The EnyAC research group develops tools and methods for energy and resource aware computing encompassing the entire system stack, from algorithms and applications to system and hardware architecture. Since its inception in 2000, the EnyAC research group has worked on novel computing paradigms and tools for energy, variability and reliability aware computing, all aiming for sustainable computing and computing for sustainability. Currently, the focus of EnyAC research group is on energy aware machine learning, hardware-machine learning model co-design, and hardware efficient data intensive applications. In recent past, EnyAC members have also worked on modeling, analysis, and optimization of social and life science applications, including river network modeling for renewable energy generation and hardware emulation of cell signaling networks by computer-aided design means.
We develop accurate, platform‐specific power and runtime models for machine learning (ML) models and new hardware-aware ML model (co-)design methodologies that allow machine learners and hardware designers to identify the most accurate model configuration that satisfies given hardware constraints.
Our work relies on statistical learning approaches for managing power and other resources in large scale computing systems and implementing computational kernels for on-chip learning in an energy efficient manner.
We have developed scalable and efficient approaches for modeling and analysis of river networks in the context of small footprint power generation and distribution. In the field of systems biology, we have demonstrated the effectiveness of hardware emulation for accelerating analysis of large cell signaling networks.
Ahmet's work on "DeepNVM++: Cross-Layer Modeling and Optimization Framework of Non-Volatile Memories for Deep Learning" is out on arXiv!
Ahmet's work on "The Architectural Implications of Distributed Reinforcement Learning on CPU-GPU Systems" has been accepted to the Energy Efficient Machine Learning and Cognitive Computing (EMC2) workshop for oral presentation! Thanks to our great collaborators at NVIDIA and Diana Marculescu!
Rudy's work on "One Weight Bitwidth to Rule Them All" has been accepted at Embedded Vision Workshop at ECCV 2020 as the best paper! Congratulations!
Rudy's work on "Improving the Adversarial Robustness of Transfer Learning via Noisy Feature Distillation" has been accepted at AdvML workshop at KDD 2020! Congratulations!
Rudy's work on "Pareto-aware Channel Optimization for Slimmable Neural Networks" has been accepted at RealML and DMMLSys workshops at ICML 2020! Congratulations!
Rudy and Ahmet's proposal were selected as finalists for Qualcomm Innovation Fellowship 2020! Congratulations!
Rudy and Ruizhou's work on ``Towards Efficient Model Compression'' has been accepted at CVPR 2020 as oral presentation! Congratulations!
Zhuo and Ruizhou's work on ``ViP: Virtual Pooling for Accelerating CNN-based Image Classification and Object Detection'' has been accepted at WACV 2020! Congratulations!
Ahmet's work on ``DeepNVM: A Framework for Modeling and Analysis of Non-Volatile Memory Technologies for Deep Learning Applications'' has been accepted at DATE 2020! Congratulations!
Dimirios and Ruizhou's work on ``Single-Path NAS: Designing Hardware-Efficient ConvNets in less than 4 Hours'' has been accepted at ECML-PKDD 2019! Congratulations!
Ruizhou and Rudy's work on ``Regularizing Activation Distribution for Training Binarized Deep Networks'' has been accepted at CVPR 2019! Congratulations!
Ruizhou and Rudy's work on ``Lightweight Quantized Deep Neural Networks for Fast and Accurate Inference'' has been accepted at DAC 2019! Congratulations!
Rudy and Ruizhou's work on ``AdaScale: Towards Real-time Video Object Detection Using Adaptive Scaling'' has been accepted at SysML 2019! Congratulations!
Congratulations to Ahmet, Rudy, and Ruizhou for winning Bob Lee Gregory, La Veerne Owen-Barakat, and Liang Ji-Dian Fellowships, respectively.
Ting-Wu (Rudy) Chin’s work on layer-compensated pruning that leverages meta-learning for pruning got accepted at the NeurIPS 2018 Workshop on Machine Learning on the Phone and other Consumer Devices (MLPCD 2) as oral contribution.
Ruizhou Ding and Ting-Wu (Rudy) Chin’s work on differentiable training for hardware efficient LightNNs got accepted at the NeurIPS 2018 Workshop on Compact Deep Neural Networks with industrial applications.
Zhuo Chen, Ruizhou Ding and Ting-Wu Chin’s work on understanding the impact of label granularity on CNN-based image classification got accepted at the IEEE International Conference on Data Mining (ICDM) Workshop on Data Science and Big Data Analytics (DSDA) 2018
EnyAC-ers Zhuo Chen and Dimitrios Stamoulis were awarded the 2018 Qualcomm Innovation Fellowship.
Dimitrios' work on Designing Adaptive Neural Networks for Energy-Constrained Image Classification got accepted at ICCAD 2018.
Ahmet Fatih Inci’s work “Solving the Non-Volatile Memory Conundrum for Deep Learning Workloads” got accepted at the Eight Workshop on Architectures and Systems for Big Data (ASBD) at International Symposium on Computer Architecture (ISCA) in June 2018!
Zhuo Chen, Ermao Cai, Ruizhou Ding, and Ting-Wu Chin took third place in the competition with their project, dubbed ParkAI.
Learn more about our recent effort on brining machine learning applications to your mobile devices.
Marculescu is the recipient of the Barbara Lazarus Award, which recognizes exemplary contributions to fostering a welcoming and nurturing environment for graduate students and young faculty at Carnegie Mellon.
While selecting the hyper-parameters of Neural Networks (NNs) has been so far treated as an art, the emergence of more complex, deeper architectures poses increasingly more challenges to designers and Machine Learning (ML) practitioners, especially when power and memory constraints need to be considered.
"How much energy is consumed for an inference made by a convolutional neural network (CNN)?" With the increased popularity of CNNs deployed on the wide-spectrum of platforms (from mobile devices to workstations), the answer to this question has drawn significant attention.
Application-specific integrated circuit (ASIC) implementations for Deep Neural Networks (DNNs) have been adopted in many systems because of their higher classification speed. However, although they may be characterized by better accuracy, larger DNNs require significant energy and area, thereby limiting their wide adoption.
Ph.D. positions are available immediately in the area of scalable power management and control for large scale computing systems, discrete modeling for dynamical systems, and hardware acceleration for social and life science applications. Highly motivated students with a strong theoretical background, especially in control theory, optimization algorithms, and modeling are encouraged to apply. Minimum qualifications include BS in CE/CS/EE/Applied Math or related fields; preferred qualifications include MS in CE/CS/EE/Applied Math or related fields. For qualified applicants holding a Ph.D., a postdoctoral appointment is possible. For additional details/questions, please contact
EnyAC-ers Ifigeneia Apostolopoulou and Ermao Cai received the Onassis Ph.D. Fellowship (Scholarship Programs for Hellenes, Onassis Foundation) and the CMU Presidential Fellowship respectively. Congratulations!!
EnyAC alumnus and NYU Assistant Professor Siddarth Garg was listed in Popular Science Magazine's annual list of the year’s most brilliant young scientists and engineers. Congratulations, Siddharth!!
Diana Marculescu has been named the next David Edward Schramm Professor of Electrical and Computer Engineering, Carnegie Mellon University.
EnyAC alumnus Guangshuo Liu and Professor Diana Marculescu co-authored an IEEE Transactions on Computers article selected as featured paper for the month of April 2016.
Ten girls from school districts in and around Pittsburgh met with Professor Diana Marculescu, and ECE Ph.D. students Ermao Cai and Zhuo Chen for a weekend activity part of the Tour Your Future Program. The students worked on a hands-on activity with electrical circuits that illustrated simple open or closed circuit operation with resistors and color LEDs.
Diana Marculescu has been named an IEEE Fellow for her contributions to design and optimization of energy aware computing systems.
Diana Marculescu has been named a 2013-2014 fellow in ELATE at Drexel, a national leadership development program designed to advance senior women faculty in academic engineering, computer science and related fields into effective institutional leadership roles within their schools and universities.
The Best Foundational Advance Prize is awarded at the World Championship Jamboree to the team that creates a biological system that could help enable the success of other systems created using synthetic biology. The CMU iGEM Team comprised of Yang Choo (E'14, BME and ChmeE), Eric Pederson (S'15, Biology), Jesse Salazar (E'13, BME and ECE), and Peter Wei (E'15, BME and ECE) was co-advised by Diana Marculescu, along with colleagues from Chemistry, BME, and CS Departments.
Intel Ph.D. fellow Da-Cheng Juan earned one of three best poster awards at the annual Intel Ph.D. Fellowship Forum, which brings together talented students from top academic research institutions around the globe. As part of the forum, the 2012 Intel Ph.D. fellows were recognized and shared their research during the poster session, where the fellows were judged based on overall presentation. The three posters receiving the most votes earned best poster honors.
Diana Marculescu was elected to the IT Honor Roll of IT History Society, for developing novel power management techniques to improve the performance delivered per unit of energy consumed for computer hardware and software.
Da-Cheng has been awarded a U.S. Corporate Intel Fellowship, a highly competitive and prestigious award reserved for Ph.D. candidates pursuing leading-edge research in fields related to Intel's business and research interests at select universities. Da-Cheng's research investigates how to apply machine learning for the performance modeling and optimization of chip-multiprocessors under physical constraints, such as power, temperature and process variations.
Diana Marculescu has been named a distinguished scientist for her significant impact on the computing field by the Association for Computing Machinery (ACM). Less than 10 percent of ACM Professional members are selected as Distinguished members.
ECE Professors Radu and Diana Marculescu, former Ph.D. student Umit Ogras and former ECE post-doc Eun Gu Jung have earned the 2011 IEEE Circuits and Systems Society VLSI Transactions Best Paper Award. The award recognizes the best paper published in the Transactions on Very Large Scale Integration Systems publication and is based on general quality, originality, contributions, subject matter and timeliness.
Carnegie Mellon and the University of Pittsburgh co-host the first CRA-W/CDC Workshop on Diversity in Design Automation and Test May 23–24 at the University of Pittsburgh's University Club. The workshop, “Putting D(iversity) in Design Automation and Test,” intends to provide senior undergraduates, graduate students and early post-doctoral researchers from underrepresented groups with an overview of the field, research directions and career paths available in design automation and test.
Diana Marculescu has received the 2010 ACM/Special Interest Group on Design Automation (SIGDA) Distinguished Service Award in recognition for her dedicated service as SIGDA Chair (2005–2009), and contributions to SIGDA, DAC and the electronic design automation profession. Marculescu has led and initiated several ACM/SIGDA activities as a volunteer since 2002, and she served as DAC's coordinator for Collocated Events and Workshops in 2009. She is the leader of the systems thrust within CSSI.
Siddharth Garg has received the 2010 Angel G. Jordan Award from Dept. of ECE. The A.G. Jordan award recognizes a graduating ECE graduating doctoral student who has combined outstanding Ph.D. Thesis work with exceptional service to the ECE or Carnegie Mellon community. In addition to performing high caliber research for his thesis, Siddharth has served as a Vice-President for the ECE Graduate student Organization (2007-2009) and EGO volunteer.
A paper authored by Siddharth Garg, Diana Marculescu and Radu Marculescu on technology-driven limits on voltage and frequency control of multi-processor platforms received a Best in Session Award at the Semiconductor Research Corporation (SRC) TECHCON.
Siddharth Garg and Diana Marculescu's research on characterizing the impact of process variations on 3D integrated circuit technology has received a Best Paper Award at the IEEE International Symposium on Quality Electronic Design (ISQED) 2009.
Kai-Chiang Wu and Diana Marculescu's paper on soft-error hardening of digital circuits using dual supply voltages has received a best paper award at the IEEE International Conference on Computer Design (ICCD) 2009.
EnyAC receives two Best Paper nominations at the Design Automation Conference (DAC) 2008 out of a total of eight papers nominated! Sebastian Herbert and Diana Marculescu's work on characterizing chip-multiprocessor (CMP) variability tolerance addresses the important issue of how next-generation single-chip parallel computing platforms are impacted by process variations. Umit Ogras, Radu Marculescu and Diana Marculescu's work on feedback control for on-chip power management is aimed at increasing the energy efficiency and variability tolerance of multiple processor systems-on-chip platforms.
Natasa Miskov-Zivanov and Diana Marculescu have been nominated for a Best Paper award at the IEEE International Sympoium on Quality Electronic Design (ISQED) 2007 for their paper on soft-error analysis and mitigation techniques for sequential circuits. Congratulations Natasa and Diana!
January 18th, 2005. Kaushik Niyogi and Diana Marculescu have won a Best Paper Award at the ACM/IEEE Asia and South Pacific Design Automation Conference (ASP-DAC) 2005 for their paper titled “Speed and Voltage Selection for GALS Systems based on Voltage/Frequency Islands.”