This PhD project concerns  “Identifying unknown species and unknown sounds using machine learning”. Standard machine-learning classifiers recognise a fixed list of categories - but new categories arise for many reasons. Undiscovered and rare species, previously-unheard vocal behaviours, anomalies, even newly-arising species. Large acoustic datasets exist - for birds, insects, anurans and more -  but in many cases the expert analysis is far from complete. Automatic recognition methods must be developed that can account for new and unknown categories arising.

In this project, you will:  

• Survey, investigate and implement machine learning methods that can coherently accommodate novel/unknown fine-grained categories. 
• Investigate how to reduce biases in sensitivity of classification and novel-class detection. 
• Investigate differing issues of desired and undesired novel classes, e.g. new species versus traffic noise. 
• Analyse large multi-year acoustic datasets.
• Implement and validate an “open-set” sound identification workflow that is compatible with public machine learning APIs.
• Implement user-friendly tools to automate and enhance wildlife monitoring processes, demonstrating a live example of an AI pipeline that wildlife organisations can adopt. 

While the main focus of this project is on AI algorithm development, you will collaborate with other researchers on various other aspects of Bioacoustic AI such as wildlife ecology. In addition to conducting your own research, you will attend network-wide training workshops and seasonal schools.    

Essential knowledge, skills and experience:

• An MSc (completed, or to be awarded before the starting date) in a topic relevant for the PhD such as machine learning, signal processing, quantitative biology, bioinformatics, bioacoustics, ecology, zoology, animal behaviour or similar;
• Evidence of the ability to conduct high-quality research and write scientific text;
• Good written and oral communication in English;
• Good planning and time management skills;
• The ability to work independently as well as collaboratively within the research group;
• Proficiency in programming e.g. Python, R, C/C++, PyTorch, Tensorflow;   
• Willingness to learn skills from multiple disciplines (e.g. computer science, ecology)                                                          

Desirable knowledge, skills and experience:

• Experience of working with audio data (sound recordings)                                            
• Experience of working with machine learning algorithms, especially deep learning                    
• An interest in/affinity with biodiversity, and the field of natural sciences and biology

Eligibility requirements: candidates can be of any nationality. You must fulfil the EU’s “mobility rule” for this funding scheme, which means you should not have resided or carried out your main activity (work, studies, etc.) in the PhD host country for more than 12 months during the 36 months prior to starting the PhD. You must not already have a doctoral degree. You must also be eligible to work according to your host country’s regulations (e.g. visa requirements), and also to travel/work to other European countries for collaborations and workshops.

The PhD is planned to begin in Jan/Feb/Mar 2024.

Fixed-term contract: 1 year, with the possibility of an extension.

We provide a vibrant and inspiring working environment, with ample support to our scientific staff by our Research Coordination Office and provide access to state-of-the-art laboratory facilities and the 43 million natural history objects in our collection. At Naturalis you will get the opportunity to develop your research career and to cooperate with inspiring colleagues. We offer a contract (36 hours per week, fulltime) for a period of one year, to be extended with three years after successful evaluation and a monthly starting salary of  € 3,187,-  gross per month. You also get an allowance for travel expenses, holiday allowance (8%) and year-end bonus (3.4%). 

 

Naturalis Biodiversity Center in Leiden

Naturalis Biodiversity Center in Leiden is the Dutch national research institute for biodiversity and systematics. We host over 120 researchers including 15 academia embedded professors and 50 PhD students. We closely collaborate with many Dutch universities, research institutes, industry, and government. With our collection of 43 million specimens, one of the world's largest natural history collections, and our state-of-the-art research facilities we offer the (inter)national research infrastructure for species identification and monitoring (for example in the ARISE project). We present the history of our planet, and the diversity of life on Earth, through permanent and temporary museum exhibitions, educational programmes, and online presence, with more than 400,000 visitors per year. All in all, a unique combination of science and culture in the Netherlands and elsewhere in the world!

Are you an enthusiastic researcher with a completed master's degree and programming skills? Would you like to develop deep learning algorithms to understand animal sounds? To collaborate with AI and biodiversity experts, as part of a team and with partners in many European countries?