- [00:00](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🤖 The course explores foundational concepts and algorithms of modern artificial intelligence, covering topics like graph search algorithms, optimization, reinforcement learning, and more. - [03:16](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🛣️ AI aims to solve problems by searching for solutions using various actions and transitions between states in a state space. - [07:00](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🧩 States represent configurations, actions are choices, and transition models define the outcome of actions. Goal tests determine if a state is the goal, while path costs measure the cost of actions. - [11:31](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🔄 A search problem involves exploring states using a frontier, a data structure containing states to be explored next. A loop-based search algorithm iteratively explores the frontier, considering possible solutions. - [19:35](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🕵♂ The search algorithm involves removing nodes from the frontier, analyzing their state, parent, action, and path cost to navigate the search space and find solutions. - [33:09](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🔄 Depth First Search (DFS): Explores one path until a dead end is reached, then backtracks and tries another path. Can lead to non-optimal solutions. - [36:24](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🌐 Breadth First Search (BFS): Explores all possible paths at a given depth level before going deeper. Guarantees optimal solutions but may require more memory. - [38:51](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 💻 Code Implementation: The video demonstrates code implementation of DFS and BFS for solving mazes, highlighting their exploration strategies and memory usage.
Brian is incredibly organized and polished. If I had professors this good back when I was in school for CS it would have been a vastly more productive experience.
@@unebonnevie I do... but forty years have come and gone since I started my journey with computer science, and the awareness of the computational nature of reality has changed my outlook considerably. It is both more and less interesting to me now.
When i did my Masters in mathematics, only one of the professors at my school had solid understanding of these topics. I learned as much as i could, but he was overwhelmed with students. I am grateful for these videos!
@@ShadowMind312 One my senior says books are non sense. He says to me go through TH-cam or other online source. But you are learning from books. Is there any reason to learn from books.
Add the times matched to each section: 1:07:00, Jul 30: Introduction to AI Concepts: The document starts with an introduction to basic AI concepts, including agents, states, actions, and transition models1. 1:52:00, Aug 2: Goal Tests and Path Costs: It explains how to define goals and evaluate the cost of reaching goal states in AI problems2. 2:46:00, Aug 3: Search Algorithms: Various search algorithms like DFS and BFS are explored, along with the concept of a frontier for solution exploration3. 3:39:00, Aug 6: Heuristics and A* Search: Heuristics and the A* search algorithm are introduced to improve search efficiency and find optimal solutions4. 5:34:00, Aug 8: Game Playing and Adversarial Search: The document discusses game playing in AI, focusing on adversarial search techniques like Minimax and alpha-beta pruning5. 7:18:00, Aug 9: Knowledge Representation and Logic: It covers knowledge representation and propositional logic, including truth tables, logical connectives, and solving logic puzzles6. 9:35:00, Aug 14: Probability and Bayesian Networks: Basic probability concepts and Bayesian networks are introduced, along with methods for constructing and inferring in these networks7. 10:45:00, Aug 15: Markov Chains and HMMs: The document explains Markov chains and Hidden Markov Models (HMMs), their applications, and common tasks performed with HMMs8. 11:35:00, Aug 17: Machine Learning Techniques: It delves into machine learning, covering supervised learning, classification, neural networks, and techniques like gradient descent9. I hope this helps! Let me know if you need any more details.
hey bud can u help me if this course is for me or not i know python and am quite familiar with OOP concepts idk anything about math used in ML and have built a few mini projects using gemini and langchain I have 2 questions : is this beginner friendly enough for me and what is the end goal of this course
I have been attending and listening to lectures for over 50 years now. You can always speed up the sound if u like. Speed is not a measure of the content of the subject and knowledge
Thank you so much! The only platform who made me take interest in programming after spending 3 years in my 4 year Bachelors degree of CS. CAN'T THANK YOU ENOUGH!🙏🏼
Please guide me brother.. I'm currently in my semester 3 (2nd year) I'm done with python programming all the basics and other concepts. Getting started with javascript
@@divyanshrajput8668 complete paid certification(which has exams) bro it would really help u in resume as well as showing that you have a knowledge in that field during job search to recruiters
how can someone with zero coding experience try to learn this?, am really impressed. i mean the concepts are easy to understand but implementing it requires programming knowledges, and not basic programing knowledges in some cases you will need to know not only the language but also third party packages, so im very impressed
@@jaylooppworld381 yeah but I guess actually it is possible. I don't have any knowledge of coding. But after getting a general view I could understand how to learn the language and what all concepts I have to stress on. 😊
Had Brian as a TA a few years ago, he’s an amazing guy. Malan is also the best professor I’ve ever had. Amazing to get this stuff for free nowadays! No excuses for anyone
This is the video that actually gets to the nitty gritty details as to how an AI actually works rather than just explaining its concepts and its history or whatever, really a gem in a coal mine if you ask me .
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
I find it fascinating how people are shocked when I tell them that I majored in philosophy and work in AI. Usually, they respond with, "Wow, completely unrelated fields." Little do they know that propositional logic is at the core of both philosophy and AI.
This was clear, concise and conclusive course, with a professor that not only knows the topic very well, but does have a way of helping us build the knowledge as a master! Thanks, Brian Yu!!!!
Wow, I have watched several courses on data science and artificial intelligence, but none of the hosts speak as well as Professor Yu. Clean, clear, no ums or ahs, making it easy to listen and follow. The flow makes more sense than the others. Really well done.
This is teaching at its finest. Thank you, Harvard, for your generosity. I am 49 and glad to be living in a time where such precious knowledge is provided for free. It's amazing how one can gain knowledge in abundance, free of charge.
