Workshops - AI@Enterprise
Here is the list of online workshops. Each one of them will be 4 hours long. Workshops will take place on 11th June 2021.
Workshop 1
AI FOR MANAGERS
9:00 AM - 1:00 PM (CET)
Description:
Many data science or machine learning projects fail due to mistakes done during the project development. We can group the mistakes into a few most popular that if known earlier can make your project successful. The training also gives a better understanding of the topic of artificial intelligence for technical and non-technical managers. We go through the process of the AI transformation and show a few tips on how to make the transformation easy.
Many managers also try to build machine learning related projects in an agile process in a similar way as it’s done in a typical software development project. We show how to do it right and structure the team. Another important topic that is covered in the training is related to quality and signs when to stop the development or research. Finally, we go through several business cases and show customers challenges, expectations and how we solved it
Target Audience:
- Managers having any kind of machine learning related project
- Data scientists or developers who want to avoid mistakes made while building a machine learning project
Participants will understand…
- what types of machine learning methods exist,
- how to make due diligence of a data science project,
- what kind of roles are important in the structure of an AI company.
Participants will be able to…
- avoid common mistakes made while developing a machine learning-based solution,
Outline
- Introduction to machine learning
- Buzzword-driven machine learning projects
- Common mistakes and failures
- AI transformation
- Data science team structure done right
- Quality of a data science team outcome
Q&A
Readings
- Catherine Nelson, Hannes Hapke, Building Machine Learning Pipelines, O’Reilly 2019
- HBR's 10 Must Reads on AI, Analytics, and the New Machine Age, Harvard Business Review Press, 2019
- Susanne Chishti (Editor-in-Chief), Ivana Bartoletti (Editor), Anne Leslie (Editor), Shân M. Millie (Editor), The AI Book: The Artificial Intelligence Handbook for Investors, Entrepreneurs and FinTech Visionaries, Wiley 2020
Lectured by:
Codete
Workshop 2
EXPLAINABLE AI
2:00 PM - 6:00 PM (CET)
Description:
In the days where we have autonomous cars, drones, and automated medical diagnostics, we want to learn more about how to interpret the decisions made by the machine learning models. Having such information we are able to debug the models and retrain it in the most efficient way.
This training is dedicated to managers, developers and data scientists that want to learn how to interpret the decisions made by machine learning models.
We explain the difference between white and black box models, the taxonomy of explainable models and approaches to XAI. Knowing XAI methods is especially useful in any regulated company. We go through the basic methods like the regression methods, decision trees, ensemble methods, and end with more complex methods based on neural networks.In each example, we use a different data set for each example. Finally, we show how to use model agnostic methods to interpret it and the complexity of the interpretability of many neural networks.
Target Audience:
- Managers who want to understand the decisions made by machine learning models.
- Data scientists who have little or no experience with XAI.
- Python developers who would like to extend their knowledge on machine learning.
- Developers who want to explain to the managers or other audience how the model she/he created works under the hood.
Participants will understand…
- the difference between white and black-box machine learning models,
- what type of models are easier to explain and why it’s easier to explain such,
- what methods are used to explain machine learning models.
Participants will be able to…
- use model agnostic methods,
- interpret machine learning models.
Outline
- What is XAI?
- Differences between black and white-box models
- Models that are easy to explain
- Model-agnostic methods
- Explaining deep neural networks
Lectured by:
Codete
Workshop 3
LOW CODE AI ON GOOGLE CLOUD PLATFORMY
10:00AM - 11:30 AM (CET)
Description:
Advanced AI models need experienced data scientists with a deep knowledge about modeling area. Moreover, building and testing can take a lot of time, when problem to solve is very complicated. During this workshop we will present 5 tools delivered by Google Cloud Platform which makes it possible to build AI models with almost no code. Google Cloud Platform allows not only to build simple models by no-data scientist but also save time of experienced ML scientists because they can build prototypes faster and test assumptions before starting big project.
• ML APIs
• AutoML Vision
• AutoML Tables
• BigQuery ML
• Dialogflow
Lectured by:
Chmura Krajowa
Chmura Krajowa
Workshop 4
MACHINE LEARNING SECURITY
9:00 AM - 1:00 PM (CET)
Description:
Neural networks are currently the most popular machine learning methods. One type of use cases are based on pattern recognition on images. Neural networks as any other solution is liable to security issues. In this training we go through potential leaks and vulnerabilities of neural networks. This training is dedicated to managers and data scientists that want to learn more on how to find leaks and secure a neural network.
In the first part of the training, we start with some examples of simple adversarial attacks and show how to generate simple images with noise that change the prediction of a network. The second part is on attack taxonomy. We explain different types of attacks and weaknesses of networks. This includes targeted and untargeted attacks. We show how to find out the vulnerability using white and black-box methods. The third section is on environment and how to inverse the gradient descent methods. We go deep into the math details of a gradient descent attack. Finally, we show the details of defense methods.
Each section ends with a simple exercise, we have four small exercises written in Python. The examples are developed in Tensorflow and Keras.
Target Audience:
- Data scientists who wants to understand how to secure a neural network
- Managers who want to understand what kind of attacks and defence methods are available
- Security specialists that want to extend the knowledge in the area of machine learning security
Expected Outcomes:
- Participants will understand…
- what a adversarial attack is,
- how to secure a neural network.
- Participants will be able to…
- find leaks in their networks,
- Defends against adversarial attacks.
Outline:
- Adversarial attacks
- Attack types and examples
- Inverse the gradient descent method
- Defence methods
Lectured by:
Codete
Workshop 5
BUILDING AND OPERATING AN OPEN SOURCE DATA SCIENCE PLATFORM
2:00 PM - 6:00 PM (CET)
Description:
There are many great tutorials for training your deep learning models using TensorFlow, Keras, Spark or one of the many other frameworks. But training is only a small part in the overall deep learning pipeline. This workshop gives an overview into building a complete automated deep learning pipeline starting with exploratory analysis, over training, model storage, model serving, and monit5ring and answer questions such as:
How can we enable data scientists to exploratively develop models?
How to automatize distributed Training, Model Optimization and serving using CI/CD?
How can we easily deploy these distributed deep learning frameworks on any public or private infrastructure?
How can we manage multiple different deep learning frameworks on a single cluster, especially considering heterogeneous resources such as GPU?
How can we store and serve models at scale?
What Metadata should be stored in a production setup?
How can we monitor the entire pipeline and track performance of the deployed models?
Target Audience:
- Data scientists who wants to extend their knowledge in applied ML and want to know how to build a ML pipeline using open source tools
- Software Engineers that starts the journey in Machine Learning
- DevOps that want to extend their knowledge in the MLOps directionOr Developers, Data Scientist, and anyone interested in Machine Learning (Operations)
Expected Outcomes:
1. Participants will understand…
- the challenges of building and operating Machine Learning Pipelines
- the different Open Source tools for building such Pipelines
- implementation strategies for stepwise implementations of Pipelines and Best Practices around
2. Participants will be able to…
The participants will build an end-to-end data analytics pipeline including:
- Pipeline Orchestration with TFX, Kubeflow, and Airflow
- Data preparation using Apache Spark
- Jupyter Notebooks
- Distributed training with TensorFlow
- Automation & CI/CD using Jenkins and Argo
- Model and metadata storage
- Model serving and monitoring
Outline:
*see description*
Readings
Some basic understanding of Machine Learning is helpful, but not required.
- TensorFlow Tutorial: https://www.tensorflow.org/tutorials/quickstart/beginner
- Introduction to TensorFlow Extended https://www.tensorflow.org/tfx
Lectured by:
ArangoDB