Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

This lecture was a really simple & practical introduction to how machines learn Bayesian Learning, Bayes Theorem, Naive Bayes from Kimia Lab by Professor H.R.Tizhoosh. The lecture also walks through - MAP - Maximum Likelihood - Bayes Optimal Classifier - Naive Bayes model A really beginner-friendly lecture if you... show more

𝗿𝗮𝗺𝗮𝗸𝗿𝘂𝘀𝗵𝗻𝗮— 𝗲/𝗮𝗰𝗰

12,763 subscribers

14,816 просмотров • 3 месяцев назад •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

It takes just 9 minutes of this lecture from the MIT to understand how a 273-year-old formula continues to generate profits Bayes' theorem is one of the best formulas for trading on the basis of market sentiment Save this video and watch it to apply the formula on the Polymarket then, read the article

It takes just 9 minutes of this lecture from the MIT to understand how a 273-year-old formula continues to generate profits Bayes' theorem is one of the best formulas for trading on the basis of market sentiment Save this video and watch it to apply the formula on the Polymarket then, read the article

bodila

43,509 просмотров • 4 месяцев назад

Percy Liang teaches Stanford students how AI actually thinks now we get to learn it too. same lectures Stanford undergrads pay nearly $90,000 a year for. it teaches the math underneath every AI tool you use: / search and planning -> why agents loop the way they do / MDPs and RL -> why models reward-hack and how to spot it / probability and Bayes nets -> why outputs are calibrated (or aren't) / machine learning fundamentals -> what fine-tuning actually changes

Percy Liang teaches Stanford students how AI actually thinks now we get to learn it too. same lectures Stanford undergrads pay nearly $90,000 a year for. it teaches the math underneath every AI tool you use: / search and planning -> why agents loop the way they do / MDPs and RL -> why models reward-hack and how to spot it / probability and Bayes nets -> why outputs are calibrated (or aren't) / machine learning fundamentals -> what fine-tuning actually changes

dunik

18,446 просмотров • 3 месяцев назад

This 50-minute lecture by the scientist who built trading algorithms for Morgan Stanley and Lehman Brothers will teach you more about how machine learning actually works in algorithmic trading than most $1,000 courses ever will. Bookmark this and watch it tonight. It's the highest-leverage thing you can do if you want to understand how professional trading bots are really built. Then read the article below.

This 50-minute lecture by the scientist who built trading algorithms for Morgan Stanley and Lehman Brothers will teach you more about how machine learning actually works in algorithmic trading than most $1,000 courses ever will. Bookmark this and watch it tonight. It's the highest-leverage thing you can do if you want to understand how professional trading bots are really built. Then read the article below.

Goaty

79,842 просмотров • 4 месяцев назад

a playlist of 30 youtube videos to learn machine learning fundamentals from scratch if you're struggling on where to start learning ML, this list goes this "Machine Learning: Teach by Doing" is a solid choice to learn both theory and code. (1) Introduction to Machine Learning Teach by Doing: (2) What is Machine Learning? History of Machine Learning: (3) Types of ML Models: (4) 6 steps of any ML project: (5) Install Python and VSCode and run your first code: (6) Linear Classifiers Part 1: (7) Linear Classifiers Part 2: (8) Jupyter Notebook, Numpy and Scikit-Learn: (9) Running the Random Linear Classifier Algorithm in Python: (10) The oldest ML model - Perceptron: (11) Coding the Perceptron: (12) Perceptron Convergence Theorem: (13) Magic of features in Machine Learning: (14) One hot encoding: (15) Logistic Regression Part 1: (16) Cross Entropy Loss: (17) How gradient descent works: (18) Logistic Regression from scratch in Python: (19) Introduction to Regularization: (20) Implementing Regularization in Python: (21) Linear Regression Introduction: (22) Ordinary Least Squares step by step implementation: (23) Ridge regression fundamentals and intuition: (24) Regression recap for interviews: (25) Neural network architecture in 30 minutes: (26) Backpropagation intuition: (27) Neural network activation functions: (28) Momentum in gradient descent: (29) Hands on neural network training in Python: (30) Introduction to Convolutional Neural Networks (CNNs):

