Discovering and optimizing commercially viable materials for clean energy applications typically takes more than a decade. Self-driving laboratories that iteratively design, execute, and learn from materials science experiments in a fully autonomous loop present an opportunity to accelerate this research process. We report here a modular robotic platform driven by a model-based optimization algorithm capable of autonomously optimizing the optical and electronic properties of thin-film materials by modifying the film composition and processing conditions. We demonstrate the power of this platform by using it to maximize the hole mobility of organic hole transport materials commonly used in perovskite solar cells and consumer electronics. This demonstration highlights the possibilities of using autonomous laboratories to discover organic and inorganic materials relevant to materials sciences and clean energy technologies.

Useful materials must satisfy multiple objectives, where the optimization of one objective is often at the expense of another. The Pareto front reports the optimal trade-offs between these conflicting objectives. Here we use a self-driving laboratory, Ada, to define the Pareto front of conductivities and processing temperatures for palladium films formed by combustion synthesis. Ada discovers new synthesis conditions that yield metallic films at lower processing temperatures (below 200 °C) relative to the prior art for this technique (250 °C). This temperature difference makes possible the coating of different commodity plastic materials (e.g., Nafion, polyethersulfone). These combustion synthesis conditions enable us to to spray coat uniform palladium films with moderate conductivity (1.1 × 105 S m−1) at 191 °C. Spray coating at 226 °C yields films with conductivities (2.0 × 106 S m−1) comparable to those of sputtered films (2.0 to 5.8 × 106 S m−1). This work shows how a self-driving laboratoy can discover materials that provide optimal trade-offs between conflicting objectives.

When analysing screening mammograms, radiologists can naturally process information across two ipsilateral views of each breast, namely the cranio-caudal (CC) and mediolateral-oblique (MLO) views. These multiple related images provide complementary diagnostic information and can improve the radiologist's classification accuracy. Unfortunately, most existing deep learning systems, trained with globallylabelled images, lack the ability to jointly analyse and integrate global and local information from these multiple views. By ignoring the potentially valuable information present in multiple images of a screening episode, one limits the potential accuracy of these systems.Here, we propose a new multi-view global-local analysis method that mimics the radiologist's reading procedure, based on a global consistency learning and local co-occurrence learning of ipsilateral views in mammograms. Extensive experiments show that our model outperforms competing methods, in terms of classification accuracy and generalisation, on a large-scale private dataset and two publicly available datasets, where models are exclusively trained and tested with global labels.

Supplemental material is available for this article. Keywords: Mammography, Screening, Convolutional Neural Network (CNN) Published under a CC BY 4.0 license. See also the commentary by Cadrin-Chênevert in this issue.

State-of-the-art (SOTA) deep learning mammogram classifiers, trained with weakly-labelled images, often rely on global models that produce predictions with limited interpretability, which is a key barrier to their successful translation into clinical practice. On the other hand, prototype-based models improve interpretability by associating predictions with training image prototypes, but they are less accurate than global models and their prototypes tend to have poor diversity. We address these two issues with the proposal of BRAIxProtoPNet++, which adds interpretability to a global model by ensembling it with a prototype-based model. BRAIxProtoPNet++ distills the knowledge of the global model when training the prototype-based model with the goal of increasing the classification accuracy of the ensemble. Moreover, we propose an approach to increase prototype diversity by guaranteeing that all prototypes are associated with different training images. Experiments on weakly-labelled private and public datasets show that BRAIx-ProtoPNet++ has higher classification accuracy than SOTA global and prototype-based models. Using lesion localisation to assess model interpretability, we show BRAIxProtoPNet++ is more effective than other prototype-based models and post-hoc explanation of global models. Finally, we show that the diversity of the prototypes learned by BRAIx-ProtoPNet++ is superior to SOTA prototype-based approaches.