This document summarizes hot topics from Microsoft Research, including distributed computing across multicore systems, natural user interfaces, understanding human motion through computer vision, machine learning from large datasets, and using machine learning to improve ad relevance. It discusses challenges like coordinating work across multiple CPU cores and distributing operating system state. It also highlights several Microsoft Research projects applying these areas like Barrelfish, a new operating system architecture, and work using probabilistic models and joint modeling to understand images and human motion.
10. The mythical man-month
If 1 accountant takes 8 hours to do my taxes,
then 8 accountants can do 8 people’s taxes in 8 hours.
But 8 accountants can’t do 1 tax return in 1 hour.
The mythical cpu-second
If it takes a 1-core machine 1 second to process 1 GB of data,
how long does a 64-core machine take to process 64GB?
Sometimes 1 second. Sometimes more.
11. Barrelfish
A new OS architecture for scalable
multicore systems
Approach: structure the OS as a distributed
system
Design principles:
• Make inter-core communication explicit
• Make OS structure hardware-neutral
• View state as replicated
12. A Multikernel vs Other Operating Systems
Spectrum of sharing and locking disciplines
Traditional OSes Multikernel
Shared state, Finer-grained Clustered objects, Distributed state,
one-big-lock locking partitioning replica maintenance
Multikernel
Windows and Unix variants are Research operating
evolving from the far left operating system:
towards the middle systems eg K42 Barrelfish
13. Barrelfish Structure
App App App App
OS node OS node OS node OS node
State State State Async messages State
replica replica replica replica
x86 x64 ARM GPU
Hardware interconnect
16. "This is a pivotal moment that will carry
with it a wave of change, the ripples of
which will reach far beyond video games“
STEVEN SPIELBERG
17. Understanding Human Motion
Andrew Blake, Kentaro Toyama,
Probablisitic tracking in a metric space,
Awarded the Marr Prize ,
IEEE International conference on Computer Vision, 2001
18. Understanding Human Motion
Ramanan Navaratnam, Andrew Fitzgibbon, Roberto Cipolla,
The Joint Manifold Model for Semi-supervised Multi-valued Regression,
IEEE International conference on Computer Vision, 2007
19. Understanding Images
J. Shotton, J. Winn, C. Rother, A. Criminisi, TextonBoost: Joint Appearance, Shape
and Context Modeling for Multi-Class Object Recognition and Segmentation.
European Conference on Computer Vision, 2006
25. Click/Ad Information Probability
1341201
1570165
Listing ID
2213187
9215433
+
Match Exact Match
Type Broad Match
ML-1
p(Click|Query, Ad)
Position SB-1
SB-2