CS and Maths Tutorials: Bringing it Together

Question and Answer

Tue, 01 Dec 2015 16:00:00 +0000

This session will be a question and answer session about every thing you’ve been through in this first semester, including data hide meetings, Daniel Bernhardt’s talk and the general scope and direction of the course.

As material for further discussion please read through the Computer Science First Year’s Tutorial Booklet on assignments and feedback (Semester 1, week 7) and review of progress (Semester 1, week 9). Think about the different forms of assessment you receive. How do DCS and SoMaS differ in their ways of providing you feedback? How does the teaching philosophy in DCS and SoMaS differ?

Probability Modelling: Language and Physics

Tue, 17 Nov 2015 16:00:00 +0000

For this week’s meeting Daniel Bernhardt from Facebook will present on spelling at Facebook. The meeting will be followed by a “Data Hide” meeting at which Daniel will also present.

Probability Modelling: Language and Physics

Tue, 03 Nov 2015 16:00:00 +0000

In two week’s time we will have Daniel Bernhardt from Facebook, London presenting to us at the Data Hide. He will spend the tutorial session beforehand discussing the details of spell checkers at Facebook.

To prepare for the 17th this we we will have a discussion of n-grams and how to model language, and starting thinking more about the spell checking problem that Daniel will talk about. We’ll have a simple review of probability and how to relate marginals and conditional probabilities and think about how probability can be used to represent sequences. In particular you should refresh yourself on the sum rule of probability and the product rule of probability.

Think about this type of modelling and how it relates to Navier Stokes. In particular what are the abstractions in Navier Stokes? What are the abstactions in n-gram models and how they are applied to language?

Which model would you think of as “Higher Level” and which model is a “Lower Level” model.

Navier Stokes is often thought of as a “Physical model” can n-gram models be thought of in this way? If so how why? If not why not?

Mathematics and Computer Science: The Interface

Tue, 20 Oct 2015 16:00:00 +0000

Take a look at Navier Stokes equations and find out what they are used for. What are the pure mathematics aspects of the equations and what are the applied mathematics aspects? Where does computation come in?

Think about the relationship between Computer Science and Maths. In absolute terms we many of the challenges of Computer Science are how to implement maths on a Computer (e.g. the Turing test, the way a processor works). For Mathematics, particularly pure mathematics, there is a lot of interest in separating the maths from the implementation. Mathematics can stand alone, outside implementation, emerging purely through abstract thought.

Have a look at the Wikipedia page for Principia Mathematica and Godel’s incompleteness theorem as well as the Barber Paradox. How does this sort of pure mathematical thinking effect Computer Science?

Tutorial will be followed by a “Data Hide”. If you are staying, you need to sign up for it here.

Non-Academic Task

Before this tutorial you also need to read through the Computer Science First Year’s Tutorial Booklet on unfair means and plagiarism. You need to fill in the form provided and prepare to submit your answers to MOLE2. (Semester 1, week 4).

Computer Science and Math Interface: Background

Tue, 06 Oct 2015 00:00:00 +0000

Many computer science departments were spun out of mathematics departments in the 1980s (including Sheffield’s!), but there is a historical connection between the two departments that has meant that dual degrees are relatively common.

In Sheffield, the original main link was between areas such as formal methods and pure mathematics. Verification of code for example. This relates to Godel’s incompleteness theorem and Alan Turing’s “Turing Machine”. The focus was on the computability of numbers.

Recently the nature of the interface has changed a lot. While areas such as theorem proving and verification of code are still very important, a major modern challenge is the overwhelming amount of data that we are generating. This data is generated as a direct consequence of the success of computers. It is also the responsibility of computer scientists to deal with it.

As background reading for this tutorial have a look at the links below:

Article in The Guardian’s media network on Digital Oligarchies in March 2015.
Article in The Guardian’s Media and Tech Network on preventing AI becoming creepy.
Article in The Guardian’s Media and Tech Network on How Africa can benefit from the data science revolution

Consider each of the articles and what the implications are for mathematicians and computer scientists. What does it mean for the future of this interface? Which issues fall within the domain of work in data and which fall outside?

Welcome to Sheffield!

Tue, 29 Sep 2015 00:00:00 +0000

This first session will be held at “The Hide” in Sheffield, near the Fire and Police Museum. There will be street food and beer and an opportunity to meet others on your course as well as research students who work at the interface of computer science and mathematics. You will also get a chance to meet some of the staff from the two departments.

