Foundation Day Meeting :
About AI :
AI helps in
1. Climate change
2. Disaster management 3. Autonomous driving, computer vision, health care, finance
How AI Language models to be developed :
Ethical ai :
1. Fairness
2. Ethics ( moral principles )
3. Moral system websites
4. Historical data matters most ( based on that large data , ai takes decision ).
5. AI - ML algorithms build and learn from data
6. Regression model (continuous thing).
Compas :
About Criminal Case
Recidivist is the act of a person repeating an undesirable behavior after they have experienced negative consequences of that behavior, or have been trained to extinguish it.
Federated learning :
1. Fine-tune models
Fine-tuning in machine learning refers to the process of taking a pre-trained model and adjusting it on a new, often smaller dataset to improve its performance for a specific task. This is commonly done by:
- Loading a Pre-Trained Model – A model, typically trained on a large dataset (e.g., ImageNet for vision tasks, GPT for NLP), is used as a starting point.
- Freezing Some Layers – Lower layers often capture general patterns, so they might remain unchanged while higher layers are modified.
- Training on New Data – The model is trained on a smaller dataset with a lower learning rate to prevent catastrophic forgetting.
- Adjusting Hyperparameters – Learning rate, batch size, optimizer settings, and regularization techniques are fine-tuned for optimal performance.
Fine-tuning is widely used in transfer learning for tasks like image classification, natural language processing (NLP), and speech recognition, reducing training time while achieving high accuracy.
ML model is fair or not :
Preprocessing
About fairness : What if tool = Google
Utility :
Laplace mechanism
Rappor noise data : about privacy
Cynthia documents on AI privacy.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Comments
Post a Comment