Dear editors,

Today I saw some news about the Royal Society which made me very uncomfortable. As a result, I am afraid I will withdraw my voluntary reviewing work for Royal Society. I will not after all be completing this review for Royal Society Open Science.

The uncomfortable news is about the Royal Society's refusal to uphold its code of conduct with regard to Elon Musk. It's very clear that Musk's continued association with the RS brings the society into disrepute. I can feel this disrepute - in my own changed opinion of the society.

I am not a member of the RS, although I have been proud to act as a reviewer for some of its journals. I don't feel proud today, and I don't feel I can continue donating my voluntary labour to the society so long as it fails to uphold its own values.

Please pass this message on to your colleagues. The society's honour can be restored, I hope.

Best wishes

Dan Stowell

We are pleased to announce a special session on "Signal Analysis for Biodiversity" to be held at the EUSIPCO 2024 conference: August 26-30 2024, Lyon, France. Please consider submitting a paper.

• Deadline for paper submission: Extended to March 10th 2024

Special session description

This Special Session will bring together practitioners interested in use of signal processing and machine learning methods to monitor biodiversity and the behavior and interaction of living organisms in an environment. For example, a typical contribution would be about using audio or video signal processing to detect animals in a forest or a farming site.

The biodiversity crisis continues to grow and becomes more visible every year. Although much monitoring is already conducted, there is a massive information gap due to the scale of the issue: for example there is currently ongoing discussion about whether the recently-identified “insect apocalypse” applies across all species and all parts of the world. Resolving these issues is of vital importance since insects and many other animals are, among other things, crucial to society as crop pollinators. On the positive side new information streams for biodiversity are becoming available, from audio and video recorders, satellite and drone imaging, and many other environmental sensors. Signal processing and statistical optimisation have a key role to play, since they are needed to turn these raw data streams into evidence.

Scientific development of such methods requires attention to the specific properties of the signals as well as the inferences required: for example, a single audio signal may embed evidence of multiple different species, and their interactions, as well as weather and human factors.

Organising committee

• Dan Stowell
• Aki Härmä
• Ricard Marxer

Instructions for authors

Deadlines, dates, and author instructions are listed on the main call for papers for EUSIPCO 2024. Papers should be submitted in the EUSIPCO main submission system, ensuring to select the correct special session in the drop-down menu.

As we head into summer, I want to tell you about the bumper crop of new research papers we have from my amazing students and postdocs.

I'm working with a great "decentralised" team of research students/postdocs at Tilburg University, Naturalis Biodiversity Centre, and QMUL.

The papers I'll link here are preprints -- i.e these are the versions we submitted to journals/conferences, and there may be corrections or other edits in the final accepted versions. So here you can read our freshest research news!

• Learning to detect an animal sound from five examples -- This one is a BIG achievement, the first journal-length analysis of a large and long collaboration on "few-shot learning" for animal sound detection. It is led by Ines Nolasco and Shubhr Singh, both PhD students working with me. There's a big set of collaborators behind this one (see the author list), and our success is certainly thanks to them.
• Comparison between transformers and convolutional models for fine-grained classification of insects -- Our first publication on fine-grained image classification, led by Rita Pucci as part of her postdoc work in our EU-funded MAMBO project. Accepted to present at the Camera Traps, AI, and Ecology workshop.
• Feature Embeddings from Large-Scale Acoustic Bird Classifiers Enable Few-Shot Transfer Learning -- This one is from Burooj Ghani, but the work you see here is a new paper from his previous collaboration. There'll be more along these lines very soon.
• Adaptive Representations of Sound for Automatic Insect Recognition -- Led by Marius Faiß, originally as part of his MSc thesis at the University of Leiden, and now developed into a larger project at Naturalis.

The big picture, that you can perhaps see in these paper titles, is that a lot of our focus is on developing machine learning methods for species recognition. There are already plenty of papers and apps for species recognition e.g. of bird sounds, but in order to provide species recognition as a service and a tool at the continental scale there's a lot still to be done. Firstly, going beyond common cases such as birdsong takes some care and attention - I'm happy that we're contributing something to insect sounds (and insect images too) though more work is needed to bring it to the same level of maturity. Second, there are plenty of new developments in deep learning architectures, and still plenty of open questions about the best input representation to use with these, and so there's plenty to explore.

For our group at Naturalis, it's a great start to 2023 because we recently welcomed two new research team members - two postdocs with expertise in AI species recognition: Rita Pucci and Burooj Ghani. Their work is funded by Horizon Europe under three different (but related) projects: MAMBO, GUARDEN, and TETTRIs. Within all three of those projects, our role is centred around providing AI species recognition algorithms.

