| 1. |
- Ivanenko, Yevhen, et al.
(författare)
-
Quasi-Herglotz functions and convex optimization
- 2020
-
Ingår i: Royal Society Open Science. - : The Royal Society Publishing. - 2054-5703. ; 7:1, s. 1-15
-
Tidskriftsartikel (refereegranskat)abstract
- We introduce the set of quasi-Herglotz functions and demonstrate that it has properties useful in the modelling of non-passive systems. The linear space of quasi-Herglotz functions constitutes a natural extension of the convex cone of Herglotz functions. It consists of differences of Herglotz functions and we show that several of the important properties and modelling perspectives are inherited by the new set of quasi-Herglotz functions. In particular, this applies to their integral representations, the associated integral identities or sum rules (with adequate additional assumptions), their boundary values on the real axis and the associated approximation theory. Numerical examples are included to demonstrate the modelling of a non-passive gain medium formulated as a convex optimization problem, where the generating measure is modelled by using a finite expansion of B-splines and point masses.
|
|
| 2. |
- Mochaourab, Rami, et al.
(författare)
-
Post-hoc Explainability for Time Series Classification: Towards a Signal Processing Perspective
- 2022
-
Ingår i: IEEE signal processing magazine (Print). - : Institute of Electrical and Electronics Engineers (IEEE). - 1053-5888 .- 1558-0792. ; 39:4, s. 119-129
-
Tidskriftsartikel (refereegranskat)abstract
- Time series data correspond to observations of phenomena that are recorded over time [1]. Such data are encountered regularly in a wide range of applications, such as speech and music recognition, monitoring health and medical diagnosis, financial analysis, motion tracking, and shape identification, to name a few. With such a diversity of applications and the large variations in their characteristics, time series classification is a complex and challenging task. One of the fundamental steps in the design of time series classifiers is that of defining or constructing the discriminant features that help differentiate between classes. This is typically achieved by designing novel representation techniques [2] that transform the raw time series data to a new data domain, where subsequently a classifier is trained on the transformed data, such as one-nearest neighbors [3] or random forests [4]. In recent time series classification approaches, deep neural network models have been employed that are able to jointly learn a representation of time series and perform classification [5]. In many of these sophisticated approaches, the discriminant features tend to be complicated to analyze and interpret, given the high degree of nonlinearity.
|
|
| 3. |
- Hidvegi, Attila, et al.
(författare)
-
A Small Portable Test System for the TileCal Digitizer System
- 2008
-
Konferensbidrag (refereegranskat)abstract
- The hadronic Tile Calorimeter (TileCal) of the ATLAS detector at LHC has a digitization, pipeline and readout system composed of nearly 2000 boards [1][2], developed and maintained by Stockholm University. Prior to now a rather complex test system been used to verify the functionality of the boards. However this system was developed nearly 10 years ago and now difficult to maintain due to several already obsolete components. A new, simpler, more reliable, and portable test system was therefore initiated. Its components have been chosen to reduce problems with obsolescence, and to allow easy migration to new platforms over the lifetime of the digitizer system.
|
|
| 4. |
- Hidvegi, Attila, et al.
(författare)
-
A system for distributing high-speed synchronous high-precision clock and trigger data over large distances
- 2008
-
Ingår i: Nuclear Science Symposium Conference Record, 2008. NSS '08. IEEE. - 9781424427147 ; , s. 2581-2584
-
Konferensbidrag (refereegranskat)abstract
- The distribution of precise timing throughout the European X-ray Free Electron Laser project [1] (XFEL) and its triggering system is a very challenging part of the system design. ADCs in data acquisition systems and DACs in control systems will require very high precision clocks. The clocks need to be synchronous to each other, both in frequency and phase, with a jitter performance better than 5 ps (RMS). At some high-speed ADCs it might even need a precision down to 0.1 ps. The frequencies that must be available are the main 1.3 GHz and some frequencies below, which are all derived from the main frequency. The phase needs to be adjustable to allow synchronization between separate devices. Triggering information needs to be distributed over the system, so that controlling instructions can be carried out at a very precise time. This is very important since the beam will travel with the speed of light, and there is no possibility for information to be sent back and forth. This requires an absolute timing to be distributed over the system. Both the main clock and triggering information will be transmitted over the same fiber cable, one to each device. An advanced synchronization method needs to be developed to synchronize the phases of the clocks throughout the whole system. The delay through the cable can change with temperature, and due to long cables the total change through a single cable can be significant. It is essential that the clocks are stable and not drifting away from each other. Therefore a continuous calibration method is needed, ensuring that the clocks are synchronous throughout the whole system. A prototype of such a system is being developed and a first version is expected to be completed in 2009 Ql.
|
|
| 5. |
- Hidvegi, Attila, et al.
(författare)
-
An Advanced FPGA Based Phase-Lock-Loop System as an Alternative Solution for the XFEL Timing System
- 2009
-
Ingår i: Nuclear Science Symposium Conference Record (NSS/MIC), 2009 IEEE. - 9781424439614 ; , s. 1871-1872
-
Konferensbidrag (refereegranskat)abstract
- The European XFEL project requires a high-speed, very precise clock and timing distribution over large distances. A prototype system which fulfils current requirements that uses high-end components has just been completed and is being tested. However, the system is quite complicated and the boards are very complex, being designed using the small micro-TCA form factor. A way to simplify the system, and perhaps reduce cost, would be to implement an Advanced PLL in the programmable logic of an FPGA, which then would control an external VCO. By doing so several major issues could be resolved at the same time, while making more use of the advanced features of modern FPGAs. Such a system could be an alternative solution to the complex part of the Timing and Triggering System for XFEL.
|
|
| 6. |
- Hidvegi, Attila, et al.