You need to register for the event here:

Eventbrite Registration

If you want to know about future events like this please ‘like’ our Facebook page or follow us on Twitter

Daniel Bernhardt Visit

Mon, 24 Nov 2014 00:00:00 +0000

Daniel is visiting from Facebook London where he’s been working for just over a year. Before that he worked for 4 years at Microsoft Search, and previous to that he did his PhD in applied machine learning for emotional gesture recognition and his Undergraduate in Computer Science (both at the University of Cambridge).

Daniel Bernhardt visit. We will learn about how Facebook deal with the challenge of spell correction for the vast range of names that exist. Daniel will give us a talk followed by a Q&A about anything you care to have answered!

Daniel spoke about the spell checking problem, two different types of spelling errors cognitive errors and typo errors, and how these are passed into the search engine. Search engine queries take several hundred milliseconds. Your typed query is passed to the search engine after spell checking.

“A System for searching the social Graph” (VLDB 2013)

First Approach

We talked about the edit distance as a way of comparing correct names (in a dictionary) to incorrect names. This is a dynamic programming approach for computing the distance of the query from a ‘correct’ query.

Through a hash table you can compute look ups quickly.

In practice the dictionary could be really massive! 1 billion user nodes, 240 billion photo nodes, 1.2 trillion posts, 1 trillion edges.

To build a dictionary of unique names: there are 73 million unique first name and 67 million unique surnames. 560,800,810 unique en-us names! The dictionary would be very large and nearly 40% of US full names are unique!

Context also needs to be included, you are often spelling friends names, you don’t want the average of all names!

Edit distances also don’t encode properly mistakes that are more likely (for example due to a keyboard layout).

Actual Approach

Language model takes any given sequence and gives a probability associated. n-gram models can be used for modelling language.

We talked about error models and how they can be derived from query rewrites. This is when a user types a search and then doesn’t seem to select any of the returned results. They then change the query and that new query gives the right result. This gives a lot of valuable information about what the original query should have been.

We talked about challenges including doing different languages, and making general solutions. In particular we talked about the challenge of domain adaptation and how a spell checker for one particular domain may not be appropriate for another domain. We talked about the importance of data and feedback.

Daniel talked about how widely machine learning is used across the site, and how, often, it’s not the learning algorithm itself that’s a problem, but evaluating the quality of the model or extracting the data for learning.

Charles asked about when the system is finished, Daniel said broadly speaking that they are always looking for continuous improvement. Whilst the target is to focus on impact, they don’t consider a system done, but if the system is working well, then they may reprioritise on something else.

** Also Non Academic Task 4 **

Non-Academic 4

Read through theComputer Science First Year’s Tutorial Booklet on exam skills. Read old exam papers and think about what the question’s aims are. How to DCS and SoMaS differ in their exam question style? How are they similar?

Preparation for Daniel Bernhardt Visit

Mon, 17 Nov 2014 00:00:00 +0000

Preparation for Daniel’s visit! Let’s go through those rules of probability once again, and have quick refresher on n-gram models. We should also try and look at dynamic programming algorithms for trying to find the most probable path through a sequence.

John Baez Topics 2

Mon, 10 Nov 2014 00:00:00 +0000

Presentation in pairs or individuals of topics from John Baez’s Google+ feed Week 2.

Group discussion, lead by Matheus and Islam, about Science, Models and Machine Learning

http://johncarlosbaez.wordpress.com/2014/09/03/science-models-and-machine-learning/

Some of the topics mentioned during the discussion were:

Software available:

http://scikit-learn.org/stable/

https://github.com/SheffieldML/GPy

http://www.automaticstatistician.com

Decision making and rational behaviour: http://www.ted.com/talks/dan_ariely_asks_are_we_in_control_of_our_own_decisions#t-890383

The sampling theory (statistical sampling):

http://www.youtube.com/watch?v=7nr9rQpm2A4

Also: Non Academic Task 3 (see below)

Non-Academic 3

Read through the Computer Science First Year’s Tutorial Booklet on assignments and feedback. Think about the different forms of assessment you receive. How do DCS and SoMaS differ in their ways of providing you feedback? (Semester 1, week 7)

John Baez Topics 1

Mon, 03 Nov 2014 00:00:00 +0000

John Baez topics Week 1

Thomas presented on the Quantum Thriller: https://www.youtube.com/watch?v=I7oZAo8hJnU which led us on to Quantum cryptography. Interesting reading is:

http://en.wikipedia.org/wiki/A_Universe_from_Nothing

http://simonsingh.net/books/the-code-book/

Presentation in pairs or individuals of topics from John Baez’s Google+ feed Week 1.