MLSP 2016 - i.e. the IEEE International Workshop on Machine Learning for Signal Processing - was a great, well-organised workshop, held last week on Italy's Amalfi coast. (Yes, lovely place to go for work - if only I'd had some spare time for sightseeing on the side! Anyway.)

Here are a few of the papers that caught my interest:

• Approximate State-Space Gaussian Processes Via Spectral Transformation by Toni Karvonen and Simo Särkkä. This is an important contribution to the current work on Gaussian processes and in particular on running efficient Gaussian process inference. It builds on other work from the Särkkä lab converting Gaussian processes to state-space models, which often involves a (mild) approximation. This paper introduces some new methods in that vein, with proofs, and in fact the paper includes various ways to approximate a GP. A veritable mathematical toolkit. It seems the Taylor expansion (the most immediately comprehensible IMHO) is not the best.

Actually, there was substantial work involving Gaussian processes at MLSP. Is it a growth area? Well, if the use of GPs can be made more scalable (as in the above paper) then yes, it certainly should be. They are a very flexible and general tool, and nicely Bayesian too. Richard Turner's keynote about Gaussian processes was a beautiful introduction - he manages to make GPs extremely understandable. If you get a chance to see him speak on them then do.

• Localizing Users And Items From Paired Comparisons by O'Shaughnessy and Davenport. This is a nicely conceived addition to the literature on recommendation algorithms, and with good demonstrations of how the approach is robust to issues such as incoherent paired comparisons.
• "Data Privacy Protection By Kernel Subspace Projection And Generalized Eigenvalue Decomposition" by Diamantaras and Kung. Privacy-preserving computing is an important area for current research. It's made obvious when we see how much a large company like Facebook or Tesco can infer about its users. Here, the authors treat privacy as a classification task - i.e. the data to be kept private is some kind of discrete label - and they apply an LDA-like method: maximise the scatter between the target classes for the "allowed" task, while minimising the scatter between the private classes. (I raised an issue with their "Privacy Index", noting that the desired accuracy for the private task was not in fact zero but ignorance. I'd presume that a metric based on mutual information would be a nice alternative.)
• "Scale and shift invariant time/frequency representation using auditory statistics: application to rhythm description" by Marchand and Peeters. They use the "Scale Transform", a class of Mellin transform. Equivalent to exponentially time-warping a signal then weighting by an exponential window. Since it's not shift-invariant you don't want to apply it directly to audio, but to e.g. autocorrelation. From there, they argue you get a good featureset for characterising musical rhythm.
• Score-Matching Estimators For Continuous-Time Point-Process Regression Models by Sahani, Bohner and Meyer - good to see this. I've been using point process models to analyse bird communication and so I'm interested in efficient ways to do such analysis, which commonly seem to come from the computational neuroscience literature at the moment. Notable that this approach doesn't require any time discretisation, so could be useful. The functions analysed need to be differentiable, so to work with impulsive time series they actually convolve/correlate them with basis functions; feels like a minor hack but there you go.

Also, I was very pleased that Pablo A Alvarado Duran presented his work on Gaussian processes for music audio modelling - his first publication as part of his PhD with me!

InterSpeech 2016 was a very interesting conference. I have been to InterSpeech before, yes - but I'm not a speech-recognition person so it's not my "home" conference. I was there specifically for the birds/animals special session (organised by Naomi Harte and Peter Jancovic), but it was also a great opportunity to check in on what's going on in speech technology research.

