- Видео 42
- Просмотров 32 313
Krishnaswamy Lab
Добавлен 15 фев 2021
Visualizing Data with PHATE | Unsupervised Learning for Big Data
PHATE is a powerful tool for visualizing high-dimensional data with (much) higher fidelity than PCA and less randomness than tSNE. This lecture explains the motivation behind PHATE and describes how the algorithm works.
This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy.
Unsupervised learning is perhaps the most beautiful and most frequently astonishing area of machine learning. It doesn't need to guzzle tons of labeled data to solve problems by brute force. Instead, it uses elegant mathematical principles to understand (in some sense) the data itself and the patterns underlying it.
Because this is a young...
This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy.
Unsupervised learning is perhaps the most beautiful and most frequently astonishing area of machine learning. It doesn't need to guzzle tons of labeled data to solve problems by brute force. Instead, it uses elegant mathematical principles to understand (in some sense) the data itself and the patterns underlying it.
Because this is a young...
Просмотров: 1 239
Видео
Generative Adversarial Networks | Unsupervised Learning for Big Data
Просмотров 1692 года назад
If you've heard of "deep fakes," then you're familiar with what GANs can do. GANs are a fascinating form of deep learning: by modeling a rivalry between competing agents, GANs can learn sophisticated structures in unlabeled data, and generate their own samples of data (art, faces, music) with compelling accuracy. This lecture introduces GANs and covers some of the common problems afflicting the...
Word-to-Vector Embeddings | Unsupervised Learning for Big Data
Просмотров 952 года назад
What if we could understand and manipulate language with the ease of manipulating numbers? This lecture covers techniques for encoding relationships between words into relationships between vectors. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most ...
Diffusion Maps | Unsupervised Learning for Big Data
Просмотров 3,8 тыс.2 года назад
One of the most elegant methods for dimensionality reduction, which makes an analogy to the diffusion of heat to learn a robust coordinate system for data on a manifold. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most frequently astonishing area o...
Finding Eigenvectors | Unsupervised Learning for Big Data
Просмотров 1422 года назад
This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most frequently astonishing area of machine learning. It doesn't need to guzzle tons of labeled data to solve problems by brute force. Instead, it uses elegant mathematical principles to understand (in so...
Backpropogation | Unsupervised Learning for Big Data
Просмотров 852 года назад
At the heart of any neural network is backpropogation, the algorithm which allows it to self-correct in response to feedback. This lecture describes the intuition behind it, and presents the "Four Fundamental Equations" of backpropogation. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learnin...
CNN Autoencoders | Unsupervised Learning for Big Data
Просмотров 1,2 тыс.2 года назад
Autoencoders are marvelous tools for manipulating the dimension of arbitrary data, for visualization, classification, denoising, or generative purposes. But what about image data? Can we outfit autoencoders with the mechanisms to properly leverage the spatial layout of images? This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Sm...
Cross Entropy Loss | Unsupervised Learning for Big Data
Просмотров 1742 года назад
Cross Entropy Loss is an effective teaching strategy for neural networks learning to classify data. This lecture motivates and derives cross entropy loss. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most frequently astonishing area of machine learn...
Autoencoders | Unsupervised Learning for Big Data
Просмотров 2362 года назад
Autoencoders are the poster child of Unsupervised Learning. By dynamically compressing and decompressing data, they can learn to extract meaningful features that can used to visualize the data, denoise the data, and even generate new bits of data. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised...
Visualizing Data with tSNE | Unsupervised Learning for Big Data
Просмотров 2522 года назад
tSNE is one of the most popular methods for visualizing high-dimensional data. This lecture explains how and why it works. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most frequently astonishing area of machine learning. It doesn't need to guzzle t...
Hierarchical Clustering and Community Detection | Unsupervised Learning for Big Data
Просмотров 4702 года назад
This lecture describes a prominent approach to clustering data, covering simple applications of the approach and culminating with an explanation of the popular Louvain method for community detection. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most...
PCA Derivation | Unsupervised Learning for Big Data
Просмотров 1442 года назад
Principal Component Analysis features so prominently in the world of data analysis that it's been re-discovered several times under different names in different fields. Here, we present one derivation which begins with the motivation of maintaining the maximal information in a small number of dimensions and uses Lagrange multipliers to solve a linear optimization problem. This is a part of a se...