Burooj Ghani is working on AI methods for recognition of European birds, bats, marine mammals and grasshoppers based on sounds. Rita Pucci is working on AI methods for recognition of European animals and plants based on images. Together with me, Vincent Kalkman and the ICT AI-team (Laurens Hogeweg and Django Brunink), we'll be refining AI methods, making the algorithms available (among other things, through the Arise online services), and working with the project partners to make these methods fully useful for ecologists and institutions all around Europe.

They will also be members of the wider Evolutionary Ecology research group here.

I decided to pick some of my favourite research papers from 2022, NOT from my group or my collaborators. Here they are:

Three good bioacoustics research papers from 2022:

1. "Acoustic indices as proxies for biodiversity: a meta-analysis" - a very nicely-done meta-analysis, for anyone wondering about "acoustic indices"!
2. "Automated annotation of birdsong with a neural network that segments spectrograms"
3. "Acoustic species distribution models (aSDMs): A framework to forecast shifts in calling behaviour under climate change" - intriguing and thought-provoking idea of applying distribution models to animal calls in themselves.

Three good AI/deep-learning research papers from 2022:

1. "Transformers learn in-context by gradient descent" - very neat: LLMs are performing unrolled gradient descent?
2. "Git Re-Basin: Merging Models modulo Permutation Symmetries" - exciting possibility of merging trained AIs.
3. "Improving biodiversity protection through artificial intelligence" - excellent novel use of reinforcement learning for biodiversity.

I'd love to hear of recent research papers you appreciated. (Not papers from your own group :) Message me on Mastodon.

It's actually astonishing how much of the super-impressive recent work in machine learning is based on vector representations of data -- simple "embeddings" in which a data item becomes a coordinate in a Euclidean space (or at least, a metric space). I remember doing something related to that back in my own PhD and spending lots of time agonishing over how bad an idea it seems to assume all datasets can be represented as a set of dots in a space!

In metric spaces such as these, there are inviolable rules, such as the triangle inequality which enforces something about the (dis)similarity between 3 items. But those fixed rules might not be true about your data. It's quite common for perceptual data to violate the triangle inequality...

In my case it was vowels and consonants plus all the strange sounds that a beatboxer can produce. The traditional "vowel space" known in phonetics can be treated as a metric space, but it has no room in it for consonants. Should vowels and consonants lie in different manifolds? They probably have different numbers of dimensions, so yes, I'd say. Can those sit in the same parent space, and should those manifolds be connected to one another? How could an algorithm learn about that complex situation?

All of this is ignored when we take the generic data-driven approach of projecting everything into some high-dimensional space and then optimising the parameters of that space. It's possible that the learnt representation embeds the data into interestingly-shaped manifolds -- but, as shown by the trivial example of the triangle inequality, there are certain relationships that simply can't be represented in a standard metric space.

I've not seen any recent work exploring these questions. But perhaps that's just because I'm too busy to find those papers. ... And I'm not claiming that our own work does anything different than this mainstream. For example, our 2021 paper "Deep perceptual embeddings for unlabelled animal sound events" uses exactly this vector embedding, driven by animals' own judgments. It performs better than previous methods, but it still makes these strong "vector space" assumptions about perception.

I'm sure there are mathematicians out there who can give a well-argued opinion. But in an era when deep learning methods seem to be able to represent entire languages, natural images and more, basically encoded into metric spaces ("embeddings"), it seems particularly apposite. Are there aspects, for example, of natural language that don't fit metric spaces, and which large language models empirically do poorly at?

Especially for MSc students starting to plan their thesis with me - here are some good examples from my past students!

Be aware that the guidelines for an MSc thesis are different depending on the programme, and they also change over the years. So always check the official requirements and the …

I'm pleased to announce that we have 2 two-year postdoctoral positions available.

We're looking for early-career researchers who would like to develop AI for sound and image recognition of nature. The positions are here in the beautiful Dutch city of Leiden, as part of projects funded by Horizon Europe. At …

I'm really excited about the developments going on around me. I'm working with great people on AI for biodiversity and sustainability, in two lovely academic departments in the Netherlands (Tilburg University and Naturalis Biodiversity Centre). You can join! That's part of what's exciting. We have job opportunities!

• Assistant professor Artificial …

If you are working with me e.g. for your MSc project, here are some starting points for reading, and for tooling up:

Recommended reading:

1. Computational Analysis of Sound Scenes and Events - a good textbook from 2018. Chapter 2 is a very good intro to many of the fundamentals in …