@@dorlock42 Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
For Future Reference: Depth First Search 28:47 Breadth First Search 31:37 Uninformed Search 55:42 Informed Search 56:22 Greedy best first search 57:02 A * Search 01:07:45 Code for the maze 44:38
@@ameen6768 Lol no one learns watching these videos or atleast not to the enough level where they can apply it in actual life. These are all exaggeration
After watching a couple of thousand presentations on youtube, this is hands down number one.. Number one in clarity, fluidity, timing, content and expression. Thank you Mr. Wu, hat is off.
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their timee
Terrific course. Watched this as part of my sabbatical. I'm an experimentalist (genomics/genetics) and this helps bridge the gap with computational approaches to make sense of large data sets and make functional inferences.
harvard really is awesome. its great that they just post this online for anyone to learn from, despite having high standards, they give so much information out for free. i started coding with another cs50 class about learning python, and that was great too.
- This course from Harvard University explores the concepts and algorithms at the foundation of modern artificial intelligence. - The course covers topics such as Graph Search algorithms, classification, optimization, reinforcement learning, and machine learning. - The course is taught by Brian Yu as part of the CS50 program. - The course starts with an introduction to how AI can search for solutions to problems, such as playing a game or finding directions. - The course explores how AI can represent and use information, including uncertain information. - The course covers optimization problems and how AI can learn from data and experiences. - The course includes an exploration of neural networks, a popular tool in modern machine learning. - The course also covers natural language processing, where AI learns to understand and interpret human language.
There are many unique aspects of those videos, but what is really nice is the depth of explaining such concepts. Even in a university, they usually cannot go that deep mostly due to time constraints.
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
Alhamdulillah, just finished it for the first round. I unintentionally stopped jotting at 'Inference Algorithms'. Going back for round 2 to understand more and continue jotting. You need notes to understand better bcoz you've to be going back and forth. Love it. Good job Sir Brian.
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
GOAT. Im from Argentina, so im not speak emglish naturally, and in every spanish video that i watched, about neural networks aididnt understand it as well as i did it here. thank u
Prof. Brian Yu, Just *amazing*!!! I am at the tail end of my career and viewing this just out of curiosity and cannot stop watching this video! You just have a talent for communicating these concepts! Just amazing, and a very big Thank You!!
By "YouSum Live" part 3 09:00:00 Process of assigning points to clusters in k-means 09:01:10 Iterative nature of k-means clustering 09:01:24 Re-centering clusters in k-means 09:03:09 Equilibrium and completion of k-means algorithm 09:03:35 Application and significance of unsupervised learning 09:04:41 Transition to neural networks in machine learning 09:05:24 Inspiration from human brain structure for neural networks 09:06:34 Explanation of artificial neural networks and activation functions 09:12:28 Illustration of neural network structure and function 09:14:34 Training a neural network for the OR function 09:15:17 Neural network basics and applications 09:15:23 Understanding activation functions and thresholds 09:16:35 Modeling simple functions like OR and AND 09:20:43 Introduction to gradient descent in training 09:24:51 Trade-offs between gradient descent methods 09:25:18 Mini-batch gradient descent for efficiency 09:29:33 Supervised machine learning and neural networks 09:30:02 Application of neural networks in reinforcement learning 09:31:36 Training neural networks with multiple outputs 09:32:50 Introduction to neural network limitations 09:33:12 Perceptron's linear separability constraint 09:34:43 Multilayer neural network proposal 09:35:46 Hidden layers enhance function complexity 09:37:18 Backpropagation for training hidden layers 09:41:17 Overfitting risk in complex neural networks 09:42:01 Dropout technique to prevent overfitting 09:43:48 TensorFlow for neural network implementation 09:46:39 Hidden layers improve data separation 09:47:55 Impact of hidden layers on decision boundaries 09:49:06 Addressing non-linear data with hidden layers 09:49:48 Understanding neural networks and backpropagation 09:50:02 Importance of hidden layers in learning data structure 09:50:13 Utilizing backpropagation to adjust weights for accurate classification 09:50:26 Training neural networks to classify data categories effectively 09:51:40 Implementing neural networks in Python using TensorFlow 09:53:01 Balancing complexity and overfitting in neural network design 09:53:15 Testing and optimizing hyperparameters for neural network performance 09:57:43 Introduction to computer vision and its applications 10:03:50 Image convolution for feature extraction in computer vision 10:07:15 Applying kernels in image processing for feature extraction 10:07:43 Detecting edges and boundaries using specific filter kernels 10:08:06 Image filtering for edge extraction and feature detection 10:09:33 Utilizing filters to extract valuable information from images 10:11:01 Pooling technique for downsizing image inputs by sampling regions 10:11:23 Max pooling to reduce image dimensions by selecting maximum values 10:13:03 Constructing convolutional neural networks for image analysis 10:14:32 Training CNNs to learn filters for feature extraction 10:17:17 Hierarchical feature learning in CNNs for image recognition 10:24:47 Saving and reusing model in TensorFlow 10:25:33 Training neural networks on handwritten digits 10:25:44 Importance of computational power in training 10:26:20 Iterative improvement of accuracy through training 10:26:49 Learning features and weights in neural networks 10:27:09 Monitoring training progress and accuracy 10:27:56 Testing accuracy on a separate dataset 10:28:13 Applying neural networks for handwriting recognition 10:30:00 Power of neural networks in image analysis 10:32:54 Recurrent neural networks for sequence data processing 10:40:15 Recurrent neural networks for video analysis 10:46:00 Understanding natural language processing challenges 10:48:18 Syntax: Structure of language 10:49:52 Semantics: Meaning of language 10:51:56 Formal grammar: Rules for sentence generation 10:55:23 Context-free grammar: Parsing sentence structure 11:00:46 Statistical approach: Analyzing n-grams for language structure 11:01:14 Analyzing ngrams in text data 11:02:02 Identifying common bigrams and trigrams 11:02:32 Tokenization process for text analysis 11:03:00 Building a Markov chain for language prediction 11:04:23 Generating sentences based on statistical patterns 11:05:09 Introduction to text classification 11:05:51 Applying sentiment analysis to text data 11:07:40 Naive Bayes classifier for text sentiment analysis 11:13:44 Challenges and solutions in text classification 11:17:13 Word representation in neural networks 11:19:24 Representation of word meanings through vectors 11:20:05 Transition from one-hot to distributed representations 11:20:45 Deriving word meanings from surrounding context 11:21:40 Utilizing Word2Vec model for word vector generation 11:23:44 Analyzing word vector distances for similarity 11:24:24 Identifying closest words based on vector representations 11:25:12 Capturing relationships between words using vectors 11:26:37 Application of word vectors in neural networks 11:34:42 Implementing attention mechanism for sequence translation 11:38:30 Attention mechanism in machine learning 11:39:25 Challenges of parallelizing recurrent neural networks 11:40:15 Evolution from recurrent neural networks to transformers 11:40:25 Transformer architecture overview 11:42:51 Importance of positional encoding in transformers 11:43:49 Self-attention for better word representation 11:44:36 Multi-headed attention for comprehensive context 11:44:49 Deep learning repetition for deeper patterns 11:46:48 Decoder's attention to encoded input representations 11:48:39 Transformer's focus on attention for effective results 11:49:04 Advancements in natural language processing
This is the best lecture trailer in computer science I have ever seen. Observe that he started naturally from states and agents(state machines) to gradually reach AI and ML.