a playlist of 30 youtube videos to learn machine learning fundamentals from scratch if you're struggling on where to start learning ML, this list goes this "Machine Learning: Teach by Doing" is a solid choice to learn both theory and code. (1) Introduction to Machine Learning Teach by Doing: (2) What is Machine Learning? History of Machine Learning: (3) Types of ML Models: (4) 6 steps of any ML project: (5) Install Python and VSCode and run your first code: (6) Linear Classifiers Part 1: (7) Linear Classifiers Part 2: (8) Jupyter Notebook, Numpy and Scikit-Learn: (9) Running the Random Linear Classifier Algorithm in Python: (10) The oldest ML model - Perceptron: (11) Coding the Perceptron: (12) Perceptron Convergence Theorem: (13) Magic of features in Machine Learning: (14) One hot encoding: (15) Logistic Regression Part 1: (16) Cross Entropy Loss: (17) How gradient descent works: (18) Logistic Regression from scratch in Python: (19) Introduction to Regularization: (20) Implementing Regularization in Python: (21) Linear Regression Introduction: (22) Ordinary Least Squares step by step implementation: (23) Ridge regression fundamentals and intuition: (24) Regression recap for interviews: (25) Neural network architecture in 30 minutes: (26) Backpropagation intuition: (27) Neural network activation functions: (28) Momentum in gradient descent: (29) Hands on neural network training in Python: (30) Introduction to Convolutional Neural Networks (CNNs):

ℏεsam

117,748 просмотров • 1 год назад

A recent Huberman Lab and @SkoolAfter collaboration was directed at focus and how to use visual focus to improve mental focus for sake of learning. In the full animation, it describes those tools and techniques, and it covers ADHD, as well as the neurotransmitter systems that underlie focus, and some of the compounds that can increase focus. It also covers optimal times of day to learn, and optimal durations of learning bouts. Anyone interested in improving their ability to focus and learn ought to find value and practical applications by watching the animation.

A recent Huberman Lab and @SkoolAfter collaboration was directed at focus and how to use visual focus to improve mental focus for sake of learning. In the full animation, it describes those tools and techniques, and it covers ADHD, as well as the neurotransmitter systems that underlie focus, and some of the compounds that can increase focus. It also covers optimal times of day to learn, and optimal durations of learning bouts. Anyone interested in improving their ability to focus and learn ought to find value and practical applications by watching the animation.

Andrew D. Huberman, Ph.D.

319,475 просмотров • 3 лет назад

if you're struggling on where to start learning ML, here’s a playlist of 30 youtube videos to learn machine learning fundamentals from scratch "Machine Learning: Teach by Doing" is a solid choice to learn both theory and code. (1) Introduction to Machine Learning Teach by Doing: (2) What is Machine Learning? History of Machine Learning: (3) Types of ML Models: (4) 6 steps of any ML project: (5) Install Python and VSCode and run your first code: (6) Linear Classifiers Part 1: (7) Linear Classifiers Part 2: (8) Jupyter Notebook, Numpy and Scikit-Learn: (9) Running the Random Linear Classifier Algorithm in Python: (10) The oldest ML model - Perceptron: (11) Coding the Perceptron: (12) Perceptron Convergence Theorem: (13) Magic of features in Machine Learning: (14) One hot encoding: (15) Logistic Regression Part 1: (16) Cross Entropy Loss: (17) How gradient descent works: (18) Logistic Regression from scratch in Python: (19) Introduction to Regularization: (20) Implementing Regularization in Python: (21) Linear Regression Introduction: (22) Ordinary Least Squares step by step implementation: (23) Ridge regression fundamentals and intuition: (24) Regression recap for interviews: (25) Neural network architecture in 30 minutes: (26) Backpropagation intuition: (27) Neural network activation functions: (28) Momentum in gradient descent: (29) Hands on neural network training in Python: (30) Introduction to Convolutional Neural Networks (CNNs):

