-criticality tasks by their deadlines to avoid catastrophic consequences, in any situation. Furthermore, in these MC systems, the peak power consumption of the system may increase, especially in an overload situation and exceed the processor Thermal Design Power (TDP) constraint. This may cause generating heat beyond the cooling capacity, resulting the system stop to avoid excessive heat and halting the processor. In this paper, we propose a technique for dependent dual-criticality tasks in fault-tolerant multi-core MC systems to manage peak power consumption and temperature. The technique develops a tree of possible task mapping and scheduling at design-time to cover all possible scenarios and reduce the low-criticality task drop rate in the high-criticality mode. At run-time, the system exploits the tree to select a proper schedule according to fault occurrences and criticality mode changes. Experimental results show that the average task schedulability is 74.14% on average for the proposed method, while the peak power consumption and maximum temperature are improved by 16.65% and 14.9 C on average, respectively, compared to a recent work. In addition, for a real-life application, our method reduces the peak power and maximum temperature by up to 20.06% and 5 C, respectively, compared to a state-of-the-art approach.

cs.DC↗

Random walks and forbidden minors III: poly(d/ε)-time partition oracles for minor-free graph classes

Consider the family of bounded degree graphs in any minor-closed family (such as planar graphs). Let d be the degree bound and n be the number of vertices of such a graph. Graphs in these classes have hyperfinite decompositions, where, for a sufficiently small \e > 0, one removes \edn edges to get connected components of size independent of n. An important tool for sublinear algorithms and property testing for such classes is the partition oracle, introduced by the seminal work of Hassidim-Kelner-Nguyen-Onak (FOCS 2009). A partition oracle is a local procedure that gives consistent access to a hyperfinite decomposition, without any preprocessing. Given a query vertex v, the partition oracle outputs the component containing v in time independent of n. All the answers are consistent with a single hyperfinite decomposition. The partition oracle of Hassidim et al. runs in time d^poly(d/\e) per query. They pose the open problem of whether poly(d/\e)-time partition oracles exist. Levi-Ron (ICALP 2013) give a refinement of the previous approach, to get a partition oracle that runs in time d^{\log(d/\e)-per query. In this paper, we resolve this open problem and give \poly(d/\e)-time partition oracles for bounded degree graphs in any minor-closed family. Unlike the previous line of work based on combinatorial methods, we employ techniques from spectral graph theory. We build on a recent spectral graph theoretical toolkit for minor-closed graph families, introduced by the authors to develop efficient property testers. A consequence of our result is a poly(d/\e)-query tester for any monotone and additive property of minor-closed families (such as bipartite planar graphs). Our result also gives poly(d/\e)-query algorithms for additive {\e}n-approximations for problems such as maximum matching, minimum vertex cover, maximum independent set, and minimum dominating set for these graph families.

cs.DS↗

Compiler Toolchains for Deep Learning Workloads on Embedded Platforms

As the usage of deep learning becomes increasingly popular in mobile and embedded solutions, it is necessary to convert the framework-specific network representations into executable code for these embedded platforms. This paper consists of two parts: The first section is made up of a survey and benchmark of the available open source deep learning compiler toolchains, which focus on the capabilities and performance of the individual solutions in regard to targeting embedded devices and microcontrollers that are combined with a dedicated accelerator in a heterogeneous fashion. The second part explores the implementation and evaluation of a compilation flow for such a heterogeneous device and reuses one of the existing toolchains to demonstrate the necessary steps for hardware developers that plan to build a software flow for their own hardware.

cs.PL↗

The Teaching Dimension of Kernel Perceptron

Algorithmic machine teaching has been studied under the linear setting where exact teaching is possible. However, little is known for teaching nonlinear learners. Here, we establish the sample complexity of teaching, aka teaching dimension, for kernelized perceptrons for different families of feature maps. As a warm-up, we show that the teaching complexity is $Θ(d)$ for the exact teaching of linear perceptrons in $\mathbb{R}^d$, and $Θ(d^k)$ for kernel perceptron with a polynomial kernel of order $k$. Furthermore, under certain smooth assumptions on the data distribution, we establish a rigorous bound on the complexity for approximately teaching a Gaussian kernel perceptron. We provide numerical examples of the optimal (approximate) teaching set under several canonical settings for linear, polynomial and Gaussian kernel perceptrons.

