Back Propagation
By Li Shen, Zhouchen Lin, Qingming Huang. 2019
The results of research on the value of exports in Indonesia can be concluded that the architectural model can make predictions with 100% accuracy with a short training time. In addition, by looking at the results of testing on architecture, it can be seen that both the speed and the prediction results, it can be concluded that the value of exports in Indonesia is increasingly declining. For future research, research should use a different algorithm or the Backpropagation algorithm can be optimized with other algorithms, such as conjugate gradient and so on, that is, the research was first carried out using the backpropagation algorithm, then after knowing the results, calculations were carried out with the algorithm. other. After that, a comparison is made between the results of backpropagation with the results of other algorithms. Images or graphs of each algorithm used, or a graphic image of the comparison of the previous data with the data that has been generated using the algorithm.
Subfields: Artificial Intelligence, Electrical and Electronic Engineering, Optimization & Training
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
In Electrical and Electronic Engineering
In Optimization & Training
Citations in the corpus
- LVIS: A Dataset for Large Vocabulary Instance Segmentation (2019, background): “The last row presents class aware sampling (CAS), an alternate oversampling method [32].”
- Pyramid Scene Parsing Network (2016, background): “Different from relay backpropagation [32] that blocks the backward auxiliary loss to several shallow layers, we let the two loss functions pass through all previous layers.”
- Path Aggregation Network for Instance Segmentation (2018, methodology): “We train these models with different random seeds, with and without balanced sampling [55] to enhance diversity between models.”
- Large-Scale Long-Tailed Recognition in an Open World (2019, background): “We also enable them with class-aware mini-batch sampling [47] for effective learning.”
- Equalization Loss for Long-Tailed Object Recognition (2020, background): “Classaware sampling [38], also called class-balanced sampling, is a typical technique of oversampling, which first samples a category and then an image uniformly that contains the sampled category.”
- BBN: Bilateral-Branch Network With Cumulative Learning for Long-Tailed Visual Recognition (2019, methodology): “, resampling [25, 1, 14, 1, 11, 2, 7, 21, 4] and cost-sensitive re-weighting [13, 28, 5, 23].”
- Disentangling Label Distribution for Long-tailed Visual Recognition (2020, methodology): “Most long-tailed visual recognition methods can be divided into two strategies: modifying the data sampler to balance the class frequency during optimization [7, 21, 25, 8, 51], and modifying the class-wise weights of the classification loss to increase the importance of tail classes in terms of emp”
- Balanced Knowledge Distillation for Long-tailed Learning (2021, methodology): “Over-sampling [44, 4] repeatedly samples training examples from the minority classes, the downside of which is the high potential risk of overfitting.”
- Rethinking the Value of Labels for Improving Class-Imbalanced Learning (2020, methodology): “Classical methods have been focusing on designing data re-sampling strategies, including over-sampling the minority classes [2,41,44] as well as under-sampling the frequent classes [5,31].”
- Balanced Meta-Softmax for Long-Tailed Visual Recognition (2020, background): “Specifically, re-sampling strategies [19, 5, 8, 10, 26, 2, 1] try to restore the true distributions from the imbalanced training data.”
- Knowledge Guided Disambiguation for Large-Scale Scene Classification With Multi-Resolution CNNs (2016, methodology): “ptimization techniques have been also proposed to reduce the difficulty of training networks, so as to improve recognition performance, such as Batch Normalization (BN) [26] and Relay Back Propagation [8]. These works focused on general aspect of applying deep networks for image classification, in par”
- Parametric Contrastive Learning (2021, background): “The most classical way to deal with long-tailed datasets is to over-sample lowfrequency class images [41, 56, 2, 3] or under-sample highfrequency class images [20, 27, 2].”
- Distributionally Robust Neural Networks (2020, methodology): “Concretely, we train our weighted model by sampling with equal probability from each group for each minibatch (Shen et al., 2016), since a recent study found this to be more effective than similar methods (Buda et al.”
- Forest R-CNN: Large-Vocabulary Long-Tailed Object Detection and Instance Segmentation (2020, methodology): “Method Setting AP AP50 AP75 APr APc APf APS APM APL APb Class-aware Sampling [33] ResNet-50-FPN 18.”
- The Devil is in Classification: A Simple Framework for Long-tail Object Detection and Instance Segmentation (2020, methodology): “Repeat Sampling [17], [37] Repeat sampling directly over-samples data (images) with a class-dependent repeating factor, so that the tail classes can be more frequently involved in optimization.”
- CBNet: A Composite Backbone Network Architecture for Object Detection (2021, background): “Although increasing the depth usually leads to performance improvement [13], it may introduce additional optimization difficulties, as in the case of image classification [61].”
- Small Sample Learning in Big Data Era (2018, methodology): “A simple amelioration strategy is re-balancing training data like sampling examples from the rare classes more frequently (Shen et al., 2016) or reducing the number of examples from the top numbered classes (He and Garcia, 2009).”