if you're struggling on where to start learning ML, here’s a playlist of 30 youtube videos to learn machine learning fundamentals from scratch "Machine Learning: Teach by Doing" is a solid choice to learn both theory and code. (1) Introduction to Machine Learning Teach by Doing: (2) What is Machine Learning? History of Machine Learning: (3) Types of ML Models: (4) 6 steps of any ML project: (5) Install Python and VSCode and run your first code: (6) Linear Classifiers Part 1: (7) Linear Classifiers Part 2: (8) Jupyter Notebook, Numpy and Scikit-Learn: (9) Running the Random Linear Classifier Algorithm in Python: (10) The oldest ML model - Perceptron: (11) Coding the Perceptron: (12) Perceptron Convergence Theorem: (13) Magic of features in Machine Learning: (14) One hot encoding: (15) Logistic Regression Part 1: (16) Cross Entropy Loss: (17) How gradient descent works: (18) Logistic Regression from scratch in Python: (19) Introduction to Regularization: (20) Implementing Regularization in Python: (21) Linear Regression Introduction: (22) Ordinary Least Squares step by step implementation: (23) Ridge regression fundamentals and intuition: (24) Regression recap for interviews: (25) Neural network architecture in 30 minutes: (26) Backpropagation intuition: (27) Neural network activation functions: (28) Momentum in gradient descent: (29) Hands on neural network training in Python: (30) Introduction to Convolutional Neural Networks (CNNs):

ℏεsam

109,447 просмотров • 1 год назад

Sukhendu Das, IIT Madras professor: "Jump Trading pays a 23 year old $500,000 a year for one skill: given the odds, pick the one move that makes the most money on average. that rule is the first thing this professor proves on a free chalkboard, and it costs nothing" this free lecture is the decision rule sitting under every trading bot and every hedge fund model on the planet. strip away the AI and the neural nets and one question is left. given everything you know right now, what is the single best bet. there is a mathematically correct answer. it is called the Bayes decision rule, and it turns a probability into the one action that maximizes your payoff. every profitable bot is just this rule running fast. Das builds it from zero, on a blackboard, in notation a beginner can follow. the odds go in, the optimal bet comes out. Wall Street rents this out as a $500,000 edge. an IIT professor hands you the rule itself for nothing.

Sukhendu Das, IIT Madras professor: "Jump Trading pays a 23 year old $500,000 a year for one skill: given the odds, pick the one move that makes the most money on average. that rule is the first thing this professor proves on a free chalkboard, and it costs nothing" this free lecture is the decision rule sitting under every trading bot and every hedge fund model on the planet. strip away the AI and the neural nets and one question is left. given everything you know right now, what is the single best bet. there is a mathematically correct answer. it is called the Bayes decision rule, and it turns a probability into the one action that maximizes your payoff. every profitable bot is just this rule running fast. Das builds it from zero, on a blackboard, in notation a beginner can follow. the odds go in, the optimal bet comes out. Wall Street rents this out as a $500,000 edge. an IIT professor hands you the rule itself for nothing.

Rossst.03

69,755 просмотров • 15 дней назад

I want you to really sit with this for a moment 👁️ Masonic 33rd Degree 🦉 A lecture on Energy Manipulation. Not conspiracy. Not speculation. An actual lecture. While the public is taught matter, mechanics, and materialism… the highest levels are taught energy, vibration, and control. ⚡ Different knowledge. Different rules. Different reality. They don’t rule through force. They rule through understanding how reality works. And now you’re seeing it. 🔺✨

I want you to really sit with this for a moment 👁️ Masonic 33rd Degree 🦉 A lecture on Energy Manipulation. Not conspiracy. Not speculation. An actual lecture. While the public is taught matter, mechanics, and materialism… the highest levels are taught energy, vibration, and control. ⚡ Different knowledge. Different rules. Different reality. They don’t rule through force. They rule through understanding how reality works. And now you’re seeing it. 🔺✨

vegastar

116,368 просмотров • 7 месяцев назад

Every time you accepted a salary, chose a price, or walked into a negotiation, the other person was running GAME THEORY in their head. You were guessing. This 1-hour Yale lecture by Professor Ben Polak will permanently change how you read people and make decisions. Most MBAs pay $150k to learn this. Yale posted it for free:

Every time you accepted a salary, chose a price, or walked into a negotiation, the other person was running GAME THEORY in their head. You were guessing. This 1-hour Yale lecture by Professor Ben Polak will permanently change how you read people and make decisions. Most MBAs pay $150k to learn this. Yale posted it for free:

