
On November 25, the Moscow office of the Mail.Ru Group will host the traditional meeting of the Moscow Data Science community. Participants will share their professional experience in solving practical data analysis tasks and communicate in an informal setting. The meeting will be dedicated to deep learning. Join now!
Meeting program:
- Deep Metric LearningSergey Ovcharenko, NTech.Lab
')
The report will consider several approaches to the training of convolutional neural networks to obtain representations with the properties of semantic proximity. There will also be a comparative analysis of the received views on the example of the MNIST dataset.
- High performance indentation of deep GPU networks with TensorRTMaxim Milakov, NVIDIA
Inference performance is one of the most serious problems in the implementation of DL-applications, since it determines what impression the service will have from the user, as well as the price of product implementation. Thus, it is important for the inference to be highly productive and energy efficient. TensorRT automatically optimizes a trained neural network for maximum performance, providing significant acceleration compared to common, frequently used frameworks.
- Generative models against cancerArthur Kadurin, Mail.Ru Group
The speaker will tell how with the help of neural networks we can (and, obviously, should) fight even diseases such as cancer.
Participation is free, but you must
register . The event starts at 19:00.
For those who will not be able to attend in person, an
online broadcast will be organized.