Trang 1
AI Saturdays Lesson Plan Here are learning roadmaps organised in weeks. Each week consists of lecture videos, reading materials and programming assignments. Feel free to head to the AI6 forums to discuss or ask questions you might have. Machine Learning from Scratch This is for anyone who wants to learn Machine Learning or TensorFlow. The only prerequisite is high school math and a ton of passion. We will follow Machine Learning Crash Course by Google - it breaks down concepts of Machine Learning in fun, bite-sized portions and provides thoughtful exercises to deepen your understanding. Timeline Learning material Assignments Week 1/ Prerequisites: Linear Algebra Assignments and Week of 4 readings August - Khan academy Linear Algebra (Finish Complete 2018 Vectors and Spaces and Matrix Transformations modules) - Still hungry for more? Try Essence of linear algebra Week 2/ Prerequisites: Python basics Finish 5 python Week of 11 exercises here August For those without any programming experience: (solutions provided) 2018 - Automate the boring stuff with python: Chapter 0 - Python for Data Science (7 hours) For those with programming experience: - Automate the boring stuff with python: Chapters 1 - 5 Week 3/ Prerequisites: Python advanced Replicate the codes in Week of 18 “A Byte of Python” August 2018 - A Byte of Python ebook: Object oriented Create a Cheat Sheet programming summarising all that - Udemy Numpy stack in Python: Numpy, you have learnt about Matplotlib, Scipy, Pandas Python. - MIT Course Object Oriented Programming Week 4/ Recommended reading: Complete “Check Your Week of 25 - Machine Learning is Fun! Understanding”
Trang 2
August 2018 Discuss: Artificial Machine Learning Crash Course by Google: Intelligence, Machine Learning, Deep - Introduction to ML Learning, Data Science - Framing - Descending into ML Week 5/ Machine Learning Crash Course by Google: Complete “Check Your Week of 1 Understanding” September - Reducing Loss 2018 Code up a simple gradient descent algorithm on a linear regression problem (example). Linear regression from scratch Another Linear regression from scratch in Python Week 6/ This week, explore different frameworks for Programming exercise Week of 8 Machine Learning. Recommended readings: with tensor flow September - What are Machine Learning Frameworks 2018 Official scikit-learn - Best Python libraries for Machine Learning tutorial and Data Science - TensorFlow or Keras? - Introduction to scikit-learn - Machine Learning Crash Course by Google: First Steps with TensorFlow Note: Don’t spend too much time on picking a framework. If you cannot decide, try out TensorFlow first, since assignments offered in the Crash Course are in TensorFlow.
Trang 3
Week 7/ Machine Learning Crash Course by Google: Complete “Check Your Week of 15 - Generalization Understanding” and September “Programming 2018 - Training and Test Sets Exercises”, if any Validation - Representation Week 8/ Machine Learning Crash Course by Google Complete “Check Your Week of 22 Understanding” and September - Feature Crosses “Programming 2018 Exercises”, if any Week 9/ Machine Learning Crash Course by Google Complete “Check Your Week of 29 Understanding” and September - Regularization: Simplicity “Programming 2018 Exercises”, if any - Logistic Regression Logistic Regression with Tensor Flow Logistic Regression from scratch in Python Titanic challenge using sklearn Alternatively, find your own dataset to perform logistic regression Week 10/ You are at the halfway mark! Use this week to take a good break and re- Week of 6 energize. October 2018 Week 11/ Machine Learning Crash Course by Google Complete “Check Your Week of 13 Understanding” and October - Classification “Programming 2018 Exercises”, if any - Regularization: Sparsity AUC curve: discussion and analogy
Trang 4
