Massoud Pedram

Bio: Massoud Pedram is an academic researcher from University of Southern California. The author has contributed to research in topics: Energy consumption & CMOS. The author has an hindex of 77, co-authored 780 publications receiving 23047 citations. Previous affiliations of Massoud Pedram include University of California, Berkeley & Syracuse University.

VIBNN: Hardware Acceleration of Bayesian Neural Networks

Abstract: Bayesian Neural Networks (BNNs) have been proposed to address the problem of model uncertainty in training and inference. By introducing weights associated with conditioned probability distributions, BNNs are capable of resolving the overfitting issue commonly seen in conventional neural networks and allow for small-data training, through the variational inference process. Frequent usage of Gaussian random variables in this process requires a properly optimized Gaussian Random Number Generator (GRNG). The high hardware cost of conventional GRNG makes the hardware implementation of BNNs challenging. In this paper, we propose VIBNN, an FPGA-based hardware accelerator design for variational inference on BNNs. We explore the design space for massive amount of Gaussian variable sampling tasks in BNNs. Specifically, we introduce two high performance Gaussian (pseudo) random number generators: 1) the RAM-based Linear Feedback Gaussian Random Number Generator (RLF-GRNG), which is inspired by the properties of binomial distribution and linear feedback logics; and 2) the Bayesian Neural Network-oriented Wallace Gaussian Random Number Generator. To achieve high scalability and efficient memory access, we propose a deep pipelined accelerator architecture with fast execution and good hardware utilization. Experimental results demonstrate that the proposed VIBNN implementations on an FPGA can achieve throughput of 321,543.4 Images/s and energy efficiency upto 52,694.8 Images/J while maintaining similar accuracy as its software counterpart.

Stratified random sampling for power estimation

Abstract: In this paper, we propose a statistical power evaluation framework at the RT-level. We first discuss the power macro-modeling formulation, and then propose a simple random sampling technique to alleviate the the overhead of macro-modeling during RTL simulation. Next, we describe a regression estimator to reduce the error of the macro-modeling approach. Experimental results indicate that the execution time of the simple random sampling combined with power macro-modeling is 50 X lower than that of conventional macro-modeling while the percentage error of regression estimation combined with power macro-modeling is 16 X lower than that of conventional macro-modeling. Hence, we provide the designer with options to either improve the accuracy or the execution time when using power macro-modeling in the context of RTL simulation.

Demand-side load scheduling incentivized by dynamic energy prices

Abstract: Demand response is an important part of the smart grid technologies. This is a particularly interesting problem with the availability of dynamic energy pricing models. Electricity consumers are encouraged to consume electricity more prudently in order to minimize their electric bill, which is in turn calculated based on dynamic energy prices. In this paper, task scheduling policies that help consumers minimize their electrical energy cost by setting the time of use (TOU) of energy in the facility. Moreover, the utility companies can reasonably expect that their customers reduce their consumption at critical times in response to higher energy prices during those times. These policies target two different scenarios: (i) scheduling with a TOU-dependent energy pricing function subject to a constraint on total power consumption; and (ii) scheduling with a TOU and total power consumption-dependent pricing function for electricity consumption. Exact solutions (based on Branch and Bound) are presented for these task scheduling problems. In addition, a rank-based heuristic and a force directed-based heuristic are presented to efficiently solve the aforesaid problems. The proposed heuristic solutions are demonstrated to have very high quality and competitive performance compared to the exact solutions. Moreover, ability of demand shaping utilizing the aforementioned pricing schemes is demonstrated by the simulation results.

Architectural energy optimization by bus splitting

Abstract: This paper proposes split shared-bus architecture to reduce the energy dissipation for global data exchange among a set of interconnected modules. The bus splitting problem for minimum energy is formulated as a minimum-exchange bus split problem, which is shown to be NP-complete. The problem is solved heuristically by using a maximum-weight matching algorithm and combinatorial search. Experimental results show that the energy saving of split-bus architecture compared to monolithic-bus architecture varies from 16% to 50%, depending on the characteristics of the data transfer among the modules and the configuration of the split-bus. The proposed split-bus architecture can be extended to multiway split-bus architecture when large numbers of modules are to be connected.

An energy and deadline aware resource provisioning, scheduling and optimization framework for cloud systems

Abstract: Cloud computing has attracted significant attention due to the increasing demand for low-cost, high performance, and energy-efficient computing. Profit maximization for the cloud service provider (CSP) is a key objective in the large-scale, heterogeneous, and multi-user environment of a cloud system. This paper addresses the problem of minimizing the operation cost of a cloud system by maximizing its energy efficiency while ensuring that user deadlines as defined in Service Level Agreements are met. The workload in the cloud system can be modeled as independent batch requests or as task graphs with dependencies. This paper adopts the latter modeling approach, which provides more opportunities for energy and performance optimizations, thus enabling the CSP to meet user deadlines at lower operation costs. However, these optimizations require additional supporting efforts e.g., resource provisioning, virtual machine placement, and task scheduling, which are addressed in a holistic manner in the proposed framework. In the envisioned cloud environment, users can construct their own services and applications based on the available set of virtual machines, but are relieved from the burden of resource provisioning and task scheduling. The CSP will then exploit data parallelism in user workloads to create an energy and deadline-aware cloud platform.

I and i

Abstract: There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality. Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

Machine learning

Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Pattern Recognition and Machine Learning

Abstract: Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.