16                                         

 Review for Classical Deep Learning Algorithms with Applications

主讲人：金海

摘要： In this talk, we shall first look at mathematical backgrounds for deep learning: loss/optimizer/metrics,overfitting vs generalization, and classification vs regression. Then, we will cover many aspects of Computer Vision (CV) algorithms: convnets(such as those in the Xception model), data augmentation, feature extraction and fine-tuning with pretrained models, and model interpretation. Next, we will discuss DL algorithms for time series or sequence data: LSTM, GRU, the transformer architecture. Last, we will study a few interesting DL applications, such as DeepDream, Neural style transfer, and image generation. The audience may expect this talk as a comprehensive review for classical DL algorithms with implementation details-the source code.

主讲人简介：金海, 副教授. 研究领域: 微分方程数值解法, 机器/深度学习算法. 2006 年博士毕业于美国华盛顿大学, 2008年入职北京理工大学. 主要从事各种数值算法、ML/DL算法、及各种算法的应用. 研究成果发表于Numerical Algorithms、Journal of Computational and Applied Mathematics、Applied Numerical Mathematics、European Journal of Physics等期刊上。

邀请人：周少波

时间：2024年8月2日上午10：00-12：00

地点：逸夫科技楼711室