Machine Learning

This tutorial section contains learning material on programmatically training, managing and deploying machine learning models in Dataiku.

Local Interpretable Model-agnostic Explanations

• This tutorial explains using LIME (Local Interpretable Model-agnostic Explanations) to provide human-readable explanations for machine learning model predictions.

Predictive maintenance

• This tutorial explains how can you predict performance before getting the ground truth.

Reinforcement learning

• This tutorial uses reinforcement learning (RL) to tune a random forest classifier’s hyperparameters automatically. The Q-learning algorithm explores and exploits hyperparameter combinations to find the best combination, using validation accuracy as the reward.

Transfer learning

Transductive transfer learning

This tutorial focuses on scenarios where labeled target data is available, the source and target tasks are the same, but the domain changes.

Unsupervised transfer learning

This tutorial addresses the challenge of data scarcity. Unsupervised transfer learning techniques are helpful for adapting a model to a new domain when you have a target dataset, but no labels.

Experiment Tracking

• With XGBoost and custom pre-processing

• With Keras/Tensorflow for sentiment analysis

• With Catboost

• With LightGBM

• With scikit-learn

Pre-trained Models

• Tensorflow

• PyTorch

• Hugging Face

• SentenceTransformers

• spaCy

• NLTK

Model Import

• Importing serialized scikit-learn pipelines as Saved Models for MLOps

Model Export

• Wrapping an exported model in a CLI tool

• Deploying a model to an edge device with AWS IoT Greengrass

Distributed training

• Distributed ML Training using Ray

Vulnerability and Bias Scanning with Protect AI Guardian

This tutorial will guide you through the process of scanning a model for vulnerabilities, biases, and security concerns using Protect AI Guardian’s Python SDK.