cs.LG↗

SeqL: Secure Scan-Locking for IP Protection

Existing logic-locking attacks are known to successfully decrypt functionally correct key of a locked combinational circuit. It is possible to extend these attacks to real-world Silicon-based Intellectual Properties (IPs, which are sequential circuits) through scan-chains by selectively initializing the combinational logic and analyzing the responses. In this paper, we propose SeqL, which achieves functional isolation and locks selective flip-flop functional-input/scan-output pairs, thus rendering the decrypted key functionally incorrect. We conduct a formal study of the scan-locking problem and demonstrate automating our proposed defense on any given IP. We show that SeqL hides functionally correct keys from the attacker, thereby increasing the likelihood of the decrypted key being functionally incorrect. When tested on pipelined combinational benchmarks (ISCAS,MCNC), sequential benchmarks (ITC) and a fully-fledged RISC-V CPU, SeqL gave 100% resilience to a broad range of state-of-the-art attacks including SAT[1], Double-DIP[2], HackTest[3], SMT[4], FALL[5], Shift-and-Leak[6] and Multi-cycle attacks[7].

cs.CR↗

Logic Synthesis Meets Machine Learning: Trading Exactness for Generalization

Logic synthesis is a fundamental step in hardware design whose goal is to find structural representations of Boolean functions while minimizing delay and area. If the function is completely-specified, the implementation accurately represents the function. If the function is incompletely-specified, the implementation has to be true only on the care set. While most of the algorithms in logic synthesis rely on SAT and Boolean methods to exactly implement the care set, we investigate learning in logic synthesis, attempting to trade exactness for generalization. This work is directly related to machine learning where the care set is the training set and the implementation is expected to generalize on a validation set. We present learning incompletely-specified functions based on the results of a competition conducted at IWLS 2020. The goal of the competition was to implement 100 functions given by a set of care minterms for training, while testing the implementation using a set of validation minterms sampled from the same function. We make this benchmark suite available and offer a detailed comparative analysis of the different approaches to learning

cs.LG↗

Deletion to Induced Matching

In the DELETION TO INDUCED MATCHING problem, we are given a graph $G$ on $n$ vertices, $m$ edges and a non-negative integer $k$ and asks whether there exists a set of vertices $S \subseteq V(G) $ such that $|S|\le k$ and the size of any connected component in $G-S$ is exactly 2. In this paper, we provide a fixed-parameter tractable (FPT) algorithm of running time $O^*(1.748^{k})$ for the DELETION TO INDUCED MATCHING problem using branch-and-reduce strategy and path decomposition. We also extend our work to the exact-exponential version of the problem.

cs.DS↗

Power-Aware Run-Time Scheduler for Mixed-Criticality Systems on Multi-Core Platform

In modern multi-core Mixed-Criticality (MC) systems, a rise in peak power consumption due to parallel execution of tasks with maximum frequency, specially in the overload situation, may lead to thermal issues, which may affect the reliability and timeliness of MC systems. Therefore, managing peak power consumption has become imperative in multi-core MC systems. In this regard, we propose an online peak power and thermal management heuristic for multi-core MC systems. This heuristic reduces the peak power consumption of the system as much as possible during runtime by exploiting dynamic slack and per-cluster Dynamic Voltage and Frequency Scaling (DVFS). Specifically, our approach examines multiple tasks ahead to determine the most appropriate one for slack assignment, that has the most impact on the system peak power and temperature. However, changing the frequency and selecting a proper task for slack assignment and a proper core for task re-mapping at runtime can be time-consuming and may cause deadline violation which is not admissible for high-criticality tasks. Therefore, we analyze and then optimize our run-time scheduler and evaluate it for various platforms. The proposed approach is experimentally validated on the ODROID-XU3 (DVFS-enabled heterogeneous multi-core platform) with various embedded real-time benchmarks. Results show that our heuristic achieves up to 5.25% reduction in system peak power and 20.33\% reduction in maximum temperature compared to an existing method while meeting deadline constraints in different criticality modes.

cs.DC↗

SIMDive: Approximate SIMD Soft Multiplier-Divider for FPGAs with Tunable Accuracy

The ever-increasing quest for data-level parallelism and variable precision in ubiquitous multimedia and Deep Neural Network (DNN) applications has motivated the use of Single Instruction, Multiple Data (SIMD) architectures. To alleviate energy as their main resource constraint, approximate computing has re-emerged,albeit mainly specialized for their Application-Specific Integrated Circuit (ASIC) implementations. This paper, presents for the first time, an SIMD architecture based on novel multiplier and divider with tunable accuracy, targeted for Field-Programmable Gate Arrays (FPGAs). The proposed hybrid architecture implements Mitchell's algorithms and supports precision variability from 8 to 32 bits. Experimental results obtained from Vivado, multimedia and DNN applications indicate superiority of proposed architecture (both SISD and SIMD) over accurate and state-of-the-art approximate counterparts. In particular, the proposed SISD divider outperforms the accurate Intellectual Property (IP) divider provided by Xilinx with 4x higher speed and 4.6x less energy and tolerating only < 0.8% error. Moreover, the proposed SIMD multiplier-divider supersede accurate SIMD multiplier by achieving up to 26%, 45%, 36%, and 56% improvement in area, throughput, power, and energy, respectively.

