José Cano

José Cano

Auto-generated profile, based on publicly available information

José Cano

University of Glasgow, Glasgow, United Kingdom

Research Focus & Metrics

Academic Career Length

21Years (2006-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 November 2019
Verified
Conference Paper
Closed access
VRValentin Radu
Valentin Radu
KKKuba Kaszyk
Kuba Kaszyk
YWYuan Wen
Yuan Wen
JTJack Turner
Jack Turner
JCJosé Cano
José Cano
MOMichael O'BOyle
Michael O'BOyle
Convolutional Neural Networks (CNN) are becoming a common presence in many applications and services, due to their superior recognition accuracy. They are increasingly being used on mobile devices, many times just by porting large models designed for server space, although several model compression techniques have been considered. One model compression technique intended to reduce computations is channel pruning. Mobile and embedded systems now have GPUs which are ideal for the parallel computations of neural networks and for their lower energy cost per operation. Specialized libraries perform these neural network computations through highly optimized routines. As we find in our experiments, these libraries are optimized for the most common network shapes, making uninstructed channel pruning inefficient. We evaluate higher level libraries, which analyze the input characteristics of a convolutional layer, based on which they produce optimized OpenCL (Arm Compute Library and TVM) and CUDA (cuDNN) code. However, in reality, these characteristics and subsequent choices intended for optimization can have the opposite effect. We show that a reduction in the number of convolutional channels, pruning 12% of the initial size, is in some cases detrimental to performance, leading to 2× slowdown. On the other hand, we also find examples where performance-aware pruning achieves the intended results, with performance speedups of 3× with cuDNN and above 10× with Arm Compute Library and TVM. Our findings expose the need for hardware-instructed neural network pruning.

1 September 2018
Verified
Conference Paper
Closed access
JTJack Turner
Jack Turner
JCJose Cano
Jose Cano
VRValentin Radu
Valentin Radu
ECElliot J. Crowley
Elliot J. Crowley
MOMichael O'BOyle
Michael O'BOyle
ASAmos Storkey
Amos Storkey
Convolutional Neural Networks (CNNs) are extremely computationally demanding, presenting a large barrier to their deployment on resource-constrained devices. Since such systems are where some of their most useful applications lie (e.g. obstacle detection for mobile robots, vision-based medical assistive technology), significant bodies of work from both machine learning and systems communities have attempted to provide optimisations that will make CNNs available to edge devices. In this paper we unify the two viewpoints in a Deep Learning Inference Stack and take an across-stack approach by implementing and evaluating the most common neural network compression techniques (weight pruning, channel pruning, and quantisation) and optimising their parallel execution with a range of programming approaches (OpenMP, OpenCL) and hardware architectures (CPU, GPU). We provide comprehensive Pareto curves to instruct trade-offs under constraints of accuracy, execution time, and memory space.

7 September 2024 in Journal of Network and Computer Applications
Verified
Journal Article
Closed access
SJShivani Jamwal
Shivani Jamwal
JCJosé Cano
José Cano
GLGyu Myoung Lee
Gyu Myoung Lee
NTNguyen H. Tran
Nguyen H. Tran

1 May 2012
Verified
Conference Paper
Closed access
QYQiaoyan Yu
Qiaoyan Yu
JCJosé Cano
José Cano
JFJosé Flich
José Flich
PAPaul Ampadu
Paul Ampadu
High-end MPSoC systems with built-in high-radix topologies achieve good performance because of the improved connectivity and the reduced network diameter. In high-end MPSoC systems, fault tolerance support is becoming a compulsory feature. In this work, we propose a combined method to address permanent and transient link and router failures in those systems. The LBDRhr mechanism is proposed to tolerate permanent link failures in some popular high-radix topologies. The increased router complexity may lead to more transient router errors than routers using simple XY routing algorithm. We exploit the inherent information redundancy (IIR) in LBDRhr logic to manage transient errors in the network routers. Thorough analyses are provided to discover the appropriate internal nodes and the forbidden signal patterns for transient error detection. Simulation results show that LBDRhr logic can tolerate all of the permanent failure combinations of long-range links and 80% of links failures at short-range links. Case studies show that the error detection method based on the new IIR extraction method reduces the power consumption and the residual error rate by 33% and up to two orders of magnitude, respectively, compared to triple modular redundancy. The impact of network topologies on the efficiency of the detection mechanism has been examined in this work, as well.

21 March 2025 in ACM Transactions on Architecture and Code Optimization
Verified
Journal Article
Open access
PGPerry Gibson
Perry Gibson
JCJose Cano
Jose Cano
ECElliot Crowley
Elliot Crowley
ASAmos Storkey
Amos Storkey
MOMichael O’bOyle
Michael O’bOyle
Deep Neural Networks (DNNs) are very computationally demanding, which presents a significant barrier to their deployment, especially on resource-constrained devices. Significant work from both the machine learning and computing systems communities has attempted to accelerate DNNs. However, the number of techniques available and the required domain knowledge for their exploration continue to grow, making design space exploration (DSE) increasingly difficult. To unify the perspectives from these two communities, this article introduces the Deep Learning Acceleration Stack (DLAS), a conceptual model for DNN deployment and acceleration. We adopt a six-layer representation that organizes and illustrates the key areas for DNN acceleration, from machine learning to software and computer architecture. We argue that the DLAS model balances simplicity and expressiveness, assisting practitioners from various domains in tackling co-design acceleration challenges. We demonstrate the interdependence of the DLAS layers, and thus the need for co-design, through an across-stack perturbation study, using a modified tensor compiler to generate experiments for combinations of a few parameters across the DLAS layers. Our perturbation study assesses the impact on inference time and accuracy when varying DLAS parameters across two datasets, seven popular DNN architectures, four compression techniques, three algorithmic primitives (with sparse and dense variants), untuned and auto-scheduled code generation, and four hardware platforms. The study observes significant changes in the relative performance of design choices with the introduction of new DLAS parameters (e.g., the fastest algorithmic primitive varies with the level of quantization). Given the strong evidence for the need for co-design, and the high costs of DSE, DLAS offers a valuable conceptual model for better exploring advanced co-designed accelerated deep learning solutions.

22 March 2021 in IEEE Transactions on Parallel and Distributed Systems
Verified
Journal Article
Closed access
SDShi Dong
Shi Dong
YSYifan Sun
Yifan Sun
NANicolas Bohm Agostini
Nicolas Bohm Agostini
EKElmira Karimi
Elmira Karimi
DLDaniel Lowell
Daniel Lowell
David KaeliDavid Kaeli
David Kaeli
Deep Neural Networks (DNNs) have emerged as an important class of machine learning algorithms, providing accurate solutions to a broad range of applications. Sparsity in activation maps in DNN training presents an opportunity to reduce computations. However, exploiting activation sparsity presents two major challenges: i) profiling activation sparsity during training comes with significant overhead due to computing the degree of sparsity and the data movement; ii) the dynamic nature of activation maps requires dynamic dense-to-sparse conversion during training, leading to significant overhead. In this article, we present Spartan, a lightweight hardware/software framework to accelerate DNN training on a GPU. Spartan provides a cost-effective and programmer-transparent microarchitectural solution to exploit activation sparsity detected during training. Spartan provides an efficient sparsity monitor, a tile-based sparse GEMM algorithm, and a novel compaction engine designed for GPU workloads. Spartan can reduce sparsity profiling overhead by 52.5× on average. For the most compute-intensive layers, i.e., convolutional layers, we can speedup AlexNet by 3.4×, VGGNet-16 by 2.14×, and ResNet-18 by 2.02×, when training on the ImageNet dataset.