This course covers a series of practices for preparing and working with data for training and running contemporary AI systems and neural networks. It further provides techniques for designing and optimizing neural networks, including approaches for measuring and tuning neural network model performance. The practices and techniques are documented as design patterns that can be applied individually or in different combinations to address a range of common AI system problems and requirements. The patterns are further mapped to the learning approaches, functional areas and neural network types that were introduced in Module 1: Fundamental Artificial Intelligence.
The following primary topics are covered:
- – Data Wrangling Patterns for Preparing Data for Neural Network Input
- – Feature Encoding for Converting Categorical Features
- – Feature Imputation for Inferring Feature Values
- – Feature Scaling for Training Datasets with Broad Features
- – Text Representation for Converting Data while Preserving Semantic and Syntactic Properties
- – Dimensionality Reduction to Reduce Feature Space for Neural Network Input
- – Supervised Learning Patterns for Training Neural Network Models
- – Supervised Network Configuration for Establishing the Number of Neurons in Network Layers
- – Image Identification for using a Convolutional Neural Network
- – Sequence Identification for using a Long Short Term Memory Neural Network
- – Unsupervised Learning Patterns for Training Neural Network Models
- – Pattern Identification for Visually Identifying Patterns via a Self Organizing Map
- – Content Filtering for Generating Recommendations
- – Model Evaluation Patterns for Measuring Neural Network Performance
- – Training Performance Evaluation for Assessing Neural Network Performance
- – Prediction Performance Evaluation for Predicting Neural Network Performance in Production
- – Baseline Modeling for Assessing and Comparing Complex Neural Networks
- – Model Optimization Patterns for Refining and Adapting Neural Networks
- – Overfitting Avoidance for Tuning a Neural Network
- – Frequent Model Retraining for Keeping a Neural Network in Synch with Current Data
- – Transfer Learning for Accelerating Neural Network Training
Duration: 1 Day
Taking the Course at a Workshop
This course can be taken as part of instructor-led workshops taught by Arcitura Certified Trainers. These workshops can be open for public registration or delivered privately for a specific organization. Certified Trainers can teach workshops in-person at a specific location or virtually using a video-enabled remote system, such as WebEx.
Visit the Workshop Calendar page to view the current calendar of public workshops.
Visit the Private Training page to learn more about Arcitura’s worldwide private workshop delivery options.
Taking the Course via an eLearning Study Kit
This course can be completed via self-study by purchasing an eLearning study kit subscription, which includes online video lessons, as well as online and offline access to the electronic course materials and additional supplements and resources designed for self-paced study and exam preparation.
Visit the eLearning Study Kits page for more information about eLearning study kits.
Visit the Arcitura online store for purchasing information.
Taking the Course via a Printed Study Kit
This course can be completed via self-study by purchasing a printed study kit, which includes the full-color course materials as well as additional supplements and resources designed specifically for self-paced study and exam preparation.
Visit the Printed Study Kits page for more information about full-color printed study kits.
Visit the Arcitura online store for purchasing information.
Certifications
This course is one of three courses that are used to prepare for Exam AI90.01. A passing grade on this exam is required to achieve the Artificial Intelligence Specialist certification.