- Assuring quality, accuracy, model resilience, and more characteristics in cases of AI business solution development, and guaranteeing performance, and outcomes. Interaction AI Use Cases.
- Level -Basic - Choice - Specialisation
- Prerequisites - Introduction to Interaction AI or any introduction course
- Assessments - Micro Assessment weekly, Full Assessment last week.
Week 1 - Key Failure Areas of An AI Algorithm
Learning OutcomeLearning Outcome Students should be able to understand why, where, and how AI Algorithms fail.
1.1 Failure on Accuracy - Data, Training, Model Selection or More.
1.2 Failure on Resilience - Understanding the cause.
1.3 Failure on usability, response time, all scenario applicability, and more
Practical Component
Practical demonstrations on various models and created solutions that fail on these various accounts. How do we detect, highlight, and set guardrails for preventing these failures?
Week 2 -Solutions for Accuracy
Learning OutcomeLearning Outcome Achieving accuracy is the most fundamental need of AI, as inaccurate AI has very low value.
2.1 Data Flaw Identification and Fixes.
2.2 Model selection nuances that can arise in problems.
2.3 Test set flaws, complete knowledge capture flaws
Practical Component
Understand major accuracy issues that arise in trained algorithms, learn to detect and identify reasons for flaws.
Week 3- Solutions for Resilience
Learning OutcomeLearning Outcome Ensure algorithms predict the correct outcome independent of minor variations in inputs, and do not jump to alternate conclusions incorrectly.
3.1 Understand the nature of variation and failed cases.
3.2 Alter training and test sets to achieve better resilient intelligence.
3.3 Add additional layers of models or alternate models to address issues
Practical Component
Practical demonstrations showcasing the resilience component of models to achieve better and more effective fit to data and knowledge.
Week 4 - Solutions for Response Time
Learning OutcomeLearning Outcome Addressing speed and overuse of compute issues in algorithms, with prominent issues in consecutive real-time input responses.
4.1 Understand whether the issue is model-related, and explore compression or structure changes.
4.2 Capture features, and data differently.
4.3 Capture features, and data differently.
Practical Component
Effects of changing various parameters on model performance in accuracy and speed. Understanding tradeoffs and optimal size.
Week 5 - Comprehensive Test Planning Strategy
Learning OutcomeLearning Outcome Plan a complete testing strategy for an AI project to achieve minimum time to live.
5.1 Set expectations and exact working outcomes of the solution.
5.2 Define test cases to ensure the model surpasses accuracy, resilience, and response time.
5.3 Plan and advise on potential algorithm structures and data choices for faster outcome realization.
Practical Component
Apply test strategies to various algorithms, ensuring effectiveness across use cases and efficiency.
Week 6- Project and Assessment
Learning OutcomeLearning Outcome Students will plan, test, and analyze Interaction AI models to determine exact flaws and suggest improvements.
6.1 Planning a testing strategy with improvement suggestions.
6.2 Evaluating models practically against test cases and selected strategies.
6.3 Final presentations and outcome scoring
Practical Component
Student-created testing plans applied to various algorithms, feedback, and presentation.