Evaluating a classification model (using scikit learn) K-means clustering algorithm (hands-on assignment) Week 12/ Complete “Check Your Week of 20 Video Lecture: What is a Neural Network? Understanding” and October “Programming 2018 Machine Learning Crash Course by Google Exercises”, if any - Introduction to Neural Nets Create a Neural Network using Keras Implementing a Neural Network from Scratch in Python Week 13/ Video Lectures: Complete “Check Your Week of 27 - Gradient descent Understanding” and October “Programming 2018 - What is backpropagation really doing? Exercises”, if any Machine Learning Crash Course by Google Tutorial on Backpropagation - Training Neural Nets (highly recommended) - Multi-Class Neural Nets Week 14/ Machine Learning Crash Course by Google Complete “Check Your Week of 3 Understanding” and November - Embeddings “Programming 2018 Exercises”, if any - All sections under “ML Engineering” and “ML Real World Examples” Deep Learning Specialization Lesson Plan This lesson plan is for those who have understood the basics of Machine Learning and would like to gain a deeper understanding in Deep Learning. It follows deeplearning.ai’s
Trang 5
Deep Learning Specialization on Coursera. This course is one of the highest rated courses on Deep Learning. Neural Networks and Deep Learning (Weeks 1-6) Period Youtube lecture videos (each video is Assignment labelled with an alphanumeric string, e.g C1W1L01) Week 1/ 4 Revision on Python August 2018 Week 2/ C1W1L01 - C1W1L06 Assignment 1 11 August 2018 Week 3/ C1W2L01 - C1W2L06 Assignment 2 18 August 2018 Week 4/ C1W2L07 - C1W2L18 Assignment 3 25 August 2018 Week 5/ C1W3L1 - C1W3L11 Assignment 4 1 Septembe r 2018 Week 6/ C1W4L01 - C1W4L18 Assignment 5 8 Septembe r 2018 Week 7/ Break 15 Septembe r 2018 Improving deep neural networks: hyperparameter (Weeks 8-13) Period Youtube lecture videos (each video is Assignment labelled with an alphanumeric string, e.g C2W1L01)
Trang 6
Week 8/ C2W1L01 - C2W1L06 Assignment 1 22 Septembe r 2018 Week 9/ C2W1L07 - C2W1L14 Assignment 2 29 Septembe r 2018 Week 10/ C2W2L01 - C2W2L05 Assignment 3 6 October 2018 Week 11/ C2W2L06 - C2W2L09 Assignment 4 13 October 2018 Week 12/ C2W3L01 - C2W3L05 Assignment 5 20 October 2018 Week 13/ C2W3L06 - C2W3L11 Assignment 6 27 NOTE: C2W3L10 and C2W3L11 are October mislabeled as C2W2L10 and C2W2L11 2018 respectively. Research-focused Lesson Plan Learners of this track are those who have acquired the prerequisites to understand and implement state of the art AI models. They can choose from one or more of the following courses : UCL course on Reinforcement Learning (UCL RL), Berkeley CS294, Stanford Natural Language Processing (CS224n) and Stanford Computer Vision (CS231n). These courses are offered by their respective universities and (fortunately for us) are freely accessible by the public. Learning materials for the courses can be quite ‘scattered’, i.e lecture videos for these courses are not found on the official website but on youtube; they are also not aligned with the course outline on the official website. We will follow one lecture video from youtube every week and refer to the official pages for lecture notes or assignments. Important links UCL RL CS231n (Computer vision) CS224n (NLP)
Trang 7
