Searcharxiv⌕ Search

arXiv subjects

Dmitry Sorokin

Publications and source records attributed to Dmitry Sorokin.

10 recordsLinked to original sources

Beyond AUC: Do Mammography Risk Models Predict When Cancer Occurs?

Longitudinal deep learning models for mammography-based breast cancer risk prediction output either per-year cumulative risks or a discrete distribution over the year of diagnosis, from which cumulative risks F1,...,F5 are derived. These models are evaluated almost exclusively with year-wise AUCs and the concordance index. We show that these metrics are largely insensitive to when cancer occurs: they can be high even for a model that carries no information about the year of diagnosis. We re-evaluate two recent longitudinal models (OA-BreaCR, LMV-Net) on CSAW-CC with a 5-year horizon using timing metrics computed on cancer cases only. Despite C-indices up to 0.75, neither orders cancer cases by time to diagnosis significantly better than chance (c_time 0.51 and 0.55, both 95% CIs covering 0.5). Weighted absolute timing error favours a constant "year 3" prediction over both models, so we use c_time as the primary timing metric. For Mirai's additive-hazard head, also used by LMV-Net, we analyse how per-year increments passed through a ReLU can collapse under severe class imbalance. We also propose a breast-level, two-stage model that separates whether from when. It matches LMV-Net in discrimination (C-index 0.745 vs. 0.746) but orders cancer cases better (c_time 0.58 vs. 0.51, paired 95% CI of the difference 0.01-0.13), on par with Mirai's published weights (0.57).

q-bio.QM↗

Reinforcement Learning Controllers for Fusion Scenario Development

Scenario development, the design of the waveforms that take a tokamak discharge from breakdown to flat-top, is still an iterative, labour-intensive process. Every candidate scenario has to be simulated with a controller in the loop, and the classical shape and current controllers need low-level actuator references. So the process is iterative and includes a number of trial-and-error steps. We propose to replace the per-scenario controller tuning by a single goal-conditioned reinforcement-learning controller trained once, in advance, against a free-boundary equilibrium and transport simulator. A scenario is then specified only by the target waveforms of the physical quantities of interest (plasma current, line-averaged electron density, LCFS shape, x-point positions), and the controller produces the coil-current and fueling commands that realize them. We demonstrate the approach on an EAST plasma-current ramp-up: the controller takes over a small limited plasma, ramps the current and the density, grows the cross-section, forms two x-points and holds a double-null diverted configuration against sampled shape targets for the rest of a 1.2 s episode. The neural network controller is trained using TQC algorithm. On 1024 test scenarios sampled from a low-discrepancy scrambled Sobol sequence in 11-dimensional target space, the controller forms the divertor in 86% of the cases, holds it for more than 0.5 s in 67% and reaches the end of the episode in a diverted state in 58%, with flat-top tracking errors of a few kA in plasma current, $\sim 0.15\times10^{19}$ m$^{-3}$ in density, $\sim 2$ cm in position and $\sim 8$ cm in x-point location; the failures concentrate at the high-current density ($J$) region which is believed to be physically or from engineering perspective not feasible within ohmic mode.

physics.plasm-ph↗

Deformable Medical Image Registration with KAN-based Implicit Neural Representations

Deformable image registration (DIR) is central to medical image analysis, supporting spatial alignment for longitudinal studies and multi-modal fusion. Learning-based methods such as CNNs and transformers provide rapid inference but often require large training datasets and can underperform classical iterative methods for specific anatomies or modalities. Implicit neural representations (INRs) offer a data-efficient alternative by modeling deformