Machine Learing

1. machine learnimng is the subset of artificial intelligence that focus on the development of Algorithms and statistical model that enables computer to imporove their performance for specific tasks.
2. Machine learning uses data and try to find patterns, make decisions or predictions on new scenarios.

Machine Learning types -

traditional 2 types of Machine learning.

1. Supervised learning - Dataset needed for trainining model should be labelled, Regression, classification.
2. Unsupervised learning - Dataset should be unlablled. Finding patterns and clustering.
3. Reinforcement learning - Agent intract with environment and each step having rewards and penalities. In this manner model learns to maximize the cumulative reward over time.

Labellled Dataset - When we pass training data that contain outputs for specific inputs, and test the models on new datapoints.
Unlabelled Dataset- When data is not contain any output and algorithm try to find hidden pattern in data.

Examples

• Given a dataset of images with labeled categories (e.g., cats and dogs), train a model to predict the category of a new, unseen image. - Supervised
  
• Analyze a dataset of customer purchase history without predefined categories and identify natural groupings of customers based on their purchasing behavior. - Unsupervised
  
• Train an agent to play a game by making a sequence of moves to maximize the cumulative score. (Rewards/ penalities) - Reinforcement learning
  

Vector Embeddings

1. It is a technique in Machinelearnig where each words assigned with unique rational number, and plotted in 2D arrays. Words which are related to eachother in some manner, placed in near to the words.
2. Words that are related to each other in some way are placed close to each other on the graph. For example, words like "India," "Kohli," "Narendra Modi," and "India Gate" would have nearby values like .222288, .222287, .222254, etc., indicating their semantic proximity. Similarly, words like "men," "boy," "actor," and "engineer" would have close values.
3. A different type of Database use for saving these vast amount of records known as Vector DBs.

Example -

• Google search
• Recommendation System
• Text summarizartion, Translation
• ChatGPT

Here is how words are grouped with unique values

Screenshot from deeplearning vector embedding visualizer