Evan Luthra

1,661,176 просмотров • 5 месяцев назад

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Jürgen Schmidhuber

232,370 просмотров • 6 месяцев назад

HARVARD FILMED THE FIRST LECTURE OF THEIR MOST POPULAR STATISTICS COURSE - TAUGHT BY A PROFESSOR WHOSE STUDENTS CALL HIM THE BEST TEACHER THEY HAVE EVER HAD - AND IT PROVES WHY EVEN ISAAC NEWTON GOT PROBABILITY WRONG This is Joe Blitzstein, Harvard, Statistics 110, lecture 1. He has won Harvard's Excellence in Teaching award multiple times, his textbook Introduction to Probability is used in over 200 universities worldwide, and his online course has been taken by over 2 million people across 190 countries. He opens by saying that after a few weeks of this course you will easily solve calculations that 300 years ago required consulting Isaac Newton - and Newton's intuition was still wrong. He traces probability to Fermat and Pascal writing letters back and forth in the 1650s analyzing gambling games. No one had mathematically derived the rules before. They invented the subject by betting on dice in correspondence. Then he shows why the naive definition - probability equals favorable outcomes divided by total outcomes - breaks immediately. Ask what the probability of life on Neptune is. Either there is or there isn't. By the naive definition the answer is 1/2. So is the probability of intelligent life on Neptune. Something is severely wrong. Then the multiplication rule. Two types of ice cream cone and three flavors gives 6 combinations - not because you memorized it but because you can draw a tree and count branches. Every counting problem in the course is just a bigger version of that tree. Then binomial coefficients - n choose k counts the number of ways to select k objects from n when order doesn't matter. The full house in poker falls out in 4 lines of multiplication once you understand the tree. Watch the moment he fills in the sampling table - with or without replacement, order matters or doesn't. Three of the four boxes are immediate from the multiplication rule. The fourth requires a proof he saves for next lecture. That one box is harder than the other three combined. A data scientist I know rewatched this lecture before switching careers into statistics. Said it was the first time probability felt like a system with rules rather than a collection of tricks. Free on YouTube, Harvard, over 2 million views. bookmark this and watch later - after this lecture you will never again confuse equally likely with obviously true

HARVARD FILMED THE FIRST LECTURE OF THEIR MOST POPULAR STATISTICS COURSE - TAUGHT BY A PROFESSOR WHOSE STUDENTS CALL HIM THE BEST TEACHER THEY HAVE EVER HAD - AND IT PROVES WHY EVEN ISAAC NEWTON GOT PROBABILITY WRONG This is Joe Blitzstein, Harvard, Statistics 110, lecture 1. He has won Harvard's Excellence in Teaching award multiple times, his textbook Introduction to Probability is used in over 200 universities worldwide, and his online course has been taken by over 2 million people across 190 countries. He opens by saying that after a few weeks of this course you will easily solve calculations that 300 years ago required consulting Isaac Newton - and Newton's intuition was still wrong. He traces probability to Fermat and Pascal writing letters back and forth in the 1650s analyzing gambling games. No one had mathematically derived the rules before. They invented the subject by betting on dice in correspondence. Then he shows why the naive definition - probability equals favorable outcomes divided by total outcomes - breaks immediately. Ask what the probability of life on Neptune is. Either there is or there isn't. By the naive definition the answer is 1/2. So is the probability of intelligent life on Neptune. Something is severely wrong. Then the multiplication rule. Two types of ice cream cone and three flavors gives 6 combinations - not because you memorized it but because you can draw a tree and count branches. Every counting problem in the course is just a bigger version of that tree. Then binomial coefficients - n choose k counts the number of ways to select k objects from n when order doesn't matter. The full house in poker falls out in 4 lines of multiplication once you understand the tree. Watch the moment he fills in the sampling table - with or without replacement, order matters or doesn't. Three of the four boxes are immediate from the multiplication rule. The fourth requires a proof he saves for next lecture. That one box is harder than the other three combined. A data scientist I know rewatched this lecture before switching careers into statistics. Said it was the first time probability felt like a system with rules rather than a collection of tricks. Free on YouTube, Harvard, over 2 million views. bookmark this and watch later - after this lecture you will never again confuse equally likely with obviously true

Abyzon

652,660 просмотров • 1 месяц назад

Introducing the 18th S R Nathan Fellow, Dr Shawn Lum, Senior Lecturer at The Asian School of the Environment, Nanyang Technological University. Through his work in research, teaching and conservation, Dr Lum has spent decades exploring the relationship between people and the natural world. Beyond the lecture hall and the field, he also has a lighter side — shaped by the experiences and stories that sparked his lifelong interest in nature. In his upcoming lecture series, “Singapore, Nature and a World Reimagined,” Dr Lum explores Singapore’s journey from the bold vision of a “Garden City” to the evolving ambition of becoming a “City in Nature.” The lectures examine how this transformation took shape over decades through careful planning, ecological understanding, and a shift in how society views nature. They also make a case for why biodiversity matters — not only for ecosystems, but for the well-being of people and communities, and how Singapore can better protect and value nature at home and beyond. 🎟️Sign up for his first lecture on 9 March 2026 here: 📺Watch his lecture series live on IPS’ Facebook page: 🌳Learn more about Dr Lum’s lectures:

Introducing the 18th S R Nathan Fellow, Dr Shawn Lum, Senior Lecturer at The Asian School of the Environment, Nanyang Technological University. Through his work in research, teaching and conservation, Dr Lum has spent decades exploring the relationship between people and the natural world. Beyond the lecture hall and the field, he also has a lighter side — shaped by the experiences and stories that sparked his lifelong interest in nature. In his upcoming lecture series, “Singapore, Nature and a World Reimagined,” Dr Lum explores Singapore’s journey from the bold vision of a “Garden City” to the evolving ambition of becoming a “City in Nature.” The lectures examine how this transformation took shape over decades through careful planning, ecological understanding, and a shift in how society views nature. They also make a case for why biodiversity matters — not only for ecosystems, but for the well-being of people and communities, and how Singapore can better protect and value nature at home and beyond. 🎟️Sign up for his first lecture on 9 March 2026 here: 📺Watch his lecture series live on IPS’ Facebook page: 🌳Learn more about Dr Lum’s lectures:

Institute of Policy Studies - IPS

237,200 просмотров • 6 месяцев назад

There're few who can deliver both great AI research and charismatic talks. OpenAI Chief Scientist Ilya Sutskever is one of them. I watched Ilya's lecture at Simons Institute, where he delved into why unsupervised learning works through the lens of compression. Sharing my notes: - Kolmogorov compressor is the theoretical shortest-length program that produces a dataset. SGD is a practical approximation of the Kolmogorov search that finds an implicit program embedded in the weights of a soft computer, i.e. big Transformers. - Unsupervised learning is about computing the conditional Kolmogorov complexity of a target dataset given an unlabelled corpus, i.e. K(Y|X) - Theory tells us that optimizing for K(X, Y), the joint complexity, is as good as K(Y|X). So simply throw all data into the mix, and "just compress everything". - Joint compression is maximum likelihood over the giant concatenated dataset. - Ilya cites iGPT, Chen et al. 2020, to illustrate the ideas. iGPT is an image compressor that learns to predict the next pixel using a 1D sequence model. This is a phenomenal lecture, very accessible, and sometimes quite entertaining. YouTube: Lecture page:

There're few who can deliver both great AI research and charismatic talks. OpenAI Chief Scientist Ilya Sutskever is one of them. I watched Ilya's lecture at Simons Institute, where he delved into why unsupervised learning works through the lens of compression. Sharing my notes: - Kolmogorov compressor is the theoretical shortest-length program that produces a dataset. SGD is a practical approximation of the Kolmogorov search that finds an implicit program embedded in the weights of a soft computer, i.e. big Transformers. - Unsupervised learning is about computing the conditional Kolmogorov complexity of a target dataset given an unlabelled corpus, i.e. K(Y|X) - Theory tells us that optimizing for K(X, Y), the joint complexity, is as good as K(Y|X). So simply throw all data into the mix, and "just compress everything". - Joint compression is maximum likelihood over the giant concatenated dataset. - Ilya cites iGPT, Chen et al. 2020, to illustrate the ideas. iGPT is an image compressor that learns to predict the next pixel using a 1D sequence model. This is a phenomenal lecture, very accessible, and sometimes quite entertaining. YouTube: Lecture page:

Jim Fan

822,498 просмотров • 3 лет назад

Learn to train an LLM with distributed data while ensuring privacy using federated learning in a new two-part short course, Intro to Federated Learning and Federated Fine-tuning of LLMs with Private Data, created with Flower and taught by Daniel J. Beutel and Nicholas Lane. Federated learning allows a single model to be trained across multiple devices, such as phones, or multiple organizations, such as hospitals, without the need to share data to a central server. This two-part course gives you an introduction to federated learning, and then teaches you how to fine-tune your large language model with distributed data using Flower Lab’s open source federated learning framework. You’ll learn: - How to use federated learning to train a variety of models, ranging from speech and vision models to LLMs, across distributed data while offering data privacy options to users and organizations. - Privacy Enhancing Technologies like differential privacy (DP), which obscures individual data by adding calibrated noise to query results. - Two variants of differential privacy - Central and Local - and how to choose depending on your use case. - How to measure and decrease bandwidth usage to make federated learning more practical and efficient with techniques like using pre-trained models and Parameter-Efficient Fine-Tuning - How federated LLM fine-tuning reduces the risk of leaking training data. Sign up here!

