23 OCT 2025 - We are back! If you have been following us over the last few years, you will know that the last 2 months have been rough. We website was practically not loading. Sorry for the mess. We are back though and everything should run smoothly now. New servers. Updated domains. And new owners. We invite you all to start uploading torrents again!
TORRENT DETAILS
Macedo D. Enhancing Deep Learning Performance...Linear Unit 2022
TORRENT SUMMARY
Status:
All the torrents in this section have been verified by our verification system
Recently, Deep Learning has caused a significant impact on computer vision, speech recognition, and natural language understanding. In spite of the remarkable advances, Deep Learning (DL) recent performance gains have been modest and usually rely on increasing the depth of the models, which often requires more computational resources such as processing time and memory usage. To tackle this problem, we turned our attention to the interworking between the activation functions and the batch normalization, which is virtually mandatory currently. In this work, we propose the activation function Displaced Rectifier Linear Unit (DReLU) by conjecturing that extending the identity function of ReLU to the third quadrant enhances compatibility with batch normalization. Moreover, we used statistical tests to compare the impact of using distinct activation functions (ReLU, LReLU, PReLU, ELU, and DReLU) on the learning speed and test accuracy performance of VGG and Residual Networks state-of-the-art models. These convolutional neural networks were trained on CIFAR-10 and CIFAR-100, the most commonly used Deep Learning computer vision datasets. The results showed DReLU speeded up learning in all models and datasets. Besides, statistical significant performance assessments (p<0.05) showed DReLU enhanced the test accuracy obtained by ReLU in all scenarios. Furthermore, DReLU showed better test accuracy than any other tested activation function in all experiments with one exception. Deep Learning is based on the simple principle of hierarchical composition of trivial uniform procedures, the artificial neuron computation. This simplicity produces a considerable advantage over classical Machine Learning approaches. In fact, this characteristic is responsible by the extreme scalability that deep learning system are known to present. Hence, if enough training time and data are available, the massive scalability of Deep Learning approaches appears to achieve lower error rate performance than traditional Machine Learning alternatives
VISITOR COMMENTS (0 )
FILE LIST
Filename
Size
Macedo D. Enhancing Deep Learning Performance...Linear Unit 2022.pdf