Thanks for this wonderful content. Professor Brian has done an incredibly well done job. I learned much and Thank all the persons involved in producing these lectures. Much grateful to Harvard University for sharing this. Professor Brian, you are a GREAT Communicator!
Thank you for your kind words! We're delighted to hear that you enjoyed the content and found Professor Brian's lectures informative. Your appreciation means a lot to us and to everyone involved in producing these lectures.
Learned a lot from this video. Two thumbs up. For the specific example he gave, the number tile, I think "reverse engineering" approach, couples with the AI process he described, will solve the problem more efficiently. That means I start with the end sequence = numbers in ascending order left to right, top to bottom. Then I map out all possible paths to "chaos" state = all tile arrangements that are not the end sequence. I can determine all possible chaos states = 16! = 2.092279e+13 assuming the hole is also a tile. The possible paths should be much less than 16! because each move along the way to a most "severe" chaos state is a chaos state itself. The map will look like a family tree, starting with the end sequence, and the last progeny of each branch is the most "severe" chaos. When user enters a chaos state, the algo finds where it is on the family tree, follow the reverse path/moves back up to the end sequence. The reverse-engineering approach will only work well when the goal/end is well defined.
I am surprised by the complexity of the study it requires, just to develop the AI that plays these games. Kudos to the professor who is just excellent at this concept in the simply digestible form.
This course is amazing. It was my first encounter with an AI lecture, and I enjoyed it. I really enjoyed the Professor's teaching method, when he introduces new concepts, he starts with the easy to understand components and increases the complexities of the contents gradually. Each segment is accompanied with numerous real-life examples that make understanding the concept easier. Keep up the good work.
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
Interesting. Prof. Yu. I use simplification on the ground floor at my work everyday and it works.Close to 70 and I am still at work. But your knowledge is a probe of new ideas. I have a lot that flows everyday. START !. (Zero emission) Please, can you put this function into simplification for me to grasp your whole concept. I want to come to Havard for CS50.
I am planning to complete this course within a week, I'll update each time I complete for each day with the time stamp: Day 1: I just finished Search functions fully 1:51:56 Day 2: knowledge completed (I forgot to edit the comment because I actually went to Harvard website and registered for the course and studied there) Day 3: half way through probability Day 4: I'm done with probability and now i'll continue back in few days Day n: Idk what day is this but It's been more than a month I think so, but finally completed the course and now I need to start making projects.
@@Pclub4ever why are you being like this? I legit went to their website and paid for a certification and studied through their website, I forgot to update it on TH-cam :/
This brought good memories of a course I took in 1989 while doing my computer science degree. We did the same algorithm and many similar ones. For the first one, the maze, we used a recursive algorithm or loops, traversing graphs and trees. I did not know we were doing AI at that time :) In fact, this is not AI, but without the AI attached to the name, people may not watch the video.😁 This video is fantastic as it illustrates the power of optimizing algorithms to improve performance. The time complexity of such an algorithm as the maze is exponential.
I'm only an hour into the lesson, but I'm clueless as to how this could be related to AI in anyway. This just seems like standard coding problems. I hope it clicks at some point.
A couple of notes I have noticed that are not explicitly taught: Note 1: P v Q ~P v R then P v ~P this is important because if we know Q v R and we know it is Q, then we know it is also ~P. Note 2: (A -> B) = ~A v B If above is true, then this must also be true: (A -> B) = ~B v A
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their timee
Export the Q*, Chat GPT, Revit, Plant 3D, Civil 3D, Inventor, ENGI file of the Building or Refinery to Excel, prepare Budget 1 and export it to COBRA. Prepare Budget 2 and export it to Microsoft Project. Solve the problems of Overallocated Resources, Planning Problems, prepare the Budget 3 with which the construction of the Building or the Refinery is going to be quoted.
Yes it would be helpful if you're familiar with Python and some basic data structures of the language like lists, tuples and dictionaries. It might also be helpful (although not necessary) to know a bit of discrete mathematics and some algorithmic concepts like recursion etc.
