Warsaw University of Technology

About: Warsaw University of Technology is a education organization based out in Warsaw, Poland. It is known for research contribution in the topics: Microstructure & Optical fiber. The organization has 14293 authors who have published 34362 publications receiving 492211 citations. The organization is also known as: Warsaw Polytechnic & Politechnika Warszawska.

Simulation model of transport system of Poland as a tool for developing sustainable transport

Abstract: Paper presents features of simulation model of proecological transport system on the example of Poland. Model allows computational experimentation and inference on transport modal split and emission of pollution in national transport system. Particular elements of the model are characterized: transport networks for different modes, stock of vehicles, demand model for passenger and freight transport, and mechanism of material and passenger flows distribution into a network. Characteristics describing infrastructure, vehicles, and harmful compounds of exhaust gases are given. Model is implemented in PTV VISUM. Road and rail vehicles for passenger and freight transport are characterized and divided into groups according to types. The demand for transport services and emission of exhaust gases components are reflected in model of proecological transport system of Poland. The last part of paper presents exemplary results of research on estimating emission from transport activities.

Coherent MapDrift Technique

Abstract: A new parametric autofocus technique with a high accuracy of flight-parameter estimation dedicated to strip-mode synthetic aperture radar (SAR) systems is presented. Most of the known autofocus techniques require high-reflectivity targets (man-made targets) to obtain a properly focused SAR image. The technique proposed in this paper allows flight parameters to be estimated effectively, even for a low-contrast scene (e.g., forests, fields, small paths, etc.). The autofocus technique is based on well-known MapDrift (MD) principles. The presented technique is a coherent one, which allows flight parameters to be estimated more precisely than in the other well-known parametric technique referred to as classical MD. The presented technique allows flight parameters to be estimated with accuracy that is independent of the initial velocity error. It can be used for real-time processing for both Earth imaging and moving-target indication.

Model Predictive Control for Three-Level Four-Leg Flying Capacitor Converter Operating as Shunt Active Power Filter

Abstract: This paper presents the new implementation of the finite control states set model predictive control (FS-MPC) applied to three-level four-leg flying capacitor converter (FCC) operating as a shunt active power filter (SAPF). The three main issues regarding the algorithm development are described and the solutions are proposed. The first is addressed to modeling of the four-wire system, which for simplification is based on wire-to-wire voltage equations. The second relates to the control of FCC by MPC, i.e., a proper interpretation of the FCC switching states and control of flying capacitor (FC) voltages, which was moved from the prediction loop. Finally, the system restriction regarding undesirable switching states’ transitions is explained and the solution is presented. The control performance was tested in the simulation model in MATLAB–Simulink and validated experimentally by measurements on the 10-kVA FCC laboratory setup.

Predicting popularity of online videos using Support Vector Regression

Abstract: In this work, we propose a regression method to predict the popularity of an online video based on temporal and visual cues. Our method uses Support Vector Regression with Gaussian Radial Basis Functions. We show that modelling popularity patterns with this approach provides higher and more stable prediction results, mainly thanks to the non-linearity character of the proposed method as well as its resistance against overfitting. We compare our method with the state of the art on datasets containing over 14,000 videos from YouTube and Facebook. Furthermore, we show that results obtained relying only on the early distribution patterns, can be improved by adding social and visual metadata.

Frustratingly Short Attention Spans in Neural Language Modeling

Abstract: Neural language models predict the next token using a latent representation of the immediate token history. Recently, various methods for augmenting neural language models with an attention mechanism over a differentiable memory have been proposed. For predicting the next token, these models query information from a memory of the recent history which can facilitate learning mid- and long-range dependencies. However, conventional attention mechanisms used in memory-augmented neural language models produce a single output vector per time step. This vector is used both for predicting the next token as well as for the key and value of a differentiable memory of a token history. In this paper, we propose a neural language model with a key-value attention mechanism that outputs separate representations for the key and value of a differentiable memory, as well as for encoding the next-word distribution. This model outperforms existing memory-augmented neural language models on two corpora. Yet, we found that our method mainly utilizes a memory of the five most recent output representations. This led to the unexpected main finding that a much simpler model based only on the concatenation of recent output representations from previous time steps is on par with more sophisticated memory-augmented neural language models.