Supplementary Materials: Augmenting Supervised Neural Networks with Unsupervised Objectives for Large-scale Image Classification
نویسندگان
چکیده
We report the learning parameters for 16-layer VGGNet-based model in Table S-1. We chose the learning rates that lead to the largest decrease in the reconstruction loss in the first 2000 iterations for each layer. The “loss weighting” are balancing factors for reconstruction losses in different layers varied to make them comparable in magnitude. In particular, we computed image reconstruction loss against RGB values normalized to [0,1], which are different in scale from intermediate features. We also did not normalize the reconstruction loss with feature dimensions for any layer.
منابع مشابه
Augmenting Supervised Neural Networks with Unsupervised Objectives for Large-scale Image Classification
Unsupervised learning and supervised learning are key research topics in deep learning. However, as high-capacity supervised neural networks trained with a large amount of labels have achieved remarkable success in many computer vision tasks, the availability of large-scale labeled images reduced the significance of unsupervised learning. Inspired by the recent trend toward revisiting the impor...
متن کاملطبقه بندی و شناسایی رخسارههای زمینشناسی با استفاده از دادههای لرزه نگاری و شبکههای عصبی رقابتی
Geological facies interpretation is essential for reservoir studying. The method of classification and identification seismic traces is a powerful approach for geological facies classification and distinction. Use of neural networks as classifiers is increasing in different sciences like seismic. They are computer efficient and ideal for patterns identification. They can simply learn new algori...
متن کاملUnsupervised Feature Learning With Symmetrically Connected Convolutional Denoising Auto-encoders
Unsupervised pre-training was a critical technique for training deep neural networks years ago. With sufficient labeled data and modern training techniques, it is possible to train very deep neural networks from scratch in a purely supervised manner nowadays. However, unlabeled data is easier to obtain and usually of very large scale. How to make use of them better to help supervised learning i...
متن کاملDetermination of Best Supervised Classification Algorithm for Land Use Maps using Satellite Images (Case Study: Baft, Kerman Province, Iran)
According to the fundamental goal of remote sensing technology, the image classification of desired sensors can be introduced as the most important part of satellite image interpretation. There exist various algorithms in relation to the supervised land use classification that the most pertinent one should be determined. Therefore, this study has been conducted to determine the best and most su...
متن کاملINTEGRATED ADAPTIVE FUZZY CLUSTERING (IAFC) NEURAL NETWORKS USING FUZZY LEARNING RULES
The proposed IAFC neural networks have both stability and plasticity because theyuse a control structure similar to that of the ART-1(Adaptive Resonance Theory) neural network.The unsupervised IAFC neural network is the unsupervised neural network which uses the fuzzyleaky learning rule. This fuzzy leaky learning rule controls the updating amounts by fuzzymembership values. The supervised IAFC ...
متن کامل