Published Mar 21, 2022

Matrix Factorization For k-Means

Sibylle Hess from TU Eindhoven delves into the intersection of matrix factorization and neural networks, revealing how these concepts enhance clustering techniques like k-means and spectral clustering to fortify AI against adversarial attacks, improve shape recognition, and redefine data representation within neural models.

Episode Highlights

Topics covered

Popular Clips

Episode Highlights

Related Episodes

Power K-Means
Answers 383 questions
Explainable K-Means
Answers 383 questions
k-means clustering
Answers 383 questions
K-Means in Practice
Answers 383 questions
Customer Clustering
Answers 383 questions
Quantum K-Means
Answers 383 questions
[MINI] k-means clustering
Answers 383 questions
k-means Image Segmentation
Answers 383 questions
Breathing K-Means
Answers 383 questions
Fair Hierarchical Clustering
Answers 383 questions
Drug Discovery with Machine Learning
Answers 383 questions
Graph Transformations
Answers 383 questions
Machine Learning Done Wrong
Answers 383 questions
Reproducing Deep Learning Models
Answers 383 questions
ML Ops Best Practices
Answers 383 questions

Dexa/Kyle Polich

Matrix Factorization For k-Means

Topics covered

Popular Clips

Enhanced Neural Networks

Robust AI Defense

Under the Hood

Clustering and Deep Learning

Embedded Space Insights

Exciting Research Opportunities

Softmax and K-means Clustering

Confidence and Loss Functions

Clustering in Embedded Space

Episode Highlights

Matrix Factorization

Spectral Clustering

Embedded Space and Clustering

Adversarial Attacks

Related Episodes

Power K-Means

Explainable K-Means

k-means clustering

K-Means in Practice

Customer Clustering

Quantum K-Means

[MINI] k-means clustering

k-means Image Segmentation

Breathing K-Means

Fair Hierarchical Clustering

Drug Discovery with Machine Learning

Graph Transformations

Machine Learning Done Wrong

Reproducing Deep Learning Models

ML Ops Best Practices

Matrix Factorization For k-Means

Topics covered

Popular Clips

Episode Highlights

Matrix Factorization

Spectral ClusteringSibylle Hess explores the transformative potential of spectral clustering in neural networks, emphasizing its ability to enhance shape recognition and robustness. This approach challenges traditional methods, offering new insights into AI's future capabilities.

Spectral Clustering

Embedded Space and ClusteringSibylle Hess discusses the intricate properties of embedded space in neural networks, highlighting its impact on clustering and model confidence. Her insights reveal how these structures can influence neural network predictions and defenses against adversarial attacks.

Embedded Space and Clustering

Adversarial AttacksSibylle Hess explores the challenges neural networks face with adversarial attacks and the potential of clustering methods to enhance their robustness. She discusses innovative approaches that could shift the paradigm from traditional neural networks to more resilient models.

Adversarial Attacks

Related Episodes