@@raidensh0gun you definitely need to know recursion and OOP. Also data structures like queues and stacks. I'm only on the first week and all this has popped up with the first two projects, I imagine it would only be getting more complicated as the course goes on. According to harvard, you should know the equivalent of the cs50 intro to programming course, which imo is equivalent to any intro to cs college level course
Section “knowledge”, “uncertainty”, and “optimization” are so interesting. Even though you don’t directly use them much in an industrial setting, they’re still beneficial to problem solving. Nevertheless, they’re always skipped in most MOOCs.
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
fantastic learning resource. please can you share the github link for the sourcecode from the demoes in the lectures. I couldnt locate it from the cs50 repoes on github
Listing some subchapters for future reference
01:07:41 - A* Algorithm - "heuristc search" for a singleagent
01:14:27 Adversarial Search (Tic-Tac-Toe)
01:39:07 Optimization - Alpha-Beta Prunning
01:46:53 Chess -Depth Limited Minimax
02:14:05 Inference Algoritms -Model Checking
02:32:42 Knowledge Ingeneering - Clue
02:43:04 Logic Puzzles - Harry Potter
02:56:16 Inference Rules
03:22:26 Inference by Resolution
03:39:39 Uncertainty - Probability theory
03:49:16 Conditional Probability
04:05:58 Bayes Rule
04:13:40 Joint Probability
thanks
Written in description
@@muhammadfaheem5213 Yes and No. I added some chapters not included there :)
Professor Yu is in his mid-twenties and teaches one of the most subscribed courses at Harvard. Amazing!
The AI branch. David Malan still teaches cs50x. Impressive nonetheless
@@coldfire6869 Don’t be petty
Everyone wants to get in on the groundfloor
Brian Yu, Senior Preceptor in Computer Science, Harvard University
thats what happens when you have passion
I am in my 70's. Last time I get excited about programming was Lotus 123 macro in the 1980s. Here we go again.
Never to old to learn as long as the subject matter wets your appetite, btw I'm in my 60's & going head long into this
glad to read this! enjoy the journey
- [00:00](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🤖 The course explores foundational concepts and algorithms of modern artificial intelligence, covering topics like graph search algorithms, optimization, reinforcement learning, and more.
- [03:16](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🛣️ AI aims to solve problems by searching for solutions using various actions and transitions between states in a state space.
- [07:00](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🧩 States represent configurations, actions are choices, and transition models define the outcome of actions. Goal tests determine if a state is the goal, while path costs measure the cost of actions.
- [11:31](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🔄 A search problem involves exploring states using a frontier, a data structure containing states to be explored next. A loop-based search algorithm iteratively explores the frontier, considering possible solutions.
- [19:35](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🕵♂ The search algorithm involves removing nodes from the frontier, analyzing their state, parent, action, and path cost to navigate the search space and find solutions.
- [33:09](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🔄 Depth First Search (DFS): Explores one path until a dead end is reached, then backtracks and tries another path. Can lead to non-optimal solutions.
- [36:24](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 🌐 Breadth First Search (BFS): Explores all possible paths at a given depth level before going deeper. Guarantees optimal solutions but may require more memory.
- [38:51](th-cam.com/video/5NgNicANyqM/w-d-xo.html) 💻 Code Implementation: The video demonstrates code implementation of DFS and BFS for solving mazes, highlighting their exploration strategies and memory usage.
Great
Brian is incredibly organized and polished. If I had professors this good back when I was in school for CS it would have been a vastly more productive experience.
@@unebonnevie I do... but forty years have come and gone since I started my journey with computer science, and the awareness of the computational nature of reality has changed my outlook considerably. It is both more and less interesting to me now.
When i did my Masters in mathematics, only one of the professors at my school had solid understanding of these topics. I learned as much as i could, but he was overwhelmed with students. I am grateful for these videos!
Would you like to taught this topic?
@nnbbbbb-jv8yz yes, I also purchased textbooks so I am working through the concepts.
I think you are experienced person. My question is ( is it good to learn code from online or through real physical mentor.) I have this doubt.
@@ShadowMind312 One my senior says books are non sense. He says to me go through TH-cam or other online source. But you are learning from books. Is there any reason to learn from books.
What job are you going for in computer science
Grateful to the Harvard University for providing this course. Thank you brian yu and all of them who are behind to provide this course.
Stop reading comments , and follow the lesson
😂😂
Guilty 😅
Stop focussing on comment box👊
😮😂
Stop writing the comment 😅
1:07:00, Jul 30
1:52:00, Aug 2,
2:46:00, Aug 3,
3:39:00, Aug 6,
5:34:00 Aug 8,
7:18:00 Aug 9,
9:35:00 Aug 14
10:45:00 Aug 15
11:35:00 Aug 17
Add the times matched to each section:
1:07:00, Jul 30: Introduction to AI Concepts: The document starts with an introduction to basic AI concepts, including agents, states, actions, and transition models1.
1:52:00, Aug 2: Goal Tests and Path Costs: It explains how to define goals and evaluate the cost of reaching goal states in AI problems2.
2:46:00, Aug 3: Search Algorithms: Various search algorithms like DFS and BFS are explored, along with the concept of a frontier for solution exploration3.
3:39:00, Aug 6: Heuristics and A* Search: Heuristics and the A* search algorithm are introduced to improve search efficiency and find optimal solutions4.
5:34:00, Aug 8: Game Playing and Adversarial Search: The document discusses game playing in AI, focusing on adversarial search techniques like Minimax and alpha-beta pruning5.
7:18:00, Aug 9: Knowledge Representation and Logic: It covers knowledge representation and propositional logic, including truth tables, logical connectives, and solving logic puzzles6.
9:35:00, Aug 14: Probability and Bayesian Networks: Basic probability concepts and Bayesian networks are introduced, along with methods for constructing and inferring in these networks7.
10:45:00, Aug 15: Markov Chains and HMMs: The document explains Markov chains and Hidden Markov Models (HMMs), their applications, and common tasks performed with HMMs8.
