Seda Ogrenci-Memik

Bio: Seda Ogrenci-Memik is an academic researcher from Northwestern University. The author has contributed to research in topics: Content-addressable memory & Pattern recognition (psychology). The author has an hindex of 8, co-authored 26 publications receiving 252 citations.

Machine Learning-Based Temperature Prediction for Runtime Thermal Management Across System Components

Abstract: Elevated temperatures limit the peak performance of systems because of frequent interventions by thermal throttling. Non-uniform thermal states across system nodes also cause performance variation within seemingly equivalent nodes leading to significant degradation of overall performance. In this paper we present a framework for creating a lightweight thermal prediction system suitable for run-time management decisions. We pursue two avenues to explore optimized lightweight thermal predictors. First, we use feature selection algorithms to improve the performance of previously designed machine learning methods. Second, we develop alternative methods using neural network and linear regression-based methods to perform a comprehensive comparative study of prediction methods. We show that our optimized models achieve improved performance with better prediction accuracy and lower overhead as compared with the Gaussian process model proposed previously. Specifically we present a reduced version of the Gaussian process model, a neural network–based model, and a linear regression–based model. Using the optimization methods, we are able to reduce the average prediction errors in the Gaussian process from $4.2^\circ$ C to $2.9^\circ$ C. We also show that the newly developed models using neural network and Lasso linear regression have average prediction errors of $2.9^\circ$ C and $3.8^\circ$ C respectively. The prediction overheads are 0.22, 0.097, and 0.026 ms per prediction for reduced Gaussian process, neural network, and Lasso linear regression models, respectively, compared with 0.57 ms per prediction for the previous Gaussian process model. We have implemented our proposed thermal prediction models on a two-node system configuration to help identify the optimal task placement. The task placement identified by the models reduces the average system temperature by up to $11.9^\circ$ C without any performance degradation. Furthermore, these models respectively achieve 75, 82.5, and 74.17 percent success rates in correctly pointing to those task placements with better thermal response, compared with 72.5 percent success for the original model in achieving the same objective. Finally, we extended our analysis to a 16-node system and we were able to train models and execute them in real time to guide task migration and achieve on average 17 percent reduction in the overall system cooling power.

User-specific skin temperature-aware DVFS for smartphones

Abstract: Skin temperature of mobile devices intimately affects the user experience. Power management schemes built into smartphones can lead to quickly crossing a user's threshold of tolerable skin temperature. Furthermore, there is a significant variation among users in terms of their sensitivity. Hence, controlling the skin temperature as part of the device's power management scheme is paramount. To achieve this, we first present a method for estimating skin and screen temperature at run-time using a combination of available on-device thermal sensors and performance indicators. In an Android-based smartphone, we achieve 99.05% and 99.14% accuracy in estimations of back cover and screen temperatures, respectively. Leveraging this run-time predictor, we develop User-specific Skin Temperature-Aware (USTA) DVFS mechanism to control the skin temperature. Performance of USTA is tested both with benchmarks and user tests comparing USTA to the standard Android governor. The results show that more users prefer to use USTA as opposed to the default DVFS mechanism.

A Reconfigurable Neural Network ASIC for Detector Front-End Data Compression at the HL-LHC

Abstract: Despite advances in the programmable logic capabilities of modern trigger systems, a significant bottleneck remains in the amount of data to be transported from the detector to off-detector logic where trigger decisions are made. We demonstrate that a neural network (NN) autoencoder model can be implemented in a radiation-tolerant application-specific integrated circuit (ASIC) to perform lossy data compression alleviating the data transmission problem while preserving critical information of the detector energy profile. For our application, we consider the high-granularity calorimeter from the Compact Muon Solenoid (CMS) experiment at the CERN Large Hadron Collider. The advantage of the machine learning approach is in the flexibility and configurability of the algorithm. By changing the NN weights, a unique data compression algorithm can be deployed for each sensor in different detector regions and changing detector or collider conditions. To meet area, performance, and power constraints, we perform quantization-aware training to create an optimized NN hardware implementation. The design is achieved through the use of high-level synthesis tools and the hls4ml framework and was processed through synthesis and physical layout flows based on a low-power (LP)-CMOS 65-nm technology node. The flow anticipates 200 Mrad of ionizing radiation to select gates and reports a total area of 3.6 mm2 and consumes 95 mW of power. The simulated energy consumption per inference is 2.4 nJ. This is the first radiation-tolerant on-detector ASIC implementation of an NN that has been designed for particle physics applications.

b-HiVE: a bit-level history-based error model with value correlation for voltage-scaled integer and floating point units

Abstract: Existing timing error models for voltage-scaled functional units ignore the effect of history and correlation among outputs, and the variation in the error behavior at different bit locations. We propose b-HiVE, a model for voltage-scaling-induced timing errors that incorporates these attributes and demonstrates their impact on the overall model accuracy. On average across several operations, b-HiVE's estimation is within 1--3% of comprehensive analog simulations, which corresponds to 5--17x higher accuracy (6--10x on average) than error models currently used in approximate computing research. To the best of our knowledge, we present the first bit-level error models of arithmetic units, and the first error models for voltage scaling of bitwise logic operations and floating-point units.

