Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
Insect-like plume tracking with reinforcement learning
Flying insects excel at solving the computational challenge of tracking of odour plumes. Many aspects of the associated behaviour and the underlying neural circuitry are well studied, but measuring neural activity directly in freely behaving insects is not tractable. Singh et al. developed a complementary in silico approach that involves recurrent neural network artificial agents that use deep reinforcement learning to locate the source of simulated odour plumes. The trained agents produce trajectories with a strong resemblance to those of flying insects and learn to compute task-relevant variables with distinct dynamic structures in population activity.
To fully leverage big data, they need to be shared across institutions in a manner compliant with privacy considerations and the EU General Data Protection Regulation (GDPR). Federated machine learning is a promising option.
A recent case of a flawed medical AI system that was backed by public funding provides an opportunity to discuss the impact of government policies and regulation in AI.
Antibodies are an essential class of therapeutics but low breadth or off-target binding are major concerns for antibody–drug efficiency and safety. To predict which targets an antibody can neutralize, a machine learning pipeline based on an adaptive graph convolutional network architecture is proposed that learns the binding landscape of antibodies to multiple mutated viruses at the same time.
Machine translation of languages can now automatically detect different cell types from single-cell transcriptomic data. Such a feat opens the prospect of dissecting complex clinical samples such as heterogenous tumours at scale.
Gathering big datasets has become an essential component of machine learning in many scientific areas, but it is unavoidable that some data values are missing. An important and growing effect that needs careful attention, especially when heterogeneous data sources are combined, is that of structured missingness, where data values are missing not at random, but with a specific structure.
The reconstruction of spatially resolved information of an extended object from an observed intensity diffraction pattern in holographic imaging is a challenging problem. By incorporating an explicit physical model, Lee and colleagues propose a deep learning method that can be used in holographic image reconstruction under physical perturbations and which generalizes well beyond object-to-sensor distances and pixel sizes seen during training.
AI language modelling and generation approaches have developed fast in the last decade, opening promising new directions in human–AI collaboration. An AI-in-the loop conversational system called HAILEY is developed to empower peer supporters in providing empathic responses to mental health support seekers.
Olfactory navigation is a well-studied topic in insect behaviour, but many aspects of the challenging task of odour plume tracking are unknown. In a deep reinforcement learning approach, artificial agents are trained to produce (in silico) trajectories to localize the source of an odour plume, showing dynamics that mimic real insect behaviours.
Despite recent improvements in microscopy acquisition methods, extracting quantitative information from biological experiments in crowded conditions is a challenging task. Pineda and colleagues propose a geometric deep-learning-based framework for automated trajectory linking and dynamical property estimation that is able to effectively deal with complex biological scenarios.
When it comes to reasoning about the motion of physical objects, humans have natural intuitive physics knowledge. To test how good artificial learning agents are in similar predictive abilities, Xue and colleagues present a benchmark based on a two-dimensional physics environment in which 15 physical reasoning skills are measured.