12-Week + Project Study Plan for AI/ML:
Module 1: Introduction to the AI Landscape (1 Week)
- Topics: Overview of AI, Data Analysis, Prediction, Decision Making, Causal Inference
- Activities: Lectures, readings, self-study
Module 2: Data Exploration - Structured Data (2 Weeks)
- Topics: Data Exploration Techniques, Dimensionality Reduction (PCA), Clustering (K-means, DBSCAN)
- Activities: Hands-on exercises, data analysis projects, self-study
Module 3: Prediction Methods - Regression (2 Weeks)
- Topics: Linear Regression, Model Evaluation, Regularization, Overfitting
- Activities: Hands-on exercises, building regression models, self-study
Module 4: Decision Systems (2 Weeks)
- Topics: Decision Trees, Classification, Ensemble Learning (Random Forests)
- Activities: Hands-on exercises, building decision trees and random forests, self-study
Module 5: Data Exploration - Unstructured Data (2 Weeks)
- Topics: Natural Language Processing (NLP), Sentiment Analysis
- Activities: Hands-on exercises, analyzing text data, building sentiment analysis models, self-study
Module 6: Recommendation Systems (1 Week)
- Topics: Collaborative Filtering, Recommendation System Techniques
- Activities: Guest lecture, case studies, self-study
Module 7: Data Exploration - Temporal Data (1 Week)
- Topics: Time Series Analysis, Stationarity, Time Series Forecasting
- Activities: Guest lecture, case studies, self-study
Module 8: Prediction Methods - Neural Networks (2 Weeks)
- Topics: Neural Networks Architecture, Forward Propagation, Backpropagation, Optimization
- Activities: Hands-on exercises, building and training neural networks, self-study
Module 9: Computer Vision Methods (2 Weeks)
- Topics: Convolutional Neural Networks (CNNs), Image Processing, Feature Extraction
- Activities: Hands-on exercises, building and training CNNs, self-study
Module 10: Workflows and Deployment (1 Week)
- Topics: Industry Case Studies, Practical Applications of AI/ML
- Activities: Guest lectures, group discussions, self-study
Projects:
- Insurance Claim Prediction (2 Weeks)
- Tools & Concepts: Linear Regression, Model evaluation, Tuning, Exploratory Data Analysis, Python
- Forecasting Monthly Sales of French Champagne (2 Weeks)
- Tools & Concepts: Time Series Analysis, Predictive Modelling, Python for Time Series
Self-study:
- Allocate time each week for self-study, reviewing concepts, and completing additional exercises.
- Utilize online resources, tutorials, and books to supplement your learning.
Additional Notes:
- Consider your learning style and adjust the pace of the study plan as needed.
- Join online communities and forums to connect with other AI/ML enthusiasts.
- Participate in hackathons and challenges to apply your knowledge and gain experience.
- Continuously update your skills and stay informed about the latest advancements in AI/ML.
I have created more details below.
Module 1: Introduction to the AI Landscape
To offer a general overview of the four blocks upon which this No Code AI and Machine Learning Program is focused.
Understanding the data: What is it telling us?
Prediction: What is going to happen?
Decision Making: What should we do?
Causal Inference: Did it work?
Module 2: Data Exploration - Structured Data
To learn the basic principles of applying data exploration techniques, such as dimensionality projection and clustering on structured data.
Asking the right questions to understand the data.
Understanding how data visualization makes data clearer.
Performing Exploratory Data Analysis using PCA.
Clustering the data through K-means & DBSCAN clustering.
Evaluating the quality of clusters obtained.
Module 3: Prediction Methods - Regression
To understand the concept of linear regression and how it can be used with historical data to build models that can predict future outcomes.
The idea of regression and predicting a continuous output.
How do you build a model that best fits your data?
How do you quantify the degree of uncertainty?
What do you do when you don’t have enough data?
What lies beyond linear regression?
Module 4: Decision Systems
To understand the concept of classification and understand how tree-based models achieve prediction of outcomes that fall into two or more categories.
Understand the Decision Tree model and the mechanics behind its predictions.
Learn to evaluate the performance of classification models.
Understand the concepts of Ensemble Learning and Bagging.
Learn how Random Forests aggregate the predictions of multiple Decision Trees.
Module 5: Data Exploration - Unstructured Data
To understand the concept of Natural Language Processing and how natural language represents an example of unstructured data, the business applications for this kind of data analysis, and how data exploration and prediction are performed on natural language data.
Understand the concept of unstructured data, and how natural language is an example.
Understand the business applications for Natural Language Processing.
Learn the techniques and methods to analyze text data.
Apply the knowledge gained towards the business use case of sentiment analysis.
Module 6: Recommendation Systems
To understand the idea behind recommendation systems and potential business applications.
Learn the concept of recommendation systems and potential business applications.
Understand the sparse data problem that necessitates recommendation systems.
Learn about potentially simple solutions to the recommendation problem.
Understand the ideas behind Collaborative Filtering Recommendation Systems.
Module 7: Data Exploration - Temporal Data
To understand the critical concept of temporal data, and its differences from structured and unstructured data, the idea behind Time Series Forecasting and the preprocessing required to obtain stationarity in Time Series.
Understand temporal data and how it represents a different data modality.
Understand the idea behind Time Series forecasting
Learn about the concept of Stationary Time Series, testing for stationarity and conversion techniques to transform non-stationary time series into stationary.
Module 8: Prediction Methods - Neural Networks
To understand the ideas behind Neural Networks, their introduction of non-linearities into the encoding and predictive process through a hierarchical structure, and the various steps involved in their forward propagation and back propagation cycle to minimize prediction error.
Understand the key concepts involved in Neural Networks.
Learn about the encoding process taking place in the neural network layers, and how non-linearities are introduced.
Understand how the forward propagation happens through the layered architecture of neural networks and how the first prediction is achieved.
Learn about the cost function used to evaluate the neural network’s performance, and how gradient descent is used in a backpropagation cycle to minimize error.
Understand the critical optimization techniques used in gradient descent
Module 9: Computer Vision Methods
To understand how images represent a spatial form of unstructured data and hence, a different data modality, how the Convolutional Neural Network (CNN) structure achieves generalized encoding abilities from image data and acquire an understanding of what CNNs learn.
Learn about spatial concepts of images such as locality and translation invariance.
Understand the working of filters and convolutions, and how they achieve feature extraction to generate encodings.
Learn about how these concepts are used in the structure of Convolutional Neural Networks (CNNs) and understand what CNNs actually learn from image data.
Module 10: Workflows and Deployment
To obtain additional perspective on how the same takeaways from the conceptual modules discussed prior have been applied in various business scenarios and problem statements by industry leaders who have achieved success in practical applications of Data Science and AI.
Hands on project:
Insurance Claim Prediction
A key challenge for the insurance industry is to charge each customer an appropriate premium for the risk they represent. In this project, we will build a regression model to predict the cost of insurance claims using user information on age, gender, bmi, blood pressure, health conditions, as well as insurance claim details.
Tools & Concepts: Linear Regression, Model evaluation, Tuning, Exploratory Data Analysis, Python
Forecasting Monthly Sales of French Champagne
Being able to make accurate predictions of future revenue can be hugely important for businesses. This project will focus on forecasting the next monthly revenue of a french chamapagne brand, which will inform the decision-making process across all areas of the business, from purchasing decisions and marketing activity to staffing levels.
Tools & Concepts: Time Series Analysis, Predictive Modelling, Python for Time Series
No comments:
Post a Comment