Vijay's posts

Post has attachment

Mohan, here is a useful new paper in the space you have been talking about.

Deadlines are the single most important boost to productivity invented by humanity.

Post has attachment

How the Internet is accelerating mathematical progress, or, going beyond MOOCs...

There is a lovely story in Wired magazine that showed up on my FB page.

http://www.wired.com/wiredscience/2013/11/prime/all/

People may recall that on April 17, an unknown lecture at U New Hampshire shocked the mathematical world by establishing that there is a number N <= 70 000 000 s.t. that there are infinitely many pairs of primes whose difference is N. This was the first proof that infinitely many primes come in pairs.

Terence Tao, a Field Medalist at UCLA, then proposed a polymath project, polymath8 (http://polymathprojects.org/2013/06/04/polymath-proposal-bounded-gaps-between-primes/#comments) for mathematicians (and computer scientists!) from all over the world to collaborate on reducing this bound. In just five months, the bound was reduced to 4 680!

Just to establish that the Lone Mathematician still has a role to play, three days ago a post doctoral researcher from U Montreal sent in a preprint to arXiv reducing the number to 600!

The polymath idea is interesting! What is fascinating to me is that a whole lot of advanced mathematics is now happening collaboratively, in the open blogosphere, involving people who may not even have known about each other before they started collaborating online.

It seems that a new area of Massively Open Online Research has begun, at least in math. Wonder which other areas of research would be able to sustain this (without running afoul of credit / intellectual property / ....)

There is a lovely story in Wired magazine that showed up on my FB page.

http://www.wired.com/wiredscience/2013/11/prime/all/

People may recall that on April 17, an unknown lecture at U New Hampshire shocked the mathematical world by establishing that there is a number N <= 70 000 000 s.t. that there are infinitely many pairs of primes whose difference is N. This was the first proof that infinitely many primes come in pairs.

Terence Tao, a Field Medalist at UCLA, then proposed a polymath project, polymath8 (http://polymathprojects.org/2013/06/04/polymath-proposal-bounded-gaps-between-primes/#comments) for mathematicians (and computer scientists!) from all over the world to collaborate on reducing this bound. In just five months, the bound was reduced to 4 680!

Just to establish that the Lone Mathematician still has a role to play, three days ago a post doctoral researcher from U Montreal sent in a preprint to arXiv reducing the number to 600!

The polymath idea is interesting! What is fascinating to me is that a whole lot of advanced mathematics is now happening collaboratively, in the open blogosphere, involving people who may not even have known about each other before they started collaborating online.

It seems that a new area of Massively Open Online Research has begun, at least in math. Wonder which other areas of research would be able to sustain this (without running afoul of credit / intellectual property / ....)

Statistical methods strike again ... Fernando, do you have specific papers to recommend in this space?

Towards Open-domain Spoken Dialogue Systems

Speaker: Steve Young

Speaker Affiliation: Cambridge University

Host: Jim Glass and Victor Zue

Host Affiliation: MIT CSAIL

Date: Monday, October 28, 2013

Time: 4:30 PM to 5:30 PM

Refreshments Time: 4:15 PM

Location: 32-155

In contrast to traditional rule-based approaches to building spoken dialogue systems, recent research has shown that it is possible to implement all of the required functionality using statistical models trained using a combination of supervised learning and reinforcement learning. This new approach to spoken dialogue is based on the mathematics of partially observable Markov decision processes (POMDPs) in which user inputs are treated as observations of some underlying belief state, and system responses are determined by a policy which maps belief states into actions.

Virtually all current spoken dialogue systems are designed to operate in a specific carefully defined domain such as restaurant information, appointment booking, product installation support, etc. However, if voice is to become a significant input modality for accessing web-based information and services, then techniques will be needed to enable spoken dialogue systems to operate within open domains.

The first part of the talk will briefly review the basic ideas of POMDP dialogue systems as currently applied to closed-domains. Unlike many other areas of machine learning, spoken dialogue systems always have a user on-hand to provide supervision. Based on this idea, the second part of the talk describes a number of techniques by which implicit user supervision can allow a spoken dialogue system to adapt on-line to extended domains.

Towards Open-domain Spoken Dialogue Systems

Speaker: Steve Young

Speaker Affiliation: Cambridge University

Host: Jim Glass and Victor Zue

Host Affiliation: MIT CSAIL

Date: Monday, October 28, 2013

Time: 4:30 PM to 5:30 PM

Refreshments Time: 4:15 PM

Location: 32-155

In contrast to traditional rule-based approaches to building spoken dialogue systems, recent research has shown that it is possible to implement all of the required functionality using statistical models trained using a combination of supervised learning and reinforcement learning. This new approach to spoken dialogue is based on the mathematics of partially observable Markov decision processes (POMDPs) in which user inputs are treated as observations of some underlying belief state, and system responses are determined by a policy which maps belief states into actions.

Virtually all current spoken dialogue systems are designed to operate in a specific carefully defined domain such as restaurant information, appointment booking, product installation support, etc. However, if voice is to become a significant input modality for accessing web-based information and services, then techniques will be needed to enable spoken dialogue systems to operate within open domains.

The first part of the talk will briefly review the basic ideas of POMDP dialogue systems as currently applied to closed-domains. Unlike many other areas of machine learning, spoken dialogue systems always have a user on-hand to provide supervision. Based on this idea, the second part of the talk describes a number of techniques by which implicit user supervision can allow a spoken dialogue system to adapt on-line to extended domains.

Post has shared content

Post has attachment

Post has shared content

If you read just one article about Facebook's new Graph Search, you have to read the Wired article by +Steven Levy. He was given early access to the developers and demos of the product. Other posts and articles generally restate what was said and shown publicly by Facebook today. This one's unique. Highly recommended.

Reshare if you like it...

(Updated to be less disparaging to other articles, thanks to a comment by +Matt McGee)

Reshare if you like it...

(Updated to be less disparaging to other articles, thanks to a comment by +Matt McGee)

Post has attachment

We are hiring -- Research Staff Members and Research Engineers in the Programming Models and Tools Department at IBM TJ Watson! Seriously, this is a great time to join our group...

Post has attachment

Post has attachment

Rahul Dravid, my hero, retired yesterday. A legend of cricket and a very fine man.

Wait while more posts are being loaded