Forgot your password?
typodupeerror

Comment: Re:Perl (Score 1) 536

I've never used Perl, but the parent makes a very good point. Why move unless you have a problem?

This topic is already full of fanboys touting thir favourite tool, but one man's meat is another's poison. You already know the options (Perl, Java, PHP, Python, C#, etc) - if you don't, you're not ready to make a decision yet - so what you really need is a decision-making process.

List out all of the things that might matter to you: team skills, staff availability, platform dependence, maturity of platform, speed of development, ease of maintenace, cost, execution speed, availability of hosting, etc. You should know what those things are; if you don't you're not yet ready to make a decision. Give each attribute a weighting from 1-5 that reflects your business priorities.

Now score each language against each attribute. Sometimes you'll have to guess, but that's not going to be too much of a problem. 1-5 is a good scale; anything else will be spurious accuracy.

Now total the scores and weight them. Keep the top 3 options and look at them more closely. Was your scoring accurate? Do you trust the result? This is where your professional judgement takes over: a scoring model can only get you so far.

Do this, and you'll find out what _you_ really need, and not what some random guy on /. thinks is good for you.

Comment: true, but not really because of R itself (Score 3, Insightful) 185

by jonnyj (#47088985) Attached to: R Throwdown Challenge

Completely right.

We use R extensively in work. Programmers talk about R's libraries, but that's not the real reason we use it. The killer blow is that the _documentation_ is written by statisticians. That means that it's reliable, easy to understand, and honestly tells you the pitfalls of the techniques you're using.

We're financial guys who are doing stuff in consumer finance that has rarely, if ever, been done in our field. The statistics aren't particularly advanced, but it's impossible to hire someone who understands the industry and knows the statistics already. Statistics text books tend to either be so basic that you already know what they say, or so advanced that you need a PhD to understand them. On the other hand, much of the R documentation is beautifully simple to read, and comes with brilliant worked examples - albeit from fields that are very different from our own. Whenever we're researching potential new statistical approaches, we find blogs stuffed full of examples written in R.

In short, the R ecosystem makes you a better statistician. Julia and Python can't offer that.

Comment: iPad with GoodReader (Score 1) 134

by jonnyj (#46255063) Attached to: Ask Slashdot: E-ink Reader For Academic Papers?

In the business world, I and many others use an iPad and GoodReader for annotating board papers. To be honest, it's the only thing that I use an iPad for, as I prefer a proper PC, a smartphone or a smaller tablet for anything else.

GoodReader allows you to annotate pdfs with a wide range of tools - I usually scribble free form text with my finger - and you can read the annotations with any pdf reader. The large format of the full size iPad simplifies finger writing, and the large retina screen means that I can read dense data tables without needing to zoom in.

Despite Apple's dumbed-down iOS, GoodReader allows you to organise documents in a hierarchical folder structure, and you can synchronise your documents with a wide range of server types and cloud storage systems.

It's not the cheapest solution around, but it's by far the best that I've ever encountered amongst my business associates.

Comment: Re: patenting statistical hypothesis? (Score 3, Insightful) 129

by jonnyj (#45628051) Attached to: Facebook Patents Inferring Income of Users

This is patently absurd. In the UK, Equifax, Experian and Call Credit already sell income predictions based on statistical modelling of credit bureau information. How is switching the underlying data set in any way a unique or clever thing to do?

This is nothing more than a fancy regression algorithm.

Comment: Re:EASY (Score 5, Informative) 310

I agree, but I wouldn't be underhand and I certainly wouldn't use read receipts. That looks horribly like the very worst kind of arse covering.

You shouldn't go over your boss's head. Juggling a large number of conflicting priorities is what managers are paid to do, and you won't do yourself or anyone else any favours by undermining your boss's judgement in that way. But you should also consider the risk that she consciously has her own best interests at heart rather than the business's interests. She might have the view that, in the event of a security debacle, she will pretend that the team messed up and failed to follow instructions, and simply ride out the storm. In the meantime, she looks efficient and appears to gets jobs done quickly with a minimum of fuss.

Instead, you should sit down with her and clearly express your concerns. You should then follow up your meeting with a very clear email that summarises the conversation. You need to start with an assertive but non-hostile comment that leaves no-one in any doubt what has happened - something like this, "As we discussed earlier, these are the security issues where I believe that we are falling short of regulatory expectations..." Print out that email and take it home with you.

At that point, your boss has three options. 1. She can fix things. 2. She can escalate up the food chain, so that someone bigger than her can decide whether poor security is really in the company's best interests. 3. At huge personal risk, she can quietly ignore you.

Middle managers tend to have pretty strong survival instincts, so option 3 is very unlikely to to fly. Option 2 is pretty likely, and her manager might well say that security is too expensive/awkward/boring/inconvenient. If that happens, you're probably better off working some place else where you can be proud to turn up in the morning.

Comment: Re:Purchased 4 so far (Score 4, Interesting) 246

by jonnyj (#45451315) Attached to: Raspberry Pi Hits the 2 Million Mark

One Raspberry Pi is a toy. The other runs my home network and presently stands at an impressive 117 days uptime with even more impressive power consumption.

