Automatic differentiation and machine learning

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=pdSCHtPJ4B0


Duration: 57:01
3,410 views
43

Derivatives, mostly in the form of gradients and Hessians, are ubiquitous in machine learning. Automatic differentiation (AD) is a technique for calculating derivatives efficiently and accurately, established in fields such as computational fluid dynamics, nuclear engineering, and atmospheric sciences. Despite its advantages and use in other fields, machine learning practitioners have been little influenced by AD and make scant use of available tools. We survey the intersection of AD and machine learning, cover applications where AD has the potential to make a big impact, and report on some recent developments in the adoption of this technique. We also aim to dispel some misconceptions that we would contend have impeded the use of AD within the machine learning community.



Other Videos By Microsoft Research


2016-06-27A Scalable Cloud-Sensor Architecture for the Internet of Things
2016-06-27Think like a programmer
2016-06-27How Different is Different: A Perspective on Music Similarity in Indian Art Music
2016-06-27A Visual Programming Language for building Artificial Biochemistries in Haskell
2016-06-27WIPTTE: Sponsor Talks: Surface Pro Digital Pen Design
2016-06-27Peer-to-Peer Rental Markets in the Sharing Economy
2016-06-27Inverting RANSAC: Global Model Detection via Inlier Rate Estimation
2016-06-27Social Media Predictive Analytics: Methods and Applications
2016-06-27Software Support for Efficient Use of Modern Parallel Systems
2016-06-27Microsoft Research Gender Diversity Lecture Series 1 Kickoff: Implicit Biases and Stereotype Threat
2016-06-27Automatic differentiation and machine learning
2016-06-27Toward Causal Machine Learning
2016-06-27Fast Dual-Tree k-means with Bounded Single-Iteration Runtime
2016-06-27Fast Interpolating Bounded Model Checking
2016-06-27Internet of Things: Support IP Smart Objects over Bluetooth Smart
2016-06-27UW - MSR Machine Learning workshop 2015 - Session 1
2016-06-27Towards a Learning Theory of Causation
2016-06-27A Real-time Augmented Reality Processor and a Smart Glasses System
2016-06-27Automating Program Transformations based on Examples of Systematic Edits
2016-06-27Causal Inference and Domain Adaptation.
2016-06-27Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data.


Tags:
microsoft research
machine learning
deep neural networks