11:35:00, Aug 17: Machine Learning Techniques: It delves into machine learning, covering supervised learning, classification, neural networks, and techniques like gradient descent9.
I hope this helps! Let me know if you need any more details.
It's the best thing I have ever seen on TH-cam. Great job, thank you for every minute of this course.
hey bud can u help me if this course is for me or not
i know python and am quite familiar with OOP concepts
idk anything about math used in ML and have built a few mini projects using gemini and langchain
I have 2 questions : is this beginner friendly enough for me and what is the end goal of this course
What I love about these Harvard CS50 videos is the speed they talk and explain things. It's captivating.
I was watching on 1.5x speed the whole video
I have been attending and listening to lectures for over 50 years now. You can always speed up the sound if u like. Speed is not a measure of the content of the subject and knowledge
@@tobiassjoholm9325I'm at 2x 😅
@@tobiassjoholm9325 3x
Thank you so much! The only platform who made me take interest in programming after spending 3 years in my 4 year Bachelors degree of CS. CAN'T THANK YOU ENOUGH!🙏🏼
Something I can relate to!!
Please guide me brother.. I'm currently in my semester 3 (2nd year)
I'm done with python programming all the basics and other concepts.
Getting started with javascript
Really? Those years must have been hard for you.
@@divyanshrajput8668 complete paid certification(which has exams) bro it would really help u in resume as well as showing that you have a knowledge in that field during job search to recruiters
@@rumanaislam4758 Might be just exaggerating
dude i cannot thank you enough, what a time to be alive :)
This is amazing! I'm a beginner with zero coding experience and I'm understanding this with clarity. Professor Yu is stellar!
how can someone with zero coding experience try to learn this?, am really impressed. i mean the concepts are easy to understand but implementing it requires programming knowledges, and not basic programing knowledges in some cases you will need to know not only the language but also third party packages, so im very impressed
@@jaylooppworld381Do you know somewhere to learn full course of AI / ML on internet..
Same
@@jaylooppworld381 this is what I was searching for. Where would you suggest I go as a beginner before I go through this course?
@@jaylooppworld381 yeah but I guess actually it is possible. I don't have any knowledge of coding. But after getting a general view I could understand how to learn the language and what all concepts I have to stress on. 😊
Had Brian as a TA a few years ago, he’s an amazing guy. Malan is also the best professor I’ve ever had. Amazing to get this stuff for free nowadays! No excuses for anyone
Except for those without access to a computer. There are many such people.
Another excuse is, finding good videos like this in a sea of billions, is like finding a needle in a million haystacks.
@@JH-no8sy that is true. Thank you for humbling me.
This is the video that actually gets to the nitty gritty details as to how an AI actually works rather than just explaining its concepts and its history or whatever, really a gem in a coal mine if you ask me .
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
@@fevad1246 bro is a menace in comments. and no one is replying
Almost two hours in and it's so good; I understand the concepts really well. Thank you, Mr. Brian. I am really enjoying the course!
What I will be able to do after completing this?
@@Urug01315 IDK but I love the logic part of the course
Did u take notes or just watch
@@superrbx I mean there are already notes on the cs50 website.
@@superrbx Take notes 😂😂😂 are you trying to take a test or trying to learn something?
I find it fascinating how people are shocked when I tell them that I majored in philosophy and work in AI. Usually, they respond with, "Wow, completely unrelated fields." Little do they know that propositional logic is at the core of both philosophy and AI.
Wow
yeah i get it trying make your useless time spent on philosophy look somehow useful
Same!
This was clear, concise and conclusive course, with a professor that not only knows the topic very well, but does have a way of helping us build the knowledge as a master! Thanks, Brian Yu!!!!
Wow, I have watched several courses on data science and artificial intelligence, but none of the hosts speak as well as Professor Yu. Clean, clear, no ums or ahs, making it easy to listen and follow. The flow makes more sense than the others. Really well done.
This is teaching at its finest. Thank you, Harvard, for your generosity. I am 49 and glad to be living in a time where such precious knowledge is provided for free. It's amazing how one can gain knowledge in abundance, free of charge.
Information/knowledge is one of the most valuable things and if we don't share it. We will get nowhere as a society.
Cierto, que suerte lo que hablas inglés como lenguaje original, la mejores información están en inglés. Pero aquí estoy auque con subtitulos
@@dorlock42 Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
His teaching style is incredibly precise and evident. It's a great pleasure to watch these lectures. Well done!
For Future Reference:
Depth First Search 28:47
Breadth First Search 31:37
Uninformed Search 55:42
Informed Search 56:22
Greedy best first search 57:02
A * Search 01:07:45
Code for the maze 44:38
.
I can’t afford to miss this. This weekend I’ll go in depth this course. A value bomb! Thank you profesors!
Did you finish it? If so was it worth it and are there parts you suggest to skip? Thanks!
@@ameen6768 Lol no one learns watching these videos or atleast not to the enough level where they can apply it in actual life. These are all exaggeration
@@fevad1246 just say you aren't made for it and move on
@@nephilim18 And who are you?An Ai engineer?
After watching a couple of thousand presentations on youtube, this is hands down number one.. Number one in clarity, fluidity, timing, content and expression. Thank you Mr. Wu, hat is off.
Thank you for mentioning this 😊
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their timee
Terrific course. Watched this as part of my sabbatical. I'm an experimentalist (genomics/genetics) and this helps bridge the gap with computational approaches to make sense of large data sets and make functional inferences.
harvard university is really changing the world by giving acceses to the inovative minds all over the world
Brian yu's speech is really perspicuous and crisp pronunciation
This is amazing! Both the course and the teacher. Thank you very much for sharing this.
harvard really is awesome. its great that they just post this online for anyone to learn from, despite having high standards, they give so much information out for free. i started coding with another cs50 class about learning python, and that was great too.
