A Dynamic Convolutional Layer for short rangeweather prediction

By Benjamin Klein, Lior Wolf, Yehuda Afek. 2015

189 citations recorded by the source corpus.

We present a new deep network layer called “Dynamic Convolutional Layer” which is a generalization of the convolutional layer. The conventional convolutional layer uses filters that are learned during training and are held constant during testing. In contrast, the dynamic convolutional layer uses filters that will vary from input to input during testing. This is achieved by learning a function that maps the input to the filters. We apply the dynamic convolutional layer to the application of short range weather prediction and show performance improvements compared to other baselines.

Subfields: Atmospheric Science, Environmental Engineering, Global and Planetary Change

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Atmospheric Science

In Environmental Engineering

In Global and Planetary Change

Citations in the corpus

DISCO: Dynamic and Invariant Sensitive Channel Obfuscation for deep neural networks (2020, methodology): “Following [47], all these recent works [46, 48, 49] utilize the same concept to learn a steerable filter [46], weather prediction filter [48], an image enhancement filter [49], and a dynamic motion motion representation filter [50] using source-target image pairs.”
Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting (2015, background): “d nowcasting is done for the following 1-6 hours, i.e., to predict the 6-60 frames ahead. From the ma2It is worth noting that our precipitation nowcasting problem is different from the one studied in [14], which aims at predicting only the central region of just the next frame. 2 chine learning pers”
ABC-CNN: An Attention Based Convolutional Neural Network for Visual Question Answering (2015, background): “Configurable convolutional neural network: In [11], a dynamic convolutional layer architecture is proposed for short range weather prediction.”
A Transdisciplinary Review of Deep Learning Research and Its Relevance for Water Resources Scientists (2017, methodology): “, filters with weights that are dynamically updated using inputs during forward runs, in short-range weather predictions (Klein et al., 2015).”
Linear Transformers Are Secretly Fast Weight Programmers (2021, methodology): “…hypernetworks (Ha et al., 2017; Perez et al., 2018; Galanti & Wolf, 2020), dynamic plasticity (Miconi et al., 2018; 2019), dynamic convolution (Klein et al., 2015; Noh et al., 2016; Jia et al., 2016), or lambda networks (Bello, 2021) used for applications including meta-learning (Munkhdalai &…”
Personalized Federated Learning using Hypernetworks (2021, background): “Hypernetworks (HNs) (Klein et al., 2015; Riegler et al., 2015; Ha et al., 2017) are deep neural networks that output the weights of another target network, that performs the learning task.”
DynamoNet: Dynamic Action and Motion Network (2019, methodology): “The filter generating network (DynamoNet) is inspired from [1, 26, 38] and is composed of 3D filters and pooling kernels, with the last fully-connected layer (i.”
Incorporating Side Information by Adaptive Convolution (2017, background): “Klein et al. (2015) and De Brabandere et al. (2016) focus on image prediction tasks (e.g., predicting the next frame from the previous frames), and the dynamically-generated ﬁlters are mainly used to transfer a pixel value in the input image to a new position in the output image (e.g., predicting…”
A Review on Deep Learning Techniques for Video Prediction (2020, background): “DFN [162] 2016 [89], [161] r-cED [74], [115] RGB RGB X BCE SR ? ? ? X X”
Classification-Driven Dynamic Image Enhancement (2017, background): “First stage (enhancement stage): The enhancement network (EnhanceNet) is inspired by [7, 18, 20], and is composed of convolutional and fully-connected layers.”
Adaptive Convolution for Object Detection (2019, methodology): “other drawback of the methods in [27]–[29] is the limitation of model parameters.”
When Neurons Fail - Technical Report (2016, background): “Neural networks are now back in the headlines for their outstanding performance [30, 33], in tasks such as function approximation, image and speech recognition, as well as weather prediction [15].”
Dynamic Filter Networks (2016, methodology): “ a ﬁltering operation on an image, hence we do not have the same problem of predicting large amounts of parameters. Most similar to our work, a dynamic convolution layer is proposed by Klein et al.in [13] in the context of short range weather prediction and by Riegler et al.in [19] for single image ”