Here's a note of some of the interesting papers I saw. I'll start with some of the birds/animals papers:

• Localizing Bird Songs Using an Open Source Robot Audition System with a Microphone Array by Suzuki et al. Some really interesting work from a Japanese group I hadn't met before - they use a simple USB multi-microphone device together with source-localisation software, and then use that to recover a multi-species transcript of activity. Potentially very useful.
• Recognition of Multiple Bird Species Based on Penalised Maximum Likelihood and HMM-Based Modelling of Individual Vocalisation Elements by Jancovic and Kokuer - I like their modulated-sinusoid detector - here it's incorporated into a system for multi-species recognition in polyphonic audio. Good to see this being developed. It'd be great if others could make use of this technique (I don't think source code is available...).
• Bird Song Synthesis Based on Hidden Markov Models by Bonada et al - this is work Jordi Bonada developed while on a research visit to our lab (C4DM), with a colleague of mine Rob Lachlan. Good probabilistic generation of birdsong needs to be developed, you see, because at the moment a lot of bird behaviour experiments use a small number of WAV files as play back, and the unnatural repetitiveness that results from this makes me uncomfortable! Hence we need work like this on flexible bird sound synthesis.
• Cost Effective Acoustic Monitoring of Bird Species by Ciira wa Maina. A neat paper about a prototype 'full solution', little Raspberry Pi sound recorders with a straightforward single-species sound detector.
• Feature Learning and Automatic Segmentation for Dolphin Communication Analysis by Kohlsdorf et al. I don't know what audio tracking methods dolphin researchers currently favour, but this is an interesting highly-scalable method based on feature-learning in spectrogram data.
• Robust Detection of Multiple Bioacoustic Events with Repetitive Structures by Frank Kurth. To be honest I don't grok this yet - it's a kind of enhanced autocorrelation, but I'm not sure what leads to it being said to be robust.
• A Real-Time Parametric General-Purpose Mammalian Vocal Synthesiser by Roger K Moore: a PureData patch for general-purpose mammal sound synthesis. Fun, and maybe useful for someone out there working on mammal sounds.

That's not all the bird/animal papers, sorry, just the ones I have comments about.

And now a sampling of the other papers that caught my interest:

• Retrieval of Textual Song Lyrics from Sung Inputs by Anna Kruspe. Nice to see work on aligning song lyrics against audio recordings - it's something that the field of MIR is in need of. The example application here is if you sing a few words, can a system retrieve the right song audio from a karaoke database?
• The auditory representation of speech sounds in human motor cortex - this journal article has some of the amazing findings presented by Eddie Chang in his fantastic keynote speech, discovering the way phonemes are organised in our brains, both for production and perception.
• Today's Most Frequently Used F0 Estimation Methods, and Their Accuracy in Estimating Male and Female Pitch in Clean Speech by Sofia Strömbergsson. This survey is a great service for the community. The general conclusion is that Praat's pitch detection is really among the best off-the-shelf recommendations (for speech analysis, here - the evaluation hasn't been done for non-human sounds!).
• Supervised Learning of Acoustic Models in a Zero Resource Setting to Improve DPGMM Clustering by Heck et al - "zero-resource" speech analysis is interesting to me because it could be relevant for bird sounds. "Zero resource" means analysing languages for which we have no corpora or other helpful data available - all we have is audio recordings. (Sounds familiar?) In this paper the authors used some adaptation techniques to improve a method introduced last year based on unsupervised nonparametric clustering.
• Speech reductions cause a de-weighting of secondary acoustic cues by Varnet et al: a study of some niche aspects of human listening. Through tests of people listening to speech examples in noise they found that people's use of secondary cues - i.e. clues that help to distinguish one phoneme from another, which clues are embedded elsewhere in the word than the phoneme itself - changes according to the nature of the stimulus. Yet more evidence that perception is an active, context-sensitive process etc.

One thing you won't realise from my own notes is that InterSpeech was heavily dominated by deep learning. Convolutional neural nets (ConvNets), recurrent neural nets (RNNs), they were everywhere. Lots of discussion about connectionist temporal classification (CTC) - some people say it's the best, some people say it requires too much data to train properly, some people say they have other tricks so they can get away without it. It will be interesting to see how that discussion evolves. However, many of the other deep-learning based papers were much of a muchness: lots of people use a ConvNet or an RNN and, as we all know, in many cases they can get good results. They apply these to many tasks in speech technology. However, in many cases there was application without a whole lot of insight. That's the way the state of the art is at the moment, I guess. Therefore, many of my most interesting moments at InterSpeech were deep-learning-less :) see above.