- This course from Harvard University explores the concepts and algorithms at the foundation of modern artificial intelligence.
- The course covers topics such as Graph Search algorithms, classification, optimization, reinforcement learning, and machine learning.
- The course is taught by Brian Yu as part of the CS50 program.
- The course starts with an introduction to how AI can search for solutions to problems, such as playing a game or finding directions.
- The course explores how AI can represent and use information, including uncertain information.
- The course covers optimization problems and how AI can learn from data and experiences.
- The course includes an exploration of neural networks, a popular tool in modern machine learning.
- The course also covers natural language processing, where AI learns to understand and interpret human language.
are you a bot?
Thank you
Are there any prerequisites of statistics, or ML?
Thanks for putting together this course in one video.. Thank you so much for all of the free courses you upload.
Best course and so perfect explained, thanks Brian Yu
There are many unique aspects of those videos, but what is really nice is the depth of explaining such concepts. Even in a university, they usually cannot go that deep mostly due to time constraints.
notes: frontier, action function, transition function, depth-first-search, breadht-first-search, greedy-best-first-search (with heuristic closest to goal function), manhattan-distance, A*search=>optimal solution, admissible heuristic never overestimates the true cost, classical search vs adversarial search, minimax
I am so glad I found this, am doing an AI Engineering Course with another institute, you've just made it so much easier to understand.
Thank you so much. I don't recall ever enjoying a lecture series as much as this. Great work! I hope to see a follow up in a few years.
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
Alhamdulillah, just finished it for the first round. I unintentionally stopped jotting at 'Inference Algorithms'. Going back for round 2 to understand more and continue jotting. You need notes to understand better bcoz you've to be going back and forth. Love it. Good job Sir Brian.
Have you written any notes?
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
@@fevad1246 did you tried it , and make sure to tell that you did learn or not
@sigmaboy9431 do you even realize the point of the reply ??💀
@@fevad1246 I see so you just went to some random videos and posted this , considering you didn't take it
GOAT. Im from Argentina, so im not speak emglish naturally, and in every spanish video that i watched, about neural networks aididnt understand it as well as i did it here. thank u
Prof. Brian Yu, Just *amazing*!!! I am at the tail end of my career and viewing this just out of curiosity and cannot stop watching this video! You just have a talent for communicating these concepts! Just amazing, and a very big Thank You!!
10 millions subscribers congratulations...❤
I love the way he explains everything. Thanks for the course
Currently pursuing an MSc in AI, and this video was a great introduction to get a bigger picture. Thank you.
What's next for you ?
Thank you so much for all of the free courses you upload
Thank you! At age 62, loving every video you release!
why are you learning tech at 62 bro
why are you learning tech at all@@fredrickmweu
@@fredrickmweu I started at age 18 and it's been my life's work so no need to stop now. Maybe at 100...maybe.
@@georgejetson9801I'm 104 years old and still learning everyday. 🙏🏼
@@fredrickmweuhe knows more than you lmao 🤡
By "YouSum Live" part 3
09:00:00 Process of assigning points to clusters in k-means
09:01:10 Iterative nature of k-means clustering
09:01:24 Re-centering clusters in k-means
09:03:09 Equilibrium and completion of k-means algorithm
09:03:35 Application and significance of unsupervised learning
09:04:41 Transition to neural networks in machine learning
09:05:24 Inspiration from human brain structure for neural networks
09:06:34 Explanation of artificial neural networks and activation functions
09:12:28 Illustration of neural network structure and function
09:14:34 Training a neural network for the OR function
09:15:17 Neural network basics and applications
09:15:23 Understanding activation functions and thresholds
09:16:35 Modeling simple functions like OR and AND
09:20:43 Introduction to gradient descent in training
09:24:51 Trade-offs between gradient descent methods
09:25:18 Mini-batch gradient descent for efficiency
09:29:33 Supervised machine learning and neural networks
09:30:02 Application of neural networks in reinforcement learning
09:31:36 Training neural networks with multiple outputs
09:32:50 Introduction to neural network limitations
09:33:12 Perceptron's linear separability constraint
09:34:43 Multilayer neural network proposal
09:35:46 Hidden layers enhance function complexity
09:37:18 Backpropagation for training hidden layers
09:41:17 Overfitting risk in complex neural networks
09:42:01 Dropout technique to prevent overfitting
09:43:48 TensorFlow for neural network implementation
09:46:39 Hidden layers improve data separation
09:47:55 Impact of hidden layers on decision boundaries
09:49:06 Addressing non-linear data with hidden layers
09:49:48 Understanding neural networks and backpropagation
09:50:02 Importance of hidden layers in learning data structure
09:50:13 Utilizing backpropagation to adjust weights for accurate classification
09:50:26 Training neural networks to classify data categories effectively
09:51:40 Implementing neural networks in Python using TensorFlow
09:53:01 Balancing complexity and overfitting in neural network design
09:53:15 Testing and optimizing hyperparameters for neural network performance
09:57:43 Introduction to computer vision and its applications
10:03:50 Image convolution for feature extraction in computer vision
10:07:15 Applying kernels in image processing for feature extraction
10:07:43 Detecting edges and boundaries using specific filter kernels
10:08:06 Image filtering for edge extraction and feature detection
10:09:33 Utilizing filters to extract valuable information from images
10:11:01 Pooling technique for downsizing image inputs by sampling regions
10:11:23 Max pooling to reduce image dimensions by selecting maximum values
10:13:03 Constructing convolutional neural networks for image analysis
10:14:32 Training CNNs to learn filters for feature extraction
10:17:17 Hierarchical feature learning in CNNs for image recognition
10:24:47 Saving and reusing model in TensorFlow
10:25:33 Training neural networks on handwritten digits
10:25:44 Importance of computational power in training