(författare)
-
Timing and Triggering System Prototype for the XFEL Project
- 2010
-
Ingår i: IEEE Transactions on Nuclear Science. - 0018-9499 .- 1558-1578. ; 58:4, s. 1852-1856
-
Tidskriftsartikel (refereegranskat)abstract
- The European X-ray Free Electron Laser (XFEL) [1] at DESY in Hamburg will begin operating in the next few years, enabling new, ground-breaking research opportunities. The entire system requires very precise clock and trigger distribution, synchronous with the 1.3 GHz system RF-frequency, over distances of more than 3.4 km. The new experiment demanded features that other commercial solutions could not yet provide. Researchers at Stockholm University and DESY have developed a prototype for the timing system of XFEL. It has been decided that XFEL will use modern ATCA and Micro-TCA systems because of their advanced features and reliability. The timing system has been adapted to the Micro-TCA bus standard and also follows the new upcoming xTCA for physics standard. The prototype is fully functional and complete. It will serve as a platform for future development of the whole timing system. This paper describes the hardware design and some test results using the prototype board.
|
|
| 7. |
- Berglund, Erik, et al.
(författare)
-
Distributed Newton Method Over Graphs : Can Sharing of Second-Order Information Eliminate the Condition Number Dependence?
- 2021
-
Ingår i: IEEE Signal Processing Letters. - : Institute of Electrical and Electronics Engineers (IEEE). - 1070-9908 .- 1558-2361. ; 28, s. 1180-1184
-
Tidskriftsartikel (refereegranskat)abstract
- One of the main advantages of second-order methods in a centralized setting is that they are insensitive to the condition number of the objective function's Hessian. For applications such as regression analysis, this means that less pre-processing of the data is required for the algorithm to work well, as the ill-conditioning caused by highly correlated variables will not be as problematic. Similar condition number independence has not yet been established for distributed methods. In this paper, we analyze the performance of a simple distributed second-order algorithm on quadratic problems and show that its convergence depends only logarithmically on the condition number. Our empirical results indicate that the use of second-order information can yield large efficiency improvements over first-order methods, both in terms of iterations and communications, when the condition number is of the same order of magnitude as the problem dimension.
|
|
| 8. |
- Khirirat, Sarit, et al.
(författare)
-
Compressed Gradient Methods With Hessian-Aided Error Compensation
- 2021
-
Ingår i: IEEE Transactions on Signal Processing. - : Institute of Electrical and Electronics Engineers (IEEE). - 1053-587X .- 1941-0476. ; 69, s. 998-1011
-
Tidskriftsartikel (refereegranskat)abstract
- The emergence of big data has caused a dramatic shift in the operating regime for optimization algorithms. The performance bottleneck, which used to be computations, is now often communications. Several gradient compression techniques have been proposed to reduce the communication load at the price of a loss in solution accuracy. Recently, it has been shown how compression errors can be compensated for in the optimization algorithm to improve the solution accuracy. Even though convergence guarantees for error-compensated algorithms have been established, there is very limited theoretical support for quantifying the observed improvements in solution accuracy. In this paper, we show that Hessian-aided error compensation, unlike other existing schemes, avoids accumulation of compression errors on quadratic problems. We also present strong convergence guarantees of Hessian-based error compensation for stochastic gradient descent. Our numerical experiments highlight the benefits of Hessian-based error compensation, and demonstrate that similar convergence improvements are attained when only a diagonal Hessian approximation is used.
|
|
| 9. |
- Magnússon, Sindri, et al.
(författare)
-
On Maintaining Linear Convergence of Distributed Learning and Optimization Under Limited Communication
- 2020
-
Ingår i: IEEE Transactions on Signal Processing. - : Institute of Electrical and Electronics Engineers (IEEE). - 1053-587X .- 1941-0476. ; 68, s. 6101-6116
-
Tidskriftsartikel (refereegranskat)abstract
- In distributed optimization and machine learning, multiple nodes coordinate to solve large problems. To do this, the nodes need to compress important algorithm information to bits so that it can be communicated over a digital channel. The communication time of these algorithms follows a complex interplay between a) the algorithm's convergence properties, b) the compression scheme, and c) the transmission rate offered by the digital channel. We explore these relationships for a general class of linearly convergent distributed algorithms. In particular, we illustrate how to design quantizers for these algorithms that compress the communicated information to a few bits while still preserving the linear convergence. Moreover, we characterize the communication time of these algorithms as a function of the available transmission rate. We illustrate our results on learning algorithms using different communication structures, such as decentralized algorithms where a single master coordinates information from many workers and fully distributed algorithms where only neighbours in a communication graph can communicate. We conclude that a co-design of machine learning and communication protocols are mandatory to flourish machine learning over networks.
|
|
| 10. |
- Berglund, Erik, et al.
(författare)
-
Revisiting the Curvature-aided IAG : Improved Theory and Reduced Complexity
- 2023
-
Ingår i: IFAC-PapersOnLine. - : Elsevier BV. - 2405-8963. ; 56:2, s. 5221-5226, s. 5221-5226
-
Tidskriftsartikel (refereegranskat)abstract
- The curvature-aided IAG (CIAG) algorithm is an efficient asynchronous optimization method that accelerates IAG using a delay compensation technique. However, existing step-size rules for CIAG are conservative and hard to implement, and the Hessian computation in CIAG is often computationally expensive. To alleviate these issues, we first provide an easy-to-implement and less conservative step-size rule for CIAG. Next, we propose a modified CIAG algorithm that reduces the computational complexity by approximating the Hessian with a constant matrix. Convergence results are derived for each algorithm on both convex and strongly convex problems, and numerical experiments on logistic regression demonstrate their effectiveness in practice.
|
|
