Understanding how explicit theoretical features are encoded in opaque neural systems is a central challenge now common to neuroscience and AI. We introduce Metric Learning Encoding Models (MLEMs) to address this challenge most directly as a metric learning problem: we fit the distance in the space of theoretical features to match the distance in neural space. Our framework improves on univariate encoding and decoding methods by building on second-order isomorphism methods, such as Representational Similarity Analysis, and extends them by learning a metric that efficiently models feature as well as interactions between them. The effectiveness of MLEM is validated through two sets of simulations. First, MLEMs recover ground-truth importance features in synthetic datasets better than state-of-the-art methods, such as Feature Reweighted RSA (FR-RSA). Second, we deploy MLEMs on real language data, where they show stronger robustness to noise in calculating the importance of linguistic features (gender, tense, etc.). MLEMs are applicable to any domains where theoretical features can be identified, such as language, vision, audition, etc.

Diffusion models have recently enabled state-of-the-art reconstruction of positron emission tomography (PET) images while requiring only image training data. However, domain shift remains a key concern for clinical adoption: priors trained on images from one anatomy, acquisition protocol or pathology may produce artefacts on out-of-distribution data. We propose integrating steerable conditional diffusion (SCD) with our previously-introduced likelihood-scheduled diffusion (PET-LiSch) framework to improve the alignment of the diffusion model's prior to the target subject. At reconstruction time, for each diffusion step, we use low-rank adaptation (LoRA) to align the diffusion model prior with the target domain on the fly. Experiments on realistic synthetic 2D brain phantoms demonstrate that our approach suppresses hallucinated artefacts under domain shift, i.e. when our diffusion model is trained on perturbed images and tested on normal anatomy, our approach suppresses the hallucinated structure, outperforming both OSEM and diffusion model baselines qualitatively and quantitatively. These results provide a proof-of-concept that steerable priors can mitigate domain shift in diffusion-based PET reconstruction and motivate future evaluation on real data.

Do architectural differences significantly affect the way models represent and process language? We propose a new approach, based on metric-learning encoding models (MLEMs), as a first step to answer this question. The approach provides a feature-based comparison of how any two layers of any two models represent linguistic information. We apply the method to BERT, GPT-2 and Mamba. Unlike previous methods, MLEMs offer a transparent comparison, by identifying the specific linguistic features responsible for similarities and differences. More generally, the method uses formal, symbolic descriptions of a domain, and use these to compare neural representations. As such, the approach can straightforwardly be extended to other domains, such as speech and vision, and to other neural systems, including human brains.

This study investigates self-supervised learning techniques to obtain representations of Event Sequences. It is a key modality in various applications, including but not limited to banking, e-commerce, and healthcare. We perform a comprehensive study of generative and contrastive approaches in self-supervised learning, applying them both independently. We find that there is no single supreme method. Consequently, we explore the potential benefits of combining these approaches. To achieve this goal, we introduce a novel method that aligns generative and contrastive embeddings as distinct modalities, drawing inspiration from contemporary multimodal research. Generative and contrastive approaches are often treated as mutually exclusive, leaving a gap for their combined exploration. Our results demonstrate that this aligned model performs at least on par with, and mostly surpasses, existing methods and is more universal across a variety of tasks. Furthermore, we demonstrate that self-supervised methods consistently outperform the supervised approach on our datasets.

X-ray computed tomography is a versatile technique for 3D structure characterization, and the pursuit of a faster yet reliable scan is never ended. A lot of methods, such as the maximum likelihood expectation maximization (MLEM) and maximum-a-posteriori (MAP), have been proposed and developed to improve the speed, but most of they are mainly for the "step-scan" mode. At synchrotron facilities such as the FXI beamline at Brookhaven National Laboratory, data is normally collected in the high-speed "fly-scan" mode, which inevitably results in a blurred image using traditional reconstruction algorithms. Figure 1 illustrates how a "fly-scan" mode can introduce artifacts due to the nature of rotation. MLEM and MAP TV methods can be employed on top of the FBP reconstruction algorithm, however, their performance is still limited (see Figure 3).

Iterative, the MLOps company dedicated to streamlining the workflow of data scientists and machine learning (ML) engineers, has launched machine learning engineering management (MLEM) - an open source model deployment and registry tool that uses an organisation's existing Git infrastructure and workflows. According to the company, MLEM is designed to bridge the gap between ML engineers and DevOps teams. DevOps teams can understand the underlying frameworks and libraries a model uses and automate deployment into a one-step process for production services and apps, Iterative states. IDC AI/ML Lifecycle Management Softwrae research director Sriram Subramanian says, "Iterative enables customers to treat AI models as just another type of software artifact. The ability to build ML model registries using Git infrastructure and DevOps principles allows models to get into production faster."

MLOps platform Iterative, which announced a $20 million Series A round almost exactly a year ago, today launched MLEM, an open-source git-based machine learning model management and deployment tool. The idea here, the company says, is to bridge the gap between ML engineers and DevOps teams by using the git-based approach that developers are already familiar with. Using MLEM, developers can store and track their ML models throughout their lifecycle. As such, it complements Iterative's open-source GTO artifact registry and DVC, the company's version control system for data and models. "Having a machine learning model registry is becoming an essential part of the machine learning technology stack. Current SaaS solutions can lead to a divergence in the lifecycle of ML models and software applications," said Dmitry Petrov, co-founder and CEO of Iterative.

When your best friend catches you in a bad mood, does she try to console you, give you space to cool off or lick her own face in an uncontrollable slobber? If your best friend is a dog, this third reaction may be familiar to you. Certain cuteness-obsessed Internet communities call it a "mlem"; some animal behavior researchers prefer to call it mouth-licking, and offer many possible explanations for the quirky canine behavior. Mouth-licking has been described as a stress-coping mechanism, a spontaneous display of arousal or a way to communicate desire to play with a certain toy or munch a certain treat. But according to a new study by animal behavior researchers from the University of Sao Paulo, Brazil, mouth-licking may actually be one of a dog's best tools for reading and responding to human faces -- in particular, angry faces.