About this webcast
The Hive and IBM invite you to an interactive webcast on the combination of AI techniques and Cloud. In this webcast you will see for yourself that training and deploying AI models in production using open source tools has never been easier.
Continuous development, training, testing, deployment and monitoring of machine learning models requires a highly-tuned and secure system with the right combination of software, machine learning libraries and infrastructure.
There’s also now a growing demand for fairness, accountability, and transparency from machine learning (ML) systems. And we need to remember that training data isn’t the only source of possible bias and adversarial contamination. Bias can also be introduced through inappropriate data handling, inappropriate model selection, or incorrect algorithmic design.
What we need is a ML pipeline that is open, transparent, secure and fair. We need an ML pipeline that fully integrates into the AI lifecycle. Such a pipeline requires a robust set of bias and adversarial checkers, “de-biasing” and “defense” algorithms, and explanations.
An open ML pipeline needs an infrastructure to parallelize, scale and automate distributed training of ML models. It needs to understand how to do advanced batch scheduling effectively, how to add concepts of priority, preemption, effective placement.
On the model serving side, we need a system to automate A/B tests and canary testing of models, monitoring concept drifts and accuracy losses,
In this talk Animesh Singh will show you how to build a continuous learning AI pipeline to train, validate, de-bias, deploy and monitor models, leveraging open source projects such as:
- AI Fairness 360 (AIF360)
- Adversarial Robustness Toolbox (ART)
- Fabric for Deep Learning (FfDL)
- Seldon for AI model serving
For additional capabilities and integration, Watson Studio, Watson Machine Learning, PowerAI, Trust and Transparency offerings from IBM will also be discussed for advanced enterprise AI solutions.