Variational Autoencoders | Unsupervised Learning for Big Data
Просмотров 2052 года назад
Variational Autoencoders | Unsupervised Learning for Big Data
Multidimensional Scaling, Distances, and Inner Products | Unsupervised Learning for Big Data
Просмотров 2282 года назад
Multidimensional Scaling, Distances, and Inner Products | Unsupervised Learning for Big Data
Clustering and K-Means | Unsupervised Learning for Big Data
Просмотров 1012 года назад
How do you learn about data given nothing but the data itself? One approach is to reduce the number of samples you need to understand, by clustering the datapoints into broadly similar categories and then taking archetypes from these categories. This lecture motivates clustering and describes a simple clustering algorithm. This is a part of a series of lectures from the Yale class "Unsupervised...
Graph Filtering and Denoising | Unsupervised Learning for Big Data
Просмотров 3142 года назад
This lecture uses spectral graph theory to describe a method of denoising graph-shaped data. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful and most frequently astonishing area of machine learning. It doesn't need to guzzle tons of labeled data to solve p...
Graph Neural Networks | Unsupervised Learning for Big Data
Просмотров 5602 года назад
Social networks, molecules, the inter-linkage of the internet all of these types of data can be described as graphs. This lecture introduces a class of neural networks created to learn from graph-shaped data. This is a part of a series of lectures from the Yale class "Unsupervised Learning for Big Data", taught by Professor Smita Krishnaswamy. Unsupervised learning is perhaps the most beautiful...
Information Theory Concepts | Unsupervised Learning for Big Data
Просмотров 732 года назад
Information Theory Concepts | Unsupervised Learning for Big Data
Spectral Clustering | Unsupervised Learning for Big Data
Просмотров 4322 года назад
Spectral Clustering | Unsupervised Learning for Big Data
Linear Transformations and Eigenvectors | Unsupervised Learning for Big Data
Просмотров 1522 года назад
Linear Transformations and Eigenvectors | Unsupervised Learning for Big Data
Kernel PCA | Unsupervised Learning for Big Data
Просмотров 4502 года назад
Kernel PCA | Unsupervised Learning for Big Data
Data Representation | Unsupervised Learning for Big Data
Просмотров 4562 года назад
Data Representation | Unsupervised Learning for Big Data
Sequential Models | Unsupervised Learning for Big Data
Просмотров 492 года назад
Sequential Models | Unsupervised Learning for Big Data
Denoising and Low Rank Approximation | Unsupervised Learning for Big Data
Просмотров 1992 года назад
Denoising and Low Rank Approximation | Unsupervised Learning for Big Data
Geometric Interpretation of the Covariance Matrix | Unsupervised Learning for Big Data
Просмотров 4472 года назад
Geometric Interpretation of the Covariance Matrix | Unsupervised Learning for Big Data
Neural Networks | Unsupervised Learning for Big Data
Просмотров 2142 года назад
Neural Networks | Unsupervised Learning for Big Data
PCA for Data Visualization | Unsupervised Learning for Big Data
Просмотров 1862 года назад
PCA for Data Visualization | Unsupervised Learning for Big Data
Probability Theory and Density Estimation | Unsupervised Learning for Big Data
Просмотров 1632 года назад
Probability Theory and Density Estimation | Unsupervised Learning for Big Data
Fourier and Wavelet Transforms Primer | Unsupervised Learning for Big Data
Просмотров 1472 года назад
Fourier and Wavelet Transforms Primer | Unsupervised Learning for Big Data
Sparse Coding and Dictionary Learning | Unsupervised Learning for Big Data
Просмотров 1,5 тыс.2 года назад
Sparse Coding and Dictionary Learning | Unsupervised Learning for Big Data
really appreciate that experts are making general educational videos to the public. much appreciated!!!!
very interesting
can I get these slides?
Excellent content, thank you!
Good lecture video. Anyone know where can I get the slides please ?
2:07 What is the complete integration formula? The screen is cut.
Thanks for these lectures! They are truly amazing! Rarely does someone explain the motivation behind these AI/statistical methods so well
I second this comment. Very intuitive explanations.
Thank you
Thank you for presenting it this way, by starting with the motivation behind the technique you framed everything into context ❤❤ Sometimes a Kaggle notebook isn't enough, you just need a human to explain it to you.
Thank you for this straightforward explanation!
Thank you!
This video is very intuitive and easy to understand. Thanks for the great explanation!
Thank you it's very informative. I have a question about the evaluation of autoencoders, what's the best way to evaluate an autoencoder for unsupervised learning?
Thank you Professor for making your lessons freely available on youtube! You rock!
Excellent explanation, but poor video quality!
totally agree - this great material deserves better quality :-)