Chapter 2: Understanding Foundation Models Overview Foundation model design choices (training data, architecture/size, post-training) are increasingly opaque. The training process splits into pre-training (makes model capable) and post-training (aligns model to human preferences). Sampling (how outputs are chosen from all possibilities) is a crucial, often-underestimated factor impacting model behavior and Read more
Chapter 1 Summary: Introduction to Building AI Applications with Foundation Models 1. The Scaling Up of AI Models Foundation models like ChatGPT and Gemini are highly powerful but require enormous compute resources and vast amounts of data. Only a few organizations have the capability to train these models, leading to Read more
Link to paper https://arxiv.org/abs/2308.10053 Notes CRS possesses the potential to: (1) understand not only users’ historical actions but also users’ (multi-turn) natural-language inputs; (2) Provide not only recommended items but also human-like responses for multiple purposes, such as preference refinement, knowledgeable discussion, or recommendation justification.Towards this a typical conversational recommender Read more
Link to paper https://arxiv.org/abs/2308.14296 Notes The paper propose a novel algorithm, Self-Inspiring, to improve the planning ability of the LLM agent. At each intermediate planning step, the LLM “self-inspires” to consider all previously explored states to plan for next step. Literature survey Architecture Tools they used 1) DB tool1) To Read more
Executive Summary Week 1: Introduction Let’s say you have a data set with six houses, so you know the size of the houses in square feet or square meters and you know the price of the house and you want to fit a function to predict the price of a house as Read more
Amid the AI revolution, diverse AI models like large language and generative AI models have come into the limelight. These novel AI models require efficient data processing, achievable using vector embeddings. By providing semantic information to the AI models, they gain a better understanding and can perform complex tasks efficiently. Read more
Introduction to Building Machine Learning Applications Building machine learning applications requires a thorough understanding of the fundamentals of machine learning and software development. This section will provide an overview of the key considerations and best practices for building machine learning applications. Best Practices for Data Preprocessing Data preprocessing is a Read more
Median Absolute deviation is one of the other techniques specifically used for analyzing the performance of regression models. Computing MAD of errors For a univariate data set X1, X2, …, Xn, the MAD is defined as the median of the absolute deviations from the data’s median, X_median = median(X) Median Read more
R-squared is a statistical measure of how close the data are to the fitted regression line. It is also known as the coefficient of determination, or the coefficient of multiple determination for multiple regression. This metric is specifically designed for regression-based algorithms where the output is a real value. Computing Read more
We can plot error distributions like probability density functionand cumulative density function and make important deductionsbased on it. We can use plot Probability Density functions(PDF) and Cumulative density function (CDF) by using the error function as a random variable Using PDF of error distribution An ideal pdf for error distributions Read more