Transfer learning (TL) can improve data-driven modeling of building thermal dynamics. Therefore, many new TL research areas emerge in the field, such as selecting the right source model for TL. However, these research directions require massive amounts of thermal building data which is lacking presently. Neither public datasets nor existing data generators meet the needs of TL research in terms of data quality and quantity. Moreover, existing data generation approaches typically require expert knowledge in building simulation. We present BuilDa, a thermal building data generation framework for producing synthetic data of adequate quality and quantity for TL research. The framework does not require profound building simulation knowledge to generate large volumes of data. BuilDa uses a single-zone Modelica model that is exported as a Functional Mock-up Unit (FMU) and simulated in Python. We demonstrate BuilDa by generating data and utilizing it for pretraining and fine-tuning TL models.

Accurate and efficient temperature prediction is critical for optimizing the preheating process of PET preforms in industrial microwave systems prior to blow molding. We propose a novel deep learning framework for generalized temperature prediction. Unlike traditional models that require extensive retraining for each material or design variation, our method introduces a data-efficient neural architecture that leverages transfer learning and model fusion to generalize across unseen scenarios. By pretraining specialized neural regressor on distinct conditions such as recycled PET heat capacities or varying preform geometries and integrating their representations into a unified global model, we create a system capable of learning shared thermal dynamics across heterogeneous inputs. The architecture incorporates skip connections to enhance stability and prediction accuracy. Our approach reduces the need for large simulation datasets while achieving superior performance compared to models trained from scratch. Experimental validation on two case studies material variability and geometric diversity demonstrates significant improvements in generalization, establishing a scalable ML-based solution for intelligent thermal control in manufacturing environments. Moreover, the approach highlights how data-efficient generalization strategies can extend to other industrial applications involving complex physical modeling with limited data.

The prediction of crystal properties is essential for understanding structure-property relationships and accelerating the discovery of functional materials. However, conventional approaches relying on experimental measurements or density functional theory (DFT) calculations are often resource-intensive, limiting their scalability. Machine learning (ML) models offer a promising alternative by learning complex structure-property relationships from data, enabling faster predictions. Yet, existing ML models often rely on labeled data, adopt representations that poorly capture essential structural characteristics, and lack integration with physical principles--factors that limit their generalizability and interpretability. Here, we introduce CLOUD (Crystal Language mOdel for Unified and Differentiable materials modeling), a transformer-based framework trained on a novel Symmetry-Consistent Ordered Parameter Encoding (SCOPE) that encodes crystal symmetry, Wyckoff positions, and composition in a compact, coordinate-free string representation. Pre-trained on over six million crystal structures, CLOUD is fine-tuned on multiple downstream tasks and achieves competitive performance in predicting a wide range of material properties, demonstrating strong scaling performance. Furthermore, as proof of concept of differentiable materials modeling, CLOUD is applied to predict the phonon internal energy and heat capacity, which integrates the Debye model to preserve thermodynamic consistency. The CLOUD-DEBYE framework enforces thermodynamic consistency and enables temperature-dependent property prediction without requiring additional data. These results demonstrate the potential of CLOUD as a scalable and physics-informed foundation model for crystalline materials, unifying symmetry-consistent representations with physically grounded learning for property prediction and materials discovery.

We propose a variational modelling method with differentiable temperature for canonical ensembles. Using a deep generative model, the free energy is estimated and minimized simultaneously in a continuous temperature range. At optimal, this generative model is a Boltzmann distribution with temperature dependence. The training process requires no dataset, and works with arbitrary explicit density generative models. We applied our method to study the phase transitions (PT) in the Ising and XY models, and showed that the direct-sampling simulation of our model is as accurate as the Markov Chain Monte Carlo (MCMC) simulation, but more efficient. Moreover, our method can give thermodynamic quantities as differentiable functions of temperature akin to an analytical solution. The free energy aligns closely with the exact one to the second-order derivative, so this inclusion of temperature dependence enables the otherwise biased variational model to capture the subtle thermal effects at the PTs. These findings shed light on the direct simulation of physical systems using deep generative models

Modern commercial Heating, Ventilation, and Air Conditioning (HVAC) devices form a complex and interconnected thermodynamic system with the building and outside weather conditions, and current setpoint control policies are not fully optimized for minimizing energy use and carbon emission. Given a suitable training environment, a Reinforcement Learning (RL) model is able to improve upon these policies, but training such a model, especially in a way that scales to thousands of buildings, presents many real world challenges. We propose a novel simulation-based approach, where a customized simulator is used to train the agent for each building. Our open-source simulator (available online: https://github.com/google/sbsim) is lightweight and calibrated via telemetry from the building to reach a higher level of fidelity. On a two-story, 68,000 square foot building, with 127 devices, we were able to calibrate our simulator to have just over half a degree of drift from the real world over a six-hour interval. This approach is an important step toward having a real-world RL control system that can be scaled to many buildings, allowing for greater efficiency and resulting in reduced energy consumption and carbon emissions.

