Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore fundamental challenges and innovations in unsupervised learning through this 29-minute conference talk from MLOps World. Discover why classical unsupervised learning problems remain highly relevant in today's data landscape, despite being designed for low-dimensional tabular data decades ago. Learn how neural embeddings have transformed unstructured data accessibility, creating a high-dimensional data environment where traditional assumptions and intuitions fail. Examine the limitations of standard algorithms when applied to high-dimensional data and understand why clustering, dimension reduction, anomaly detection, and density estimation require new approaches. Gain insights into building next-generation algorithms specifically designed for modern high-dimensional data challenges. The presentation is delivered by Leland McInnes, a researcher at the Tutte Institute for Mathematics and Computing, who has contributed significantly to the field through algorithms like UMAP for dimension reduction and accelerated HDBSCAN for clustering, along with maintaining various open source data science tools including PyNNDescent, DataMapPlot, and Toponymy.