● Lecture Videos from ● Lecture Videos from ● Lecture Videos from youtube youtube youtube ● Official Page ● Official Page ● Official Page ● Code and exercises ● Assignments ● Lecture Notes (click on ● Berkeley CS294 ‘notes1.pdf’ or similar) (advanced RL) ● Independent Review of this course Lesson schedule Week of UCL RL CS231n (Computer CS224n (NLP) vision) 4 August Revision of prerequisites: See “Module 0” See “Suggested readings 2018 Python, Linear Algebra etc on Jan 9” Assignments 11 August UCL/Deep Mind Youtube lecture video: Youtube lecture video: 2018 Reinforcement Learning Image Classification Word Vector Lecture 1 – Intro to Representations:word2v Reinforcement Learning Course Notes ec Assignments/ Links to Assignments Assignments and materials readings 18 August UCL/Deep Mind Youtube lecture video: Youtube lecture video: 2018 Reinforcement Learning Loss Functions and Global Vectors for Word Lecture 2 – Markov Decision Optimization Representation Processes Assignments and readings Lecture Notes Assignments/ Links to materials Assignments and readings 25 August UCL/Deep Mind Youtube lecture video: Youtube lecture video: 2018 Reinforcement Learning Introduction to Neural Word Window Lecture 3 – Planning by Networks Classification and Neural Dynamic Programming Networks Assignment and readings Assignments/ Links to Lecture Notes materials 1 UCL/Deep Mind Youtube lecture video: Youtube lecture video: Septembe Reinforcement Learning Convolutional Neural Backpropogation r 2018 Lecture 4 – Model-free Networks Assignments and Prediction readings Assignments and readings
Trang 8
Assignments/ Links to materials 8 UCL/Deep Mind Youtube lecture video: Youtube lecture video: Septembe Reinforcement Learning Training Neural Networks Dependency parsing r 2018 Lecture 5 – Model-free Part 1 Control Lecture Notes Assignments and readings 15 Break Septembe r 2018 22 UCL/Deep Mind Youtube lecture video: Youtube lecture video: Septembe Reinforcement Learning Training Neural Networks Introduction to r 2018 Lecture 6 – Value Function Part 2 TensorFlow (optional) Approximation Youtube lecture video: Recurrent Neural Networks and Language Models Lecture Notes Assignments and readings 29 UCL/Deep Mind Youtube lecture video: Youtube lecture video: Septembe Reinforcement Learning Deep Learning Software Machine Translation and r 2018 Lecture 7 – Policy Gradient Advanced Recurrent Methods LSTMs and GRUs 6 October UCL/Deep Mind Youtube lecture video: Youtube lecture video: 2018 Reinforcement Learning CNN Architectures Neural Machine Lecture 8 – Integrating Translation and Models Learning & Planning with Attention 13 UCL/Deep Mind Youtube lecture video: Youtube lecture video: October Reinforcement Learning Recurrent Neural Gated Recurrent Units 2018 Lecture 9 – Exploration & Networks and Further Topics in Exploitation NMT 20 UCL/Deep Mind Youtube lecture video: Youtube lecture video: October Reinforcement Learning Detection and End-to-End Models for 2018 Lecture 10 – Classic Games Segmentation Speech Processing
Trang 9
27 Break October 2018 3 Berkeley CS294 Deep Youtube lecture video: Youtube lecture video: November Reinforcement Learning Visualizing and Tree Recursive Neural 2018 Understanding Networks and Constituency Parsing 10 Berkeley CS294 Deep Youtube lecture video: Youtube lecture video: November Reinforcement Learning Generative Models Coreference Resolution 2018 17 Berkeley CS294 Deep Youtube lecture video: Youtube lecture video: November Reinforcement Learning Deep Reinforcement Dynamic Neural 2018 Learning Networks for Question Answering 24 Berkeley CS294 Deep Youtube lecture video: Youtube lecture video: November Reinforcement Learning Efficient Methods and Issues in NLP and 2018 Hardware for Deep Possible Architectures in Learning NLP 1 Youtube lecture video: Youtube lecture video: December Berkeley CS294 Deep Adversarial Examples and Tackling the Limits of 2018 Reinforcement Learning Adversarial Training Deep Learning for NLP Until Work on projects January 2019 Bonus tips - The lecture videos go up to 2 hours; watch the lecture video at 2x speed, slowing down only at parts that cover important concepts - Revise and make sure that you have mastered the fundamentals of deep learning - Don’t aim to understand everything and anything.
Hiển thị 9 trang đầu. Tải tài liệu để xem đầy đủ.
Bình luận (0)
Chưa có bình luận nào. Hãy là người đầu tiên!