Monitoring of Vital Signs with Flexible and Wearable Medical Devices

Abstract: Advances in wireless technologies, low-power electronics, the internet of things, and in the domain of connected health are driving innovations in wearable medical devices at a tremendous pace. Wearable sensor systems composed of flexible and stretchable materials have the potential to better interface to the human skin, whereas silicon-based electronics are extremely efficient in sensor data processing and transmission. Therefore, flexible and stretchable sensors combined with low-power silicon-based electronics are a viable and efficient approach for medical monitoring. Flexible medical devices designed for monitoring human vital signs, such as body temperature, heart rate, respiration rate, blood pressure, pulse oxygenation, and blood glucose have applications in both fitness monitoring and medical diagnostics. As a review of the latest development in flexible and wearable human vitals sensors, the essential components required for vitals sensors are outlined and discussed here, including the reported sensor systems, sensing mechanisms, sensor fabrication, power, and data processing requirements.

Machine Learning-Based Temperature Prediction for Runtime Thermal Management Across System Components

Abstract: Elevated temperatures limit the peak performance of systems because of frequent interventions by thermal throttling. Non-uniform thermal states across system nodes also cause performance variation within seemingly equivalent nodes leading to significant degradation of overall performance. In this paper we present a framework for creating a lightweight thermal prediction system suitable for run-time management decisions. We pursue two avenues to explore optimized lightweight thermal predictors. First, we use feature selection algorithms to improve the performance of previously designed machine learning methods. Second, we develop alternative methods using neural network and linear regression-based methods to perform a comprehensive comparative study of prediction methods. We show that our optimized models achieve improved performance with better prediction accuracy and lower overhead as compared with the Gaussian process model proposed previously. Specifically we present a reduced version of the Gaussian process model, a neural network–based model, and a linear regression–based model. Using the optimization methods, we are able to reduce the average prediction errors in the Gaussian process from $4.2^\circ$ C to $2.9^\circ$ C. We also show that the newly developed models using neural network and Lasso linear regression have average prediction errors of $2.9^\circ$ C and $3.8^\circ$ C respectively. The prediction overheads are 0.22, 0.097, and 0.026 ms per prediction for reduced Gaussian process, neural network, and Lasso linear regression models, respectively, compared with 0.57 ms per prediction for the previous Gaussian process model. We have implemented our proposed thermal prediction models on a two-node system configuration to help identify the optimal task placement. The task placement identified by the models reduces the average system temperature by up to $11.9^\circ$ C without any performance degradation. Furthermore, these models respectively achieve 75, 82.5, and 74.17 percent success rates in correctly pointing to those task placements with better thermal response, compared with 72.5 percent success for the original model in achieving the same objective. Finally, we extended our analysis to a 16-node system and we were able to train models and execute them in real time to guide task migration and achieve on average 17 percent reduction in the overall system cooling power.

A score-based classification method for identifying hardware-trojans at gate-level netlists

Abstract: Recently, digital ICs are often designed by outside vendors to reduce design costs in semiconductor industry, which may introduce severe risks that malicious attackers implement Hardware Trojans (HTs) on them. Since IC design phase generates only a single design result, an RT-level or gate-level netlist for example, we cannot assume an HT-free netlist or a Golden netlist and then it is too difficult to identify whether a generated netlist is HT-free or HT-inserted. In this paper, we propose a score-based classification method for identifying HT-free or HT-inserted gate-level netlists without using a Golden netlist. Our proposed method does not directly detect HTs themselves in a gate-level netlist but a net included in HTs, which is called Trojan net, instead. Firstly, we observe Trojan nets from several HT-inserted benchmarks and extract several their features. Secondly, we give scores to extracted Trojan net features and sum up them for each net in benchmarks. Then we can find out a score threshold to classify HT-free and HT-inserted netlists. Based on these scores, we can successfully classify HT-free and HT-inserted netlists in all the Trust-HUB gate-level benchmarks. Experimental results demonstrate that our method successfully identify all the HT-inserted gate-level benchmarks to be “HT-inserted” and all the HT-free gate-level benchmarks to be “HT-free” in approximately three hours for each benchmark.