10:26:20 Iterative improvement of accuracy through training
10:26:49 Learning features and weights in neural networks
10:27:09 Monitoring training progress and accuracy
10:27:56 Testing accuracy on a separate dataset
10:28:13 Applying neural networks for handwriting recognition
10:30:00 Power of neural networks in image analysis
10:32:54 Recurrent neural networks for sequence data processing
10:40:15 Recurrent neural networks for video analysis
10:46:00 Understanding natural language processing challenges
10:48:18 Syntax: Structure of language
10:49:52 Semantics: Meaning of language
10:51:56 Formal grammar: Rules for sentence generation
10:55:23 Context-free grammar: Parsing sentence structure
11:00:46 Statistical approach: Analyzing n-grams for language structure
11:01:14 Analyzing ngrams in text data
11:02:02 Identifying common bigrams and trigrams
11:02:32 Tokenization process for text analysis
11:03:00 Building a Markov chain for language prediction
11:04:23 Generating sentences based on statistical patterns
11:05:09 Introduction to text classification
11:05:51 Applying sentiment analysis to text data
11:07:40 Naive Bayes classifier for text sentiment analysis
11:13:44 Challenges and solutions in text classification
11:17:13 Word representation in neural networks
11:19:24 Representation of word meanings through vectors
11:20:05 Transition from one-hot to distributed representations
11:20:45 Deriving word meanings from surrounding context
11:21:40 Utilizing Word2Vec model for word vector generation
11:23:44 Analyzing word vector distances for similarity
11:24:24 Identifying closest words based on vector representations
11:25:12 Capturing relationships between words using vectors
11:26:37 Application of word vectors in neural networks
11:34:42 Implementing attention mechanism for sequence translation
11:38:30 Attention mechanism in machine learning
11:39:25 Challenges of parallelizing recurrent neural networks
11:40:15 Evolution from recurrent neural networks to transformers
11:40:25 Transformer architecture overview
11:42:51 Importance of positional encoding in transformers
11:43:49 Self-attention for better word representation
11:44:36 Multi-headed attention for comprehensive context
11:44:49 Deep learning repetition for deeper patterns
11:46:48 Decoder's attention to encoded input representations
11:48:39 Transformer's focus on attention for effective results
11:49:04 Advancements in natural language processing
Thank You!!
Is this free
This is a great course for me to though ML after my 12week online course from MIT!
This is the best lecture trailer in computer science I have ever seen.
Observe that he started naturally from states and agents(state machines) to gradually reach AI and ML.
I agree, but did you finish it? If so was it worth it and are there parts you suggest to skip?
Just finished the course, now let's see how it compares to my graduate course that I start next week. Thank you so much for this.
Thank You very much Professor.
Thanks for this wonderful content. Professor Brian has done an incredibly well done job. I learned much and Thank all the persons involved in producing these lectures. Much grateful to Harvard University for sharing this. Professor Brian, you are a GREAT Communicator!
Thank you for your kind words! We're delighted to hear that you enjoyed the content and found Professor Brian's lectures informative. Your appreciation means a lot to us and to everyone involved in producing these lectures.
thanks for disponibility this course, i really apreciate the initiative. kisses from brazil
Learned a lot from this video. Two thumbs up. For the specific example he gave, the number tile, I think "reverse engineering" approach, couples with the AI process he described, will solve the problem more efficiently. That means I start with the end sequence = numbers in ascending order left to right, top to bottom. Then I map out all possible paths to "chaos" state = all tile arrangements that are not the end sequence. I can determine all possible chaos states = 16! = 2.092279e+13 assuming the hole is also a tile. The possible paths should be much less than 16! because each move along the way to a most "severe" chaos state is a chaos state itself. The map will look like a family tree, starting with the end sequence, and the last progeny of each branch is the most "severe" chaos. When user enters a chaos state, the algo finds where it is on the family tree, follow the reverse path/moves back up to the end sequence. The reverse-engineering approach will only work well when the goal/end is well defined.
I agree with you, btw did you finish it? If so was it worth it and are there parts you suggest to skip?
I have an exam tomorrow in chemistry, and I can't seem to stop watching this lecture today! Man, I'm gonna be screwed!
Thanks for putting together this course in one video.
9:14:08 Isn't it better to work with bias of -0.5 in case of the OR unction
9:16:51 and bias of -1.5 in case of the AND unction
I am surprised by the complexity of the study it requires, just to develop the AI that plays these games. Kudos to the professor who is just excellent at this concept in the simply digestible form.
Thank you Brian Yu for such a great lection!
I am very grateful for the excellent free courses at Harvard. ❤
This course is amazing. It was my first encounter with an AI lecture, and I enjoyed it. I really enjoyed the Professor's teaching method, when he introduces new concepts, he starts with the easy to understand components and increases the complexities of the contents gradually. Each segment is accompanied with numerous real-life examples that make understanding the concept easier. Keep up the good work.
I'm 104 years old and I'm loving this. Thank you 🙏🏼
If you 104 years old then you should love oldness intelligence not Artificial Intelligence.
im 420 years old
I am 2023 years old😊
@@mr_saam Ok, Jesus
@@Abdullah-fn1kz I'm not old, just experienced 😅
Love the BFS concept
Very useful and informative. The course provides in-depth knowledge. I learned a lot. I am grateful for providing this great course.
I wish I could give so much more than just the one thumbs up for this course, it was incredible. Thank you so much
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
Prof. Brian Yu, he's got a knack for making even the most complex topics seem like a walk in the park.
Interesting. Prof. Yu. I use simplification on the ground floor at my work everyday and it works.Close to 70 and I am still at work. But your knowledge is a probe of new ideas. I have a lot that flows everyday. START !. (Zero emission) Please, can you put this function into simplification for me to grasp your whole concept. I want to come to Havard for CS50.