(Also, I had to miss the final day, to catch my return flight. Wish I'd been able to go to the VAD and Audio Events session, for example.)

Another aspect of speech technology is the emphasis on public data challenges - there are lots of them! Speech recognition, speaker recognition, language recognition, distant speech recognition, zero-resource speech recognition, de-reverberation... Some of these have been running for years and the dedication of the organisers is worth praising. Useful to check in on how these things are organised, as we develop similar initiatives in general and natural sound scene analysis.

I've been trying to think of times when I've heard a proper Lancashire accent in music recently. (Pop, rock, rap, whatever.) It's not easy! Suggestions please? We want someone who's done for the Lancs accent what the Arctic Monkeys have done for Yorkshire...

A couple of nice examples are Shaun Ryder and Guy Garvey. But then it's a bit of a mixup because the county of Lancashire used to include Manchester, but the modern county doesn't include Shaun or Guy's hometowns, so, well, they've got the right voices but they might get disqualified on a technicality :(

Looking back to earlier eras... the same thing happens with The Beatles. Liverpool used to be part of Lancashire but not any more. Anyway The Beatles are confusing because they sometimes used quite genuine regional accents, sometimes transatlantic rock'n'roll accents - quite explicitly hopping about. Much more dependable, and much older, George Formby's voice is a proper representative sample.

Here's one modern example, though not famous: The Eccentronic Research Council - the woman in that track is Maxine Peake (another Manc/Lanc stowaway).

Oh and I don't really want to mention The Lancashire Hotpots because they're too daft.

Chumbawamba were from Burnley, hoorah. But their singing doesn't have much of the accent as far as I can tell.

FFS come on! This list needs some vim. Let's chuffing represent! Answers on a postcard.

---

Edit: top tip from Lucy: The Lovely Eggs - that's what I'm talking about!

The LRB has an excellent article by Philippe Sands about the Chilcot Report and the Iraq War.

The UK had a key role in the Iraq War, and even before it happened there were millions of us on the streets marching against it: we said in advance that it was unjustified and would escalate terrorism in the region. (There's a video going round at the moment of Jeremy Corbyn back in the day, saying exactly that.) Now, looking back from a 2016 in which we have Isis/Da'esh and waves of refugees, there's no pleasure in the confirmation that we were right. The consequences reverberated not just through the region, but through to the EU and the UK too. Millions of us ignored, and so many killed (not least, directly killed in the war), because Tony Blair had pledged to Bush: "I'm with you, whatever".

Some quotes from the article:

"[The inquiry said] 'we consider that the UK was, in fact, undermining the Security Council's authority.'"

The inquiry received 37 legal responses regarding the war's legality, "reflecting the views of 57 expert individuals and six organisations. Just one of them supported the claim that the war was lawful."

"On 31 January [2003], Blair met Bush and offered a commitment that contradicted the legal advice given to him by Goldsmith the previous day."

"[Goldsmith's] formal advice - the 7 March document permeated with an understanding of the uncertainty and risk involved in going to war - was deliberately withheld from cabinet."

Our government (and related organisations such as the UN Security Council) are built with checks and balances, so that things such as ill-advised wars on the basis of misconstrued information should be less likely.

The article is well worth a read.

So I was on holiday in the mountains bordering Germany and the Czech Republic. Walking around, it was actually a bit odd to be simply walking between Germany and the Czech Republic simply by pootling down a street - i.e. making good use of the Schengen free-movement area that's been one focal point of the migration and refugee news this year. So that was on my mind a bit. (And no I didn't spot any migrants!)

Then in the woods, I saw something unexpected. - A grid of little concrete pyramids, each about a metre high. Just sitting there, not part of a building, or an art installation, or anything, as far as I could tell. Looking fairly old and unkempt. An area of about 15 by 20 metres, weird.

