Development of modern deep learning methods has been driven primarily by the push for improving model efficacy (accuracy metrics). This sole focus on efficacy has steered development of large-scale models that require massive resources, and results in considerable carbon footprint across the model life-cycle. In this work, we explore how physics inductive biases can offer useful trade-offs between model efficacy and model efficiency (compute, energy, and carbon). We study a variety of models for spatio-temporal forecasting, a task governed by physical laws and well-suited for exploring different levels of physics inductive bias. We show that embedding physics inductive biases into the model design can yield substantial efficiency gains while retaining or even improving efficacy for the tasks under consideration. In addition to using standard physics-informed spatio-temporal models, we demonstrate the usefulness of more recent models like flow matching as a general purpose method for spatio-temporal forecasting. Our experiments show that incorporating physics inductive biases offer a principled way to improve the efficiency and reduce the carbon footprint of machine learning models. We argue that model efficiency, along with model efficacy, should become a core consideration driving machine learning model development and deployment.

The modern theory of stochastic control typically assumes complete knowledge of the underlying system dynamics. While significant theoretical advancements have been made in this area, see Øksendal and Sulem 2019; Fleming and Soner 2006, the practical application of stochastic control often faces challenges when the system model is uncertain or unknown. In recent years, Reinforcement learning (RL) has emerged as a promising approach to address this issue, enabling agents to learn optimal control policies through trial-and-error interactions with the environment. However, RL's success often hinges on the availability of vast amounts of data, and the learned control policies can be difficult to interpret, especially when deep learning techniques are employed, see Sutton 2018. To bridge the gap between fully model-based and model-free approaches, research has increasingly focused on model-based reinforcement learning.

We consider ordinary least squares estimation and variations on least squares estimation such as penalized (regularized) least squares and spectral shrinkage estimates for problems with p > n and associated problems with prediction of new observations. After the introduction of Section 1, Section 2 examines a number of commonly used estimators for p > n. Section 3 introduces prediction with p > n. Section 4 introduces notational changes to facilitate discussion of overfitting and Section 5 illustrates the phenomenon of double descent. We conclude with some final comments.

Autoencoders are powerful machine learning models used to compress information from multiple data sources. However, autoencoders, like all artificial neural networks, are often unidentifiable and uninterpretable. This research focuses on creating an identifiable and interpretable autoencoder that can be used to meld and combine climate data products. The proposed autoencoder utilizes a Bayesian statistical framework, allowing for probabilistic interpretations while also varying spatially to capture useful spatial patterns across the various data products. Constraints are placed on the autoencoder as it learns patterns in the data, creating an interpretable consensus that includes the important features from each input. We demonstrate the utility of the autoencoder by combining information from multiple precipitation products in High Mountain Asia.

Atmospheric models used for weather and climate prediction are traditionally formulated in a deterministic manner. In other words, given a particular state of the resolved scale variables, the most likely forcing from the sub-grid scale processes is estimated and used to predict the evolution of the large-scale flow. However, the lack of scale-separation in the atmosphere means that this approach is a large source of error in forecasts. Over recent years, an alternative paradigm has developed: the use of stochastic techniques to characterise uncertainty in small-scale processes. These techniques are now widely used across weather, sub-seasonal, seasonal, and climate timescales. In parallel, recent years have also seen significant progress in replacing parametrisation schemes using machine learning (ML). This has the potential to both speed up and improve our numerical models. However, the focus to date has largely been on deterministic approaches. In this position paper, we bring together these two key developments, and discuss the potential for data-driven approaches for stochastic parametrisation. We highlight early studies in this area, and draw attention to the novel challenges that remain.

Artificial intelligence and machine learning (AI/ML) are key to enabling drug discovery and development, and Pfizer is leading the biopharma industry into the next wave of innovation. The company is rapidly scaling up and recruiting talent for a collaborative effort intended to get transformative medicines to patients faster. The mandate is "uncompromising and extremely high-quality science," Sandeep Menon, chief scientific officer, AI digital sciences, SVP and head of early clinical development told BioSpace. The vision is three-fold: uncover disease biology with AI; use these insights to design the right molecules; determine the right patient population for clinical trial success. "We're building the next generation of tools to use across the preclinical and clinical development spectrum," said Jared Christensen, vice president and head of early clinical development, clinical AI/ML and quantitative sciences.

Leading Senior Constable Dr Janis Dalins is looking for 100,000 happy images of children – a toddler in a sandpit, a nine-year-old winning an award at school, a sullen teenager unwrapping a present at Christmas and pretending not to care. The search for these safe, happy pictures is the goal of a new campaign to crowdsource a database of ethically obtained images that Dalins hopes will help build better investigative tools to use in the fight against what some have called a "tsunami" of child sexual assault material online. Dalins is the co-director of AiLecs lab, a collaboration between Monash University and the Australian federal police, which builds artificial intelligence technologies for use by law enforcement. In its new My Pictures Matter campaign, people above 18 are being asked to share safe photos of themselves at different stages of their childhood. Once uploaded with information identifying the age and person in the image, these will go into a database of other safe images.

An app called TapCounter records each time I touch my phone's screen. My swipes and jabs are averaging about 1,000 a day, though I notice that's falling as I steer shy of social media to meet my deadline. The European company behind it, QuantActions, promises that through capturing and analysing the data it will be able to "detect important indicators related to mental/neurological health". Arko Ghosh is the company's cofounder and a neuroscientist at Leiden University in the Netherlands. "Tappigraphy patterns" – the time series of my touches – can, he says, confidently be used not only to infer slumber habits (tapping in the wee hours means you are not sleeping) but also mental performance level (the small intervals in a series of key-presses represent a proxy for reaction time), and he has published work to support it.

Xelex Digital announced the release of its new audio and text annotation platform, WebChartAi, designed to accelerate the adoption of machine learning applications by simplifying the creation of training data at scale. "There's an explosion of companies seeking to leverage the wealth of intelligence available in audio and text-based data," said Mark Christensen, CEO of Xelex Digital. "WebChartAi lowers the technology barrier to entry, enabling a broader base of companies to more easily harness the power of NLP-driven task automation." In machine learning applications utilizing natural language processing (NLP), the NLP engine is trained to automatically identify actionable intelligence within media posts, customer service interactions, search queries, product reviews, and other audio and text-based sources. The training process requires large volumes of data to be manually annotated, and that annotation process (sometimes called labeling or classifying) is accomplished through WebChartAi.

Inside a Schnucks grocery store in St. Louis, Missouri, the toilet paper and baking ingredients are mostly cleared out. A rolling robot turns a corner and heads down an aisle stocked with salsa and taco shells. It comes up against a masked customer wearing shorts and sneakers; he's pushing a shopping cart carrying bread. The robot looks something like a tower speaker on top of an autonomous home vacuum cleaner--tall and thin, with orb-like screen eyes halfway up that shift left and right. A red sign on its long head makes the introductions. Tally freezes, sensing the human, and the customer pauses, seeming unsure of what to do next. Should he maneuver around the robot? Or wait for it to move along on its own?