cs.AR↗

ExPAN(N)D: Exploring Posits for Efficient Artificial Neural Network Design in FPGA-based Systems

The recent advances in machine learning, in general, and Artificial Neural Networks (ANN), in particular, has made smart embedded systems an attractive option for a larger number of application areas. However, the high computational complexity, memory footprints, and energy requirements of machine learning models hinder their deployment on resource-constrained embedded systems. Most state-of-the-art works have considered this problem by proposing various low bit-width data representation schemes, optimized arithmetic operators' implementations, and different complexity reduction techniques such as network pruning. To further elevate the implementation gains offered by these individual techniques, there is a need to cross-examine and combine these techniques' unique features. This paper presents ExPAN(N)D, a framework to analyze and ingather the efficacy of the Posit number representation scheme and the efficiency of fixed-point arithmetic implementations for ANNs. The Posit scheme offers a better dynamic range and higher precision for various applications than IEEE $754$ single-precision floating-point format. However, due to the dynamic nature of the various fields of the Posit scheme, the corresponding arithmetic circuits have higher critical path delay and resource requirements than the single-precision-based arithmetic units. Towards this end, we propose a novel Posit to fixed-point converter for enabling high-performance and energy-efficient hardware implementations for ANNs with minimal drop in the output accuracy. We also propose a modified Posit-based representation to store the trained parameters of a network. Compared to an $8$-bit fixed-point-based inference accelerator, our proposed implementation offers $\approx46\%$ and $\approx18\%$ reductions in the storage requirements of the parameters and energy consumption of the MAC units, respectively.

cs.AR↗

Average-case Complexity of Teaching Convex Polytopes via Halfspace Queries

We examine the task of locating a target region among those induced by intersections of $n$ halfspaces in $\mathbb{R}^d$. This generic task connects to fundamental machine learning problems, such as training a perceptron and learning a $ϕ$-separable dichotomy. We investigate the average teaching complexity of the task, i.e., the minimal number of samples (halfspace queries) required by a teacher to help a version-space learner in locating a randomly selected target. As our main result, we show that the average-case teaching complexity is $Θ(d)$, which is in sharp contrast to the worst-case teaching complexity of $Θ(n)$. If instead, we consider the average-case learning complexity, the bounds have a dependency on $n$ as $Θ(n)$ for \tt{i.i.d.} queries and $Θ(d \log(n))$ for actively chosen queries by the learner. Our proof techniques are based on novel insights from computational geometry, which allow us to count the number of convex polytopes and faces in a Euclidean space depending on the arrangement of halfspaces. Our insights allow us to establish a tight bound on the average-case complexity for $ϕ$-separable dichotomies, which generalizes the known $\mathcal{O}(d)$ bound on the average number of "extreme patterns" in the classical computational geometry literature (Cover, 1965).

cs.LG↗

ParaLarH: Parallel FPGA Router based upon Lagrange Heuristics

Routing of the nets in Field Programmable Gate Array (FPGA) design flow is one of the most time consuming steps. Although Versatile Place and Route (VPR), which is a commonly used algorithm for this purpose, routes effectively, it is slow in execution. One way to accelerate this design flow is to use parallelization. Since VPR is intrinsically sequential, a set of parallel algorithms have been recently proposed for this purpose (ParaLaR and ParaLarPD). These algorithms formulate the routing process as a Linear Program (LP) and solve it using the Lagrange relaxation, the sub-gradient method, and the Steiner tree algorithm. Out of the many metrics available to check the effectiveness of routing, ParaLarPD, which is an improved version of ParaLaR, suffers from large violations in the constraints of the LP problem (which is related to the minimum channel width metric) as well as an easily measurable critical path delay metric that can be improved further. In this paper, we introduce a set of novel Lagrange heuristics that improve the Lagrange relaxation process. When tested on the MCNC benchmark circuits, on an average, this leads to halving of the constraints violation, up to 10% improvement in the minimum channel width, and up to 8% reduction in the critical path delay as obtained from ParaLarPD. We term our new algorithm as ParaLarH. Due to the increased work in the Lagrange relaxation process, as compared to ParaLarPD, ParaLarH does slightly deteriorate the speedup obtained because of parallelization, however, this aspect is easily compensated by using more number of threads.

cs.DC↗