On The Edge Of AI. Breakthrough cases
The session focuses on deeper, applied, and at the same time fundamental challenges on the way of AI application today - on the way of ML deployment in the company - both with the vendor and with the customer corporation.
Attacks on trained models are becoming increasingly frequent. Companies should take priority over protecting the model architecture. Misinterpretation of the model operation or lack of control over the use of its results can lead to disaster or financial losses.
The speakers will discuss:
- How to build machine learning if there are no proprietary data or they are critically scarce and you need to run the process as soon as possible?
- How to generate data from a small dataset and train a model based on it?
- How to handle Big Data with interruptible virtual machines?
- What is Meta Learning and how to use it?