Finally finished this course before applying for uni.
huge thanks to all of you who created this wonderful course, I really learnt a lot.
39:00 code to solve maze problem
Getting to know the search problem is actually very helpful to educators.
As a student in Aisa, I sincerely appreciate that one of the world's best universities can provide its wonderful courses for free.
回覆
I am 17mins into this video it’s wonderful. But yet all the best for those without an AI background
Amazing content, thank you CS50 Team. You are on top of the world
I suggest that we should move and check knowledge.
Thank you for sharing this with the world.
Learnt a lot will go for Deep learning from here along with web development with python as it will help to deploy model.... good work!
I am planning to complete this course within a week, I'll update each time I complete for each day with the time stamp:
Day 1: I just finished Search functions fully 1:51:56
Day 2: knowledge completed (I forgot to edit the comment because I actually went to Harvard website and registered for the course and studied there)
Day 3: half way through probability
Day 4: I'm done with probability and now i'll continue back in few days
Day n: Idk what day is this but It's been more than a month I think so, but finally completed the course and now I need to start making projects.
Let's go. Commented so that I can see your progress. 👨🎓
@@preythapp He won't stick to it. People who talk about their plans rarely do it because talking about it already gives them their dopamine hit.
@@Pclub4ever nooo. Please don't say that 😬. I hope he stands by his plan.
cmon bro, only one day missed don't give up, good luck
@@Pclub4ever why are you being like this? I legit went to their website and paid for a certification and studied through their website, I forgot to update it on TH-cam :/
thanks for the lecture i understood data structure more easily here than when i took the data structure class ten years ago
A very informative course, i learnt a lot of new techniques and information thank you.
thank you for your explanation
Brian Yu is best, i wish i had half of his talents.
A very much thank you sir for supporting me in this journey. I will look forward to learn more and more.. Thank you
This brought good memories of a course I took in 1989 while doing my computer science degree. We did the same algorithm and many similar ones. For the first one, the maze, we used a recursive algorithm or loops, traversing graphs and trees. I did not know we were doing AI at that time :) In fact, this is not AI, but without the AI attached to the name, people may not watch the video.😁
This video is fantastic as it illustrates the power of optimizing algorithms to improve performance. The time complexity of such an algorithm as the maze is exponential.
The guy seems cooking a good AI vinaigrette
I'm only an hour into the lesson, but I'm clueless as to how this could be related to AI in anyway. This just seems like standard coding problems. I hope it clicks at some point.
Thank you, i thought i am the crazy old computer guy here.
Wow...this prof really explain his stuff so well!
11:44 It is like useReducer in ReactJS, you pas state and and dispatch action
A couple of notes I have noticed that are not explicitly taught:
Note 1:
P v Q
~P v R
then
P v ~P
this is important because if we know
Q v R
and we know it is Q, then we know it is also ~P.
Note 2:
(A -> B) = ~A v B
If above is true, then this must also be true:
(A -> B) = ~B v A
The way Bryan explained a* search is so much better than the other videos i have watched!
Brian is the best lecturer I've ever seen across all my years studying across 4 universities
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their timee
This is unbelievable. Much appreciated.🛐🛐🛐
Export the Q*, Chat GPT, Revit, Plant 3D, Civil 3D, Inventor, ENGI file of the Building or Refinery to Excel, prepare Budget 1 and export it to COBRA. Prepare Budget 2 and export it to Microsoft Project. Solve the problems of Overallocated Resources, Planning Problems, prepare the Budget 3 with which the construction of the Building or the Refinery is going to be quoted.
Que contéudo maravilhoso, era tudo que eu precisava, grata
Olá Valmira! Vc domina o ingles?conseguiu acompanhar de boa ou precisou de legendas em português?
Any prerequisites for this? Before one gets started...
Can't wait to get hand on this before my next semester ❣️
When's your next semester
Same😊
@@mikeagoya end of August
does it cover up the artificial intelligence course's concepts in the bachelors of that of similar subjects
@@frontire_ace most probably yes.
SO AMAZING I WILL BE WATCHING ALL OF THIS ON REPEAT FOR DAYS AND THEN IM MAKING MY LITTLE MODEL FOR MY IDEAS
This man is the Hero
such an awesome lecture, this makes me enjoy learning even more!
Any prerequisites for this course especially should we know python before taking this course?
Yes it would be helpful if you're familiar with Python and some basic data structures of the language like lists, tuples and dictionaries. It might also be helpful (although not necessary) to know a bit of discrete mathematics and some algorithmic concepts like recursion etc.
@@raidensh0gun you definitely need to know recursion and OOP. Also data structures like queues and stacks. I'm only on the first week and all this has popped up with the first two projects, I imagine it would only be getting more complicated as the course goes on.
According to harvard, you should know the equivalent of the cs50 intro to programming course, which imo is equivalent to any intro to cs college level course
You should first be familiar with the concepts from the first CS50 course: th-cam.com/video/8mAITcNt710/w-d-xo.html
I couldn't even imagine watching this without solid fundamentals.
As a linguist, I find this class so meaningful...
Section “knowledge”, “uncertainty”, and “optimization” are so interesting. Even though you don’t directly use them much in an industrial setting, they’re still beneficial to problem solving. Nevertheless, they’re always skipped in most MOOCs.
This the best delivery I have seen in a longtime
Exaggerations lol. I have yet to meet someone or atleast get a reply in comments from people who are claiming that courses are good,useful etc . They all go silent because they know they learnt nothing and wasted hours of their time
fantastic learning resource. please can you share the github link for the sourcecode from the demoes in the lectures. I couldnt locate it from the cs50 repoes on github