| 291. |
|
|
| 292. |
|
|
| 293. |
|
|
| 294. |
|
|
| 295. |
- Saqib, Eiraj, et al.
(författare)
-
Optimizing the IoT Performance : A Case Study on Pruning a Distributed CNN
- 2023
-
Ingår i: 2023 IEEE Sensors Applications Symposium (SAS). - 9798350323078
-
Konferensbidrag (refereegranskat)abstract
- Implementing Convolutional Neural Networks (CNN) based computer vision algorithms in Internet of Things (IoT) sensor nodes can be difficult due to strict computational, memory, and latency constraints. To address these challenges, researchers have utilized techniques such as quantization, pruning, and model partitioning. Partitioning the CNN reduces the computational burden on an individual node, but the overall system computational load remains constant. Additionally, communication energy is also incurred. To understand the effect of partitioning and pruning on energy and latency, we conducted a case study using a feet detection application realized with Tiny Yolo-v3 on a 12th Gen Intel CPU with NVIDIA GeForce RTX 3090 GPU. After partitioning the CNN between the sequential layers, we apply quantization, pruning, and compression and study the effects on energy and latency. We analyze the extent to which computational tasks, data, and latency can be reduced while maintaining a high level of accuracy. After achieving this reduction, we offloaded the remaining partitioned model to the edge node. We found that over 90% computation reduction and over 99% data transmission reduction are possible while maintaining mean average precision above 95%. This results in up to 17x energy savings and up to 5.2x performance speed-up.
|
|
| 296. |
-
Scalable multi-core architectures : Design methodologies and tools
- 2012
-
Samlingsverk (redaktörskap) (övrigt vetenskapligt/konstnärligt)abstract
- As Moore's law continues to unfold, two important trends have recently emerged. First, the growth of chip capacity is translated into a corresponding increase of number of cores. Second, the parallelization of the computation and 3D integration technologies lead to distributed memory architectures. This book describes recent research that addresses urgent challenges in many-core architectures and application mapping. It addresses the architectural design of many core chips, memory and data management, power management, design and programming methodologies. It also describes how new techniques have been applied in various industrial case studies.
|
|
| 297. |
|
|
| 298. |
- Seceleanu, Tiberiu, et al.
(författare)
-
Communicating with Synchronized Environments
- 2006
-
Ingår i: Proceedings of the Sixth International Conference on Application of Concurrency to System Design. - 0769525563 - 9780769525563 ; , s. 15-24
-
Konferensbidrag (refereegranskat)abstract
- In the modern design environments, different modules, available in existent libraries, may obey different architectural styles and execution models. Reaching a well-behaved composition of such modules is a very important task of the system designer. In the framework of the action systems formalism, we analyze the co-existence of two models of execution, one synchronized, the other, interleaved. We devise a communication scheme, similar to the classical paradigm of polling, which allows us to model synchronized components that correctly exchange information, within the borders of a global system, with their non-synchronized partners. Derivations of such mechanisms follow specific correctness rules for refinement. We illustrate our methods on an audio system example, implementable as either a software or a hardware device.
|
|
| 299. |
- Seceleanu, Tiberiu, et al.
(författare)
-
Modeling Communication with Synchronized Environments
- 2008
-
Ingår i: Fundamenta Informaticae. - 0169-2968 .- 1875-8681. ; 86:3, s. 343-369
-
Tidskriftsartikel (refereegranskat)abstract
- A deterministic behavior of systems composed of several modules is a desirable design goal. Assembling a complex system from components requires also a high degree of re-usability. The compatibility of the selected components may become a problem even at abstract design levels, due to possible different degrees of model determinacy, possible different execution models, etc. In this cases, an overall deterministic system behavior is difficult to achieve. The development of communication mechanisms between such components will have then to accommodate the differences, so that both correct processing and information exchange (data and control, appropriate choices and relative timing or sequencing) are achieved. For instance, human-machine interaction offers a good example of cooperation between deterministic models (machines) communicating with highly non-deterministic counterparts (the human models, if not restricted). We analyze here such communication mechanisms by "confronting" synchronized and un-synchronized models of execution, in the framework of action systems, a state based formalism. We "force" the two models to coexist within the same context and explore the possibilities of building trustworthy communication channels between them. We base our approach on a combined polling - interrupt scheme, which allows us to mitigate communication issues that may otherwise compromise the correct input-output system behavior. More robust system models are obtained by applying specific correctness rules of refinement. We illustrate our methods on an audio system example, implementable as either a software or a hardware device.
|
|
| 300. |
|
|