I didn't take a picture, so here's an artist's impression:

Weird huh?

Eventually I saw there was a little notice by the side, explaining that these were a leftover part of the tank-trap strip that defended the Iron Curtain a few years ago. Thick forest provided a lot of the border defence against tanks - but in the channels between forests, these tank traps would be built so that tanks would scupper themselves on the pyramids. Then the guards watching from the nearby towers could do the rest.

Here's a picture I found online, of some similar leftover pyramids in the Czech Republic border area:

There were a couple of photos included, of tests where they drove tanks on top of the pyramids. The thought of it, in this peaceful forest. An "enemy" (or any) tank coming hurling down the way, aiming for the border, crashing over the landscape, through fences etc. This is what borders in Europe used to be like. Within my lifetime. These are the eventualities that planners had in mind.

It puts in perspective how we currently talk about borders, with the EU and Schengen and Brexit. I presume that Germans who remember their country before unification already have this perspective...? But as a mainland Brit, it's not like that.

(I know it's also how some borders are right now. For example between Israel and Palestine. Maybe we should all go and experience the physicality of those things first hand.)

Politically, the world is quite unstable right now, and it's really unclear how the multiple crises of 2016 will play out. Whatever happens, the future will not be exactly the same as the past. But when you bump up against our very recent history (the Iron Curtain) in physical form it really brings it home, for someone who never experienced a war - never really even felt the brunt of the troubles in Ireland, though I remember the Manchester Arndale bombing - that the peace and stability that happens (?) to coincide with the EU era is not something to be taken cheaply.

Further reading: I found this nice travelogue from someone who travelled along exploring the Iron Curtain etc, while I was looking for the photo.

Just back from a trip to the Bavarian Forest - hiking, eating, wandering around. Here are the highlights of what I ate!

Traditional:

• Pork Schnitzel Wiener art - classic around here, and in Eisensteiner Hof in Bayerisch Eisenstein I had a really excellent one. The meat was tender and flavourful, and the coating delightfully smooth with some lemon juice sprinkled over.
• At the Schwellhaus'l guesthaus a lovely deer goulash, with red cabbage and spaetzle (soft dumplingy noodles).
• Trout is a main fish around there, and at Haus zur Wildnis in Ludwigsthal they did a delicious smoked trout salad. For afters a nice traditional pancake, with a date preserve (I think that's what it was!).
• I also had very nice little pancakes at the Eisensteiner Hof - that time I did manage to write down what they were: "liwanzen" apparently, Viennese. This time served with a nice and tart berry compote.
• At Poeschl in Bayerisch Eisenstein I had echt Bayerisch brotzeit ("bread time"), having two boiled white sausages, with a pretzel, plus apfelschorle and some baerwurst to wash it down.

Not particularly traditional:

• Utaty in Bayerisch Eisenstein is a vegan indian restaurant, open only on Friday and Saturday evenings, and with a set menu written in a notebook in a very dear-diary style:

• Also on a veggie tip - I've been searching for veggieburgers worth making/eating. Most are... a bit meh. At Adam's Brauerei in Bodenmais I had one of the best veggieburgers I've had. So what was it? In fact it was just a straightforward veg pattie, but you couldn't tell since it was overwhelmed by: two slices fried courgette, a fried egg, tomato, cheese, and some micro salad leaves, all in a beautiful "alkaline" burger bun ("alkaline" according to automatic translation of "Laugen-Burger-Broetchen"...) with guacamole on the side. So that's how.
• Best ice cream was at Cafe Charlotte in Zelezna Ruda. I had a berry one and another one whose name I couldn't translate.

But my special prize goes to this which is... not traditional but has some "tradition" to it:

• "Country" pizza on the German/Czech border (in the village of Zelezna Ruda, and almost the closest eatery to the border there - a place called Cukrárna Sněhurka), featuring sauerkraut, sausage, ham, and soured cream. Whether a cynical invention or not, it's a great combination, and a nicely done pizza base, I was very nicely surprised.

