Artificial Neural Networks and Machine Learning – ICANN 2023: 32nd International Conference on Artificial Neural Networks, Heraklion, Crete, Greece, September 26–29, 2023, Proceedings, Part VII

... F1-score heavily. Near 1 balanced accuracy score for both experiment setups ... 2011-0117. References 1. Ahangaryan, F.P., Mohammadpour, T., Kianisarkaleh ... (2011) 5. Brodersen, K.H., Ong, C.S., Stephan, K.E., Buhmann, J.M.: The balanced accu ...

... Customer Service Dialogue Text Shengjie Ji12 , Chun Chen1 , 2 , and Fang Kong1 , 2 ( ) Laboratory for Natural Language Processing , Soochow University , Suzhou , China { 20215227043 ... Customer Service Dialogue Text 1 Introduction.

... probability of the entity's existence and orientation represents properties ... theory of neural networks and a new direction in the field of deep learning. The ... 10 [4]. Using the single CapsNet model same as [3], Xi et al. [5] reached ...

... NetCloud ← Net.Split(L i + 1, Last); 6 Engine Edge ← Net Edge (DataType <INT8>); 7 EngineCloud ← NetCloud (DataType <FP32>); 8 (Li, info) ← PredictPerformance(Engine Edge, EngineCloud); 9 P ← P∪(Li, info); 10 Env = GetEnvironment ...

... training workers. This coarse-grained parallelism strategy significantly accelerates the training of network models ... Caffe, and name the FiLayer-based system as LP-Caffe. The experimental results show that speedups of 1.58×–2.19× are ...

... summarization model that decouples the syntactic analysis process from the ... sentence. Although this approach can theoretically handle both extractive and abstractive ... Fast abstractive summarization with reinforce-selected sentence ...

... dropout layers only in the two layers just before the YOLOv3 detection ... uncertainty representing variance when detecting multiple times using MC ... Bayesian convolutional neural networks with Bernoulli approximate variational ...

... abstractive summarization. In: Proceedings of the 58th Annual Meeting of the ... Fast abstractive summarization with reinforce-selected sentence rewriting ... abstractive summarization using inconsistency loss. In: ACL 2018: 56th Annual Meeting ...

... Introduction to Algorithms, 3rd edn. The MIT Press, Cambridge (2009) Daniel ... Newman, M.E.J.: Community structure in social and biological networks. In ... Networks: An Introduction. Oxford University Press, New York (2010) Newman ...

... Dropout on an FPGA using edge computing. References 1. Feng, D., Rosenbaum, L., Dietmayer, K.: Towards safe autonomous driving: capture uncertainty in the deep neural network for Lidar 3D vehicle detection. In: 2018 21st International ...

... Training a neural network from scratch requires learning weights and biases for all functions f (1),..., f (L). However ... Caltech 101 and Caltech 256 - reduced image classification. The reduced version of the Caltech 256 dataset was ...

... HGR applications since the parameter estimation in non-stationary processes is difficult. Therefore, machine learning (ML) and deep learning (DL) techniques are typically used to classify EMG signals [4,7]. In particular, supervised ...

... uncertainty estimation methods. We reveal that label noise has a stronger impact on the reliability of uncertainty ... Dropout as a Bayesian approximation: representing model uncertainty in deep learning. International Conference on Machine ...

... (RSB) RSN-18 26.0 27.6 Table 3. Ablation study on different modules, 1s means single-stage, Input size of backbone network is 64 × 48. Ours (Res) and Ours (RSB) mean that bottleneck is Residual block and Residual Steps Block ...

... IMDB4. The four datasets have 3,772, 10,662, 10624, 50,000 instances respectively and the positive and negative instances rates are 0.64/0.36, 0.31/0.69, 0.5/0.5,0.5/0.5. In this research, for the first three datasets, we randomly split ...

... agent training for scene graph generation. In: Proceedings of the IEEE/CVF International Conference on Computer ... Wu, Z., Ernst, J., Batra, D., Parikh, SubCrime 317 5 Conclusion References.

... Caltech-101 dataset keeps images of objects belonging to 101 categories. For the Scenes15 and Caltech-101 databases, features were extracted using a deep learning strategy through VGG-19 [24], a deep neural network. The output of the ...

... scale model of the cerebellum composed of more than 1 billion spiking neurons, which is almost the same size of the whole cat cerebellum. Fig. 3. Schematic of implementation of multiple microcomplexes on PEZY-SC chips. The 1,280 chips ...

... Caltech-101 Silhouettes [15]. As they have been quite well studied ... training samples systematically resulting in higher probability being ... training samples, and hidden biases c were initialized to negative values (−4) in order to ...

... phoneme classification is a subtask) are learned with equal learning rate ... Table 1compares the frame level articulatory feature and phoneme ... Table 1. Frame level articulatory feature and phoneme classification accuracies. 1 http ...

This two volume set (LNCS 6791 and LNCS 6792) constitutes the refereed proceedings of the 21th International Conference on Artificial Neural Networks, ICANN 2011, held in Espoo, Finland, in June 2011.

... XXnet-target: –. net.)” p : where net-target' is the desired input value of the ith unit of the output layer, and net, is the net input to unit i and is given by the weighted sum of all the inputs to the unit. In principle, minimizing ...

... AXON's fascicle functions . Another important point with re- spect to expressibility is that we incorporated object ... User's Manual , Release V6.0 , The MITRE Corporation . [ 6 ] Linden , A. et al . ( 1993 ) An object - oriented framework ...

... STDP Curve in the Presence/Absence of GABA Inhibition. In: Alippi, C., et al. (eds.) ICANN 2009. LNCS, vol. 5768, pp. 229–238. Springer, Heidelberg (2009) Cutsuridis, V., Cobb, S., Graham, B.P.: Modelling the STDP Symmetry-toAsymmetry ...

... CMU test set, which are the best results published so far on this test set. Locating a face and recognizing it [10] ... logos in a straightforward way. Most logo detection approach consider opaque logos superimposed on video frames. In ...