My lowly workhorse Pi with its ARM 6 processor performs admirably as a:

- DNS server
- DHCP server
- Authentication server (Kerberos, OpenLDAP server and phpLDAPadmin) and publication service for network assets (OpenLDAP again)
- Mail server (Dovecot, Postfix, Squirrelmail, Spamassassin, ClamAV, Amavis)
- HTML image gallery
- Home wiki (MediaWiki)

Performance is no issue with any of this. MediaWiki is the slowest, but most pages load in 1-2 seconds. We're a busy, high-tech household so it serves up to seven laptops, five destops and nine mobile devices, many of which dual boot. Device management was a nightmare before the Pi saved the day.

Comment: Re:Afraid not (Score 5, Interesting) 90

The whole industry is in sharp decline and everyone knows it, especially those within.

True. But Linux Format has been bucking the trend in recent years. Its circulation has been rising steadily and, at 21,784 print copies per issue in 2012, it has a similar circulation to the venerable New Statesman (24,910). It trounces many other very familiar specialist mags such as Mac Format (6,842), PC format (6,249) and What Mountain Bike (13,870). It's not even too far behind the 100-year old Autocar (40,168).

All figures from ABC.

Comment: Re: More details? (Score 2) 465

by jonnyj (#45157281) Attached to: Ask Slashdot: Best Language To Learn For Scientific Computing?

R is by far the best solution that I've found for statistical analysis and data mining. It's ugly, inconsistent, quirky and old fashioned but it's absolutely brilliant.

The whole syntax of R is based around processing data sets without ever needing to worry about loops. Read up on data tables - not data frames - in R and you'll learn how to filter data, aggregate it, add columns, perform a regression and beautifully plot the results all in one line of code. The Zoo package will sort out your time series analysis and longitudinal analysis. With R, you can calculate the statistical significance of you hypotheses and apply the model you've developed to your hold-out sample using built-in functions. And the concept of workspacecd means that you don't need to think of funky ways to store your interim results.

Using knitr, R will produce publication quality documents and presentations. ggplot will give you the best data visualisation tools in the business.

R is the tool that has been purpose-built for the task in front of you. Anything else might be easier to learn or more widely supported - but it won't be as effective.

Comment: Re:We'll never have a sane debate about nuclear po (Score 2) 380

by jonnyj (#44901707) Attached to: Its Nuclear Plant Closed, Maine Town Is Full of Regret

I'm a strong supporter of nuclear power, but I believe that the 'stupid and irrational' people actually bring insights into important issues that are often overlooked by technical folk. And this article raises thought-provoking issues that I've never heard acknowledged in the media by any nuclear expert.

Any conceivable nuclear safety regime requires plant employees to act with honesty, integrity and procedural rigour. But what happens to honesty and integrity when the future economic prosperity of your family, friends and community depend on the answer? You will be under huge internal, personal pressure to downplay risks, underestimate costs, cut corners to save money, cover up poor practice, lie to inspectors and rebut any conceivable negative news item.

Technologists are human. No matter how rational they appear, the answers they provide us with are always subject to considerable personal and emotional bias and must be regarded with an appropriate level of scepticism.

Comment: Re:And no one was surprised... (Score 1) 474

by jonnyj (#43798701) Attached to: The Canadian Government's War On Science

The Left has its blind spots, too. GMOs, perhaps? Nuclear power? Michael Shermer, in his recent book The Liberals' War on Science, says, "Surveys show that moderate liberals and conservatives embrace science roughly equally." I think he's probably right: many people prefer to use science to rationalise rather then overturn their prejudices.

Comment: Re:Too Expensive (Score 2) 181

by jonnyj (#43504641) Attached to: Ars Reviewer is Happily Bored With Dell's Linux Ultrabook

Actually 50% more than the new MacBook Pro I bought last summer.

The nearest equivalent Apple laptop is the 13" Macbook Air (disclaimer: I have one and it's very good). In the UK, the two machines are almost exactly same price and are effectively dimensionally identical too. But the Air has less RAM (4GB vs 8GB), a slower processor (i5 vs i7) and a lower resolution screen (1440x900 vs 1920x1080).

I bought my Air to run Linux; I like OS X, but I much prefer Ubuntu. If I were buying today, I'd take the XPS over the Air. Both machines seem good but, for my use case, the XPS has the edge: better innards, better screen and manufacturer support for my OS of choice.

Comment: Re:Would I buy one? (Score 2, Informative) 290

by jonnyj (#43345955) Attached to: Falling Windows RT Tablet Prices Signify Slow Adoption

No, I wouldn't buy one either. But the inconsistency of the technical press is quite entertaining.

Apple strips most of the functionality out of OS X, erects a walled garden around the system, dumps it onto an ARM-based tablet and, voila, a cool, hip, trendy iPad that the critics adore.

Microsoft strips a small part of the functionality out of Windows, erects a walled garden around the system, dumps it onto an ARM-based tablet and, voila, a vile, loathed RT device that the critics lambast for being dumbed down and failing to run Excel macros.

I don't want either device, but it's clear which one has been dumbed down the most. Microsoft needs a new PR department.

Comment: Re:It Believes (Score 4, Interesting) 222

Credit risk profiling is part of my job and these models do indeed wok. Unfortunately, they need large sample sizes to be effective. Unless the UKBA has intercepted more than 1,000 terrorists about to jump on a plane, I'd be very sceptical indeed.

Another big concern is that these models all assume that the future is the same as the past. Feeding the model data on Islamic terrorists isn't likely to help you detect extreme right nationalist groups, for example. As conflict moves around the world, there's a risk that the model will find last year's terrorist-turned-nobel-peace-prize-winner and completely ignore the perpetrator of next year's atrocity.

Testing can show the presense of bugs, but not their absence. -- Dijkstra

Working...