Inspired by Nigel Slater's recipe I made a great and simple vegetarian stew (vegan, in fact), using my black bean chorizo to help add depth of flavour. (If you haven't got any of that, you could probably do something similar just with a blob of black bean sauce, even though the flavour is different?)

Serves 1 (fairly big portion), takes 25 mins.

• 125g ready-to-eat puy lentils
• 1 onion
• 1 tbsp black bean chorizo
• 1/4 tsp ground paprika
• 1/4 tsp ground nutmeg
• 1/4 tsp ground cinnamon
• Optional: a few mangetouts
• A few leaves of fresh parsley
• Optional: a squeeze of fresh lemon juice

In a deep pan which has a lid, heat up about 1 tbsp vegetable oil, while you chop the onion. You want to chop about three-quarters of the onion into whatever size pieces, and the remaining one-quarter of the onion slice it into nice rings, about half a centimetre thick.

The misc pieces of onion, put them in the pan and give them a good fry to get them softened. Add the three spices and stir around. Then add the chorizo - not too much, it's mainly for flavour. Let this cook for two minutes or so.

Then add the lentils and stir, then add enough boiling water to only-just-cover. Put a lid on, turn the heat right down, and let this bubble for 15 minutes.

In the final five minutes, heat about 2 tbsp of vegetable oil in a frying pan. Make sure the onion rings are separated into circles, and put them in the pan to fry briskly for 5 minutes, turning halfway. While these are getting a little crispy, chop the mangetouts roughly into maybe 3 pieces each and chuck them into the stew, and also add the bits of parsley and the lemon juice.

When the onion rings are ready, simply put the stew in a bowl and sprinkle the onion rings on top.

Just before the Brexit referendum I was wondering how Brexit would affect the kind of people coming to work with us. That's a long-term effect and very hard to measure. But really, like most of the country I hadn't really thought deeply about the direct practical consequences of an Exit vote, in this case the consequences for research that would show themselves within the first month.

The effect is on EU funding in particular. Since the UK hasn't actually left the EU, you might think that things carry on "as normal" until that point - existing projects continue, and you can even apply for new projects. (In fact, that's essentially the official guidance so far.) The problem is that collaborative EU grants are the lifeblood of a lot of research, and they're also very competitive. I know of at least one colleague who's been taken off a grant proposal (which is being organised by someone in another EU country), because a UK partner now means a risk factor that could easily cause a reviewer or a programme administrator to mark the proposal down.

Similarly, at least two colleagues who have been leading on EU grant proposals, they're now in a difficult situation. After having put a lot of work into preparing a proposal, do they submit and risk getting marked down as a risk factor? Do they rewrite the proposal with Brexit backup strategies? Do they stop and wait to see what happens?

(I'm not writing this down to change anyone's mind about Brexit, by the way. Just documenting.)

Less concrete, but in my own first-hand experience: we were intending to invite a good researcher to come and work with us under the Marie Sklodowska Curie scheme (which funds researchers to spend time in another country), but I'm not sure how we can do that now. The funding is still there, but apart from the "risk factor" effect mentioned above, the potential researchers would obviously need to know how it affects their right to work in the UK (will they need a visa?) and what career options might follow on afterwards; and there's pretty much nothing we can say in answer to such questions.

This Guardian article, "UK scientists dropped from EU projects because of post-Brexit funding fears" puts the same phenomena in a wider context. This quote, for example:

Joe Gorman, a senior scientist at Sintef, Norway's leading research institute, said he believed UK industry and universities would see "a fairly drastic and immediate reduction in the number of invitations to join consortiums. [...] I strongly suspect that UK politicians simply don’t understand this, and think it is 'business as usual', at least until negotiations have been completed. They are wrong, the problems start right now."