A goal of unsupervised machine learning is to build representations of complex high-dimensional data, with simple relations to their properties. Such disentangled representations make easier to interpret the significant latent factors of variation in the data, as well as to generate new data with desirable features. Methods for disentangling representations often rely on an adversarial scheme, in which representations are tuned to avoid discriminators from being able to reconstruct information about the data properties (labels). Unfortunately adversarial training is generally difficult to implement in practice. Here we propose a simple, effective way of disentangling representations without any need to train adversarial discriminators, and apply our approach to Restricted Boltzmann Machines (RBM), one of the simplest representation-based generative models. Our approach relies on the introduction of adequate constraints on the weights during training, which allows us to concentrate information about labels on a small subset of latent variables. The effectiveness of the approach is illustrated with four examples: the CelebA dataset of facial images, the two-dimensional Ising model, the MNIST dataset of handwritten digits, and the taxonomy of protein families. In addition, we show how our framework allows for analytically computing the cost, in terms of log-likelihood of the data, associated to the disentanglement of their representations.

This work describes experiments on thermal dynamics of pure H2O excited by hydrodynamic cavitation, which has been reported to facilitate the spin conversion of para- and ortho-isomers at water interfaces. Previous measurements by NMR and capillary methods of excited samples demonstrated changes of proton density by 12-15%, the surface tension up to 15.7%, which can be attributed to a non-equilibrium para-/ortho- ratio. Beside these changes, we also expect a variation of heat capacity. Experiments use a differential calorimetric approach with two devices: one with an active thermostat for diathermic measurements, another is fully passive for long-term measurements. Samples after excitation are degassed at -0.09MPa and thermally equalized in a water bath. Conducted attempts demonstrated changes in the heat capacity of experimental samples by 4.17%--5.72% measured in the transient dynamics within 60 min after excitation, which decreases to 2.08% in the steady-state dynamics 90-120 min after excitation. Additionally, we observed occurrence of thermal fluctuations at the level of 10^-3 C relative temperature on 20-40 min mesoscale dynamics and a long-term increase of such fluctuations in experimental samples. Obtained results are reproducible in both devices and are supported by previously published outcomes on four-photon scattering spectra in the range from -1.5 to 1.5 cm^-1 and electrochemical reactivity in CO2 and H2O2 pathways. Based on these results, we propose a hypothesis about ongoing spin conversion process on mesoscopic scales under weak influx of energy caused by thermal, EM or geomagnetic factors; this enables explaining electrochemical and thermal anomalies observed in long-term measurements.

Graph Neural Networks (GNNs) have recently gained attention in the machine learning community. They have achieved state-of-the-art performance in a number of tasks by leveraging the geometric prior inherent to many real-world problems [1]. Concurrently, several model-agnostic algorithms for meta-learning have been developed, such as Model-Agnostic Meta-Learning (MAML) [2] and Reptile [3]. Although as their name suggests these algorithms are model agnostic, works in the literature have mainly applied them to classical fully-connected and convolutional neural networks. In this paper, we explore the application of Reptile to GNN regression tasks. We show that modelagnostic algorithms for meta-learning are also applicable to GNNs and specifically, that meta-learning can exploit the underlying structure of molecules to quickly adapt models to learning new molecular regression tasks. We experimentally demonstrate that GNN expressivity is correlated to metalearning performance. Finally, we also show that using GNN ensembles can even further improve meta-learning.

Metal-organic frameworks (MOFs) are a class of materials that contain nano-sized pores. These pores give MOFs record-breaking internal surface areas, which make them extremely versatile for a number of applications: separating petrochemicals and gases, mimicking DNA, producing hydrogen, and removing heavy metals, fluoride anions, and even gold from water are just a few examples. MOFs are the focus of Professor Berend Smit's research at EPFL School of Basic Sciences, where his group employs machine learning in the discovery, design, and even categorization of the ever-increasing MOFs that currently flood chemical databases. In a new study, Smit and his colleagues have developed a machine-learning model that predicts the heat capacity of MOFs. "This is about very classical thermodynamics," says Smit. "How much energy is needed to heat up a material by one degree? Until now, all engineering calculations have assumed that all MOFs have the same heat capacity, for the simple reason that there is hardly any data available."

However, discovering new phenomena is not the only challenge: utilizing the freshly obtained knowledge for real-world applications is crucial. With the availability of computers and large amounts of experimental/- computational data nowadays [1-4], machine learning [5-9] has proven an effective tool for this purpose. Machine learning is a class of methods that start from existing data to train a model and then predict the quantities of interest useful for a given application. For example, machine learning can predict many properties of a putative material [10-18], and moreover can understand the uncertainty in those predictions. This uncertainty can be used to design the material that is most likely to satisfy the set target criteria [19-21], avoiding the typical expensive and time-consuming cycles of trial and improvement experiments. Furthermore, the uncertainty is useful for accelerating materials discovery by guiding where new experiments should be performed in the materials space [22-24], and also for the identification of outliers and erroneous entries in materials databases [25]. While uncertainty is crucial for focusing on the most viable candidates for a given application, uncertainty itself could be a useful value for property prediction. This strategy is motivated by Wilson's Renormalization Group theory [26], in which fluctuations on all scales determine the macroscopic state of the system.