Machine Learning :

ML Jargons:

• Data Augmentation: Data augmentation involves creating new training examples by applying - transformations like **rotations**, **flips**, or color changes to existing data. This process is typically used to increase the diversity of the training dataset and improve the model’s robustness. It is not related to using the model for predictions on new data.
• Named Entity Recognition: NLP technique used for text and not images.
• Support Vector Machines: Used for classification tasks.SVMs work well when you have a lot of features (dimensions) compared to the number of data points. For example, in text classification (e.g., spam detection), where every word can be a feature.
• Inpainting: Computer vision technique to restore damaged images.
• Object Detection: Computer vision technique to detect objects in images by creating boxes around them.
• CFG Scale can be increased to increase the specificity of the images and this can be tuned to increase or decrease randomness in image generation.
• Number of Generation steps: image increase the image quality and not specificity.
• XGBOOST: Is used for predicting data and not classification tasks.
• Sagemaker DEEP-AR Algorithm: Used for forecasting values in sagemaker. Works on time-series data.
• Adversarial Prompting: To expose vulnerability in AI by simulating prompt injection attacks.
• Least-to-most prompting is a teaching or problem-solving strategy where minimal guidance is provided initially, and gradually more explicit help is given only if needed to achieve the desired outcome.
• RESNETS are used for deep learning in image recognition and they contain skip connections not used for DATA GENERATION
• Wavenets is a generative synthesis speech.
• Ongoing pre training increases the performance of the model but does not improve the inference time. It just trains the model on the data.
• Inference time depends on the user prompt and input tokens and output tokens depending on how long the response given by the model.
• Confusion Matrix: Used for evaluation of classification tasks by showing true positives, true negative, false positives and false negatives.
• Correlation Matrix: Shows relationship between different variables or features and not used for performance evaluation for classification.