Source author record

Kevin Trebing

Kevin Trebing appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Machine Learning eess.IV

Catalog footprint

What is connected

2works

2topics

2close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2021arXiv

SmaAt-UNet: Precipitation Nowcasting using a Small Attention-UNet Architecture

Weather forecasting is dominated by numerical weather prediction that tries to model accurately the physical properties of the atmosphere. A downside of numerical weather prediction is that it is lacking the ability for short-term forecasts using the latest available information. By using a data-driven neural network approach we show that it is possible to produce an accurate precipitation nowcast. To this end, we propose SmaAt-UNet, an efficient convolutional neural networks-based on the well known UNet architecture equipped with attention modules and depthwise-separable convolutions. We evaluate our approaches on a real-life datasets using precipitation maps from the region of the Netherlands and binary images of cloud coverage of France. The experimental results show that in terms of prediction performance, the proposed model is comparable to other examined models while only using a quarter of the trainable parameters.

preprint2020arXiv

Wind speed prediction using multidimensional convolutional neural networks

Accurate wind speed forecasting is of great importance for many economic, business and management sectors. This paper introduces a new model based on convolutional neural networks (CNNs) for wind speed prediction tasks. In particular, we show that compared to classical CNN-based models, the proposed model is able to better characterise the spatio-temporal evolution of the wind data by learning the underlying complex input-output relationships from multiple dimensions (views) of the input data. The proposed model exploits the spatio-temporal multivariate multidimensional historical weather data for learning new representations used for wind forecasting. We conduct experiments on two real-life weather datasets. The datasets are measurements from cities in Denmark and in the Netherlands. The proposed model is compared with traditional 2- and 3-dimensional CNN models, a 2D-CNN model with an attention layer and a 2D-CNN model equipped with upscaling and depthwise separable convolutions.