Machine learning for beginners
Machine learning course for absolute beginners.
This course is designed for absolute beginners in Machine Learning (ML). No prior knowledge in ML is required. A beginner level python skill is a prerequisite to understand the code snippets. We will also use python packages like numpy, matplotlib which you can learn easily by reviewing the code in this course.
This is a code-first course; we will implement the techniques and algorithms discussed using python. This helps in understanding the concepts better.
Lecture 1. Machine Learning Intuition:
This lecture provides a basic intuition behind many Machine Learning (ML) algorithms.
When we write software, for example a sorting algorithm, we will explicitly write step-by-step instructions in a programming language so that the computer can execute them to achieve the desired results. However, for a class of problems such as predicting stock prices, identifying objects in an image etc., writing instructions or creating rules to solve them is very brittle and very hard to achieve even by domain experts.
Instead, we will design algorithms that ‘learns’ the rules from the historical data on their own. In short, we teach the machines how to learn from the data themselves. Please read the below notebook and feel free to play around with the code provided.
Lecture 2. Vectorization and Broadcasting:
Vectorization helps to optimize the cost of model training. Broadcasting is an important aspect to get right when vectorizing the ML operations, any mistakes are very difficult to identify and debug.
Lecture 3. Classification - Logistic Regression:
Classification is a bread and butter task in Machine Learning. Given a few categories, we train a model using a dataset, that predicts the category/class of the new data point.
As usual, we will define a toy dataset and build out the complete logistic classification model from scratch in python. We will leverage cross entropy loss function to train the model.
More to follow…
Comments