Summary of Course Highlights

A computer program that analyzes given data to learn patterns and rules, subsequently making predictions and decisions based on the new data.

The process of retraining an already trained AI model to improve its performance for specific tasks or purposes.

It means extracting features from numerous examples to learn patterns and developing the ability to process new data accurately.

Technical Explanation: It involves creating an algorithm (a step-by-step procedure for performing tasks) that determines the output for newly inputted data.

AI adjusts its parameters, including weights and biases, to learn the data patterns and make predictions for new data.

Copy
y = w1x1 + w2x2 + ... + wnxn + b

A lightweight data format used for storing and exchanging data.

JSON consists of objects and arrays, with objects enclosed in curly braces { } and arrays in square brackets [ ].

Copy
// An array enclosed in square brackets
[
     // Objects enclosed in curly braces
    {  
        "name": "John Doe",
        "math": 85,
        "english": 90
    },
    {
        "name": "Jane Smith",
        "math": 88,
        "english": 80
    }
]

A JSONL file used for fine-tuning stores one JSON formatted data entry per line.

A collection of data gathered and organized for specific purposes, such as AI model training and validation.

• Training Dataset: About 60-80% of the total data

• Validation Dataset: About 10-20% of the total data

• Test Dataset: About 10-20% of the total data

A function that measures the difference between model predictions and actual values, taking predictions and actual values (correct answers) as input and returning the loss (error) as output.

A smaller value of the loss function indicates that the model's predictions are closer to the actual values, and the goal of AI training is to minimize the value of the loss function.

The gradient of the loss function, known as the gradient, informs how to adjust the model's parameters to reduce the loss.

Parameters (settings) established when training an AI model.