Learn to train an LLM with distributed data while ensuring privacy using federated learning in a new two-part short course, Intro to Federated Learning and Federated Fine-tuning of LLMs with Private Data, created with Flower and taught by Daniel J. Beutel and Nicholas Lane. Federated learning allows a single model to be trained across multiple devices, such as phones, or multiple organizations, such as hospitals, without the need to share data to a central server. This two-part course gives you an introduction to federated learning, and then teaches you how to fine-tune your large language model with distributed data using Flower Lab’s open source federated learning framework. You’ll learn: - How to use federated learning to train a variety of models, ranging from speech and vision models to LLMs, across distributed data while offering data privacy options to users and organizations. - Privacy Enhancing Technologies like differential privacy (DP), which obscures individual data by adding calibrated noise to query results. - Two variants of differential privacy - Central and Local - and how to choose depending on your use case. - How to measure and decrease bandwidth usage to make federated learning more practical and efficient with techniques like using pre-trained models and Parameter-Efficient Fine-Tuning - How federated LLM fine-tuning reduces the risk of leaking training data. Sign up here!

Andrew Ng

64,609 просмотров • 2 лет назад

Did a very different format with Reiner Pope – a blackboard lecture where he walks through how frontier LLMs are trained and served. It's shocking how much you can deduce about what the labs are doing from a handful of equations, public API prices, and some chalk. It’s a bit technical, but I encourage you to hang in there - it’s really worth it. There are less than a handful of people who understand the full stack of AI, from chip design to model architecture, as well as Reiner. It was a real delight to learn from him. Recommend watching this one on YouTube so you can see the chalkboard. 0:00:00 – How batch size affects token cost and speed 0:31:59 – How MoE models are laid out across GPU racks 0:47:02 – How pipeline parallelism spreads model layers across racks 1:03:27 – Why Ilya said, “As we now know, pipelining is not wise.” 1:18:49 – Because of RL, models may be 100x over-trained beyond Chinchilla-optimal 1:32:52 – Deducing long context memory costs from API pricing 2:03:52 – Convergent evolution between neural nets and cryptography

Did a very different format with Reiner Pope – a blackboard lecture where he walks through how frontier LLMs are trained and served. It's shocking how much you can deduce about what the labs are doing from a handful of equations, public API prices, and some chalk. It’s a bit technical, but I encourage you to hang in there - it’s really worth it. There are less than a handful of people who understand the full stack of AI, from chip design to model architecture, as well as Reiner. It was a real delight to learn from him. Recommend watching this one on YouTube so you can see the chalkboard. 0:00:00 – How batch size affects token cost and speed 0:31:59 – How MoE models are laid out across GPU racks 0:47:02 – How pipeline parallelism spreads model layers across racks 1:03:27 – Why Ilya said, “As we now know, pipelining is not wise.” 1:18:49 – Because of RL, models may be 100x over-trained beyond Chinchilla-optimal 1:32:52 – Deducing long context memory costs from API pricing 2:03:52 – Convergent evolution between neural nets and cryptography

Dwarkesh Patel

1,292,143 просмотров • 4 месяцев назад

$Last semester a professor pulled me aside after class. "You trade prediction markets?" - he asked. I said yeah, mostly vibes. He laughed and said that's why you're losing. He spent 20 minutes on a whiteboard showing me Bayes' Theorem, Kelly Criterion, and LMSR pricing. Said this is literally how quant funds approach binary markets. I took photos of everything. Went home, opened Claude, pasted my notes and said: “build me a bot that does exactly this 3 hours later I had a working Bayesian trading bot for Polymarket BTC markets. Here's the math it runs on every single trade: - Bayesian update update beliefs before entering P(H|D) = P(D|H) · P(H) / P(D) - Edge detection -only trade if you have an edge EV = p̂ − p_market - Kelly sizing-bet exactly the right amount f* = EV / p̂(1−p̂) - LMSR pricing how Polymarket actually prices markets p(q) = e^{q/b} / Σe^{qⱼ/b} (it's literally the softmax function from neural networks) 65% win rate. Losses happen. Kelly keeps them from blowing the account. Compounding does the rest. My CLAWD BOT turned $100 into $1000 in 30min The professor was right. Vibes don't compound. Math does.$

Last semester a professor pulled me aside after class. "You trade prediction markets?" - he asked. I said yeah, mostly vibes. He laughed and said that's why you're losing. He spent 20 minutes on a whiteboard showing me Bayes' Theorem, Kelly Criterion, and LMSR pricing. Said this is literally how quant funds approach binary markets. I took photos of everything. Went home, opened Claude, pasted my notes and said: “build me a bot that does exactly this 3 hours later I had a working Bayesian trading bot for Polymarket BTC markets. Here's the math it runs on every single trade: - Bayesian update update beliefs before entering P(H|D) = P(D|H) · P(H) / P(D) - Edge detection -only trade if you have an edge EV = p̂ − p_market - Kelly sizing-bet exactly the right amount f* = EV / p̂(1−p̂) - LMSR pricing how Polymarket actually prices markets p(q) = e^{q/b} / Σe^{qⱼ/b} (it's literally the softmax function from neural networks) 65% win rate. Losses happen. Kelly keeps them from blowing the account. Compounding does the rest. My CLAWD BOT turned $100 into $1000 in 30min The professor was right. Vibes don't compound. Math does.

AdiiX

174,137 просмотров • 6 месяцев назад

Jensen Huang on "distillation" On his new interview with axios, he was asked this question "Should open source model companies be allowed to distill closed models" "Distillation—learning from AI, learning from other people, and learning from other sources of knowledge, is fundamental to intelligence. We are constantly learning from other people. I am learning from you through the questions you are asking, and you are learning from me. All day long, we are learning from one another. AI also has to learn from something. The original AI models, whether they were open or closed, were trained on previously created knowledge from the internet. Now, AI is generating more content than humans. In a few more years, the internet could be 99% AI-generated content, and that content will have been created by some form of AI. As a result, AI systems will constantly be distilling knowledge and intelligence from other AI systems. The fact that AI can learn is a good thing. We want AI systems to be intelligent because a smarter AI can also be a safer AI." ---- From "Axios" YouTube channel, (full video link in comment)

Jensen Huang on "distillation" On his new interview with axios, he was asked this question "Should open source model companies be allowed to distill closed models" "Distillation—learning from AI, learning from other people, and learning from other sources of knowledge, is fundamental to intelligence. We are constantly learning from other people. I am learning from you through the questions you are asking, and you are learning from me. All day long, we are learning from one another. AI also has to learn from something. The original AI models, whether they were open or closed, were trained on previously created knowledge from the internet. Now, AI is generating more content than humans. In a few more years, the internet could be 99% AI-generated content, and that content will have been created by some form of AI. As a result, AI systems will constantly be distilling knowledge and intelligence from other AI systems. The fact that AI can learn is a good thing. We want AI systems to be intelligent because a smarter AI can also be a safer AI." ---- From "Axios" YouTube channel, (full video link in comment)

Rohan Paul

443,477 просмотров • 1 месяц назад

Drake has always been strategic, throwback to how he was thinking when it came to picking a label. He discussed Roc-A-Fella Records or Def Jam, listen closely. New young artists can learn a lot here, but this also tells you nothing Drake accomplished is by luck, homie really had a locked in vision from day one.

Drake has always been strategic, throwback to how he was thinking when it came to picking a label. He discussed Roc-A-Fella Records or Def Jam, listen closely. New young artists can learn a lot here, but this also tells you nothing Drake accomplished is by luck, homie really had a locked in vision from day one.

Hip-Hop Unison

34,357 просмотров • 3 месяцев назад

Really authentic and vulnerable post from jesse.base.eth this morning. A lot was said and there’s clearly a major shift happening in how Base is thinking about the future. If you don’t want to read the whole thing, here’s my quick summary and thoughts 👇

Really authentic and vulnerable post from jesse.base.eth this morning. A lot was said and there’s clearly a major shift happening in how Base is thinking about the future. If you don’t want to read the whole thing, here’s my quick summary and thoughts 👇

Bill- Late Night Onchain

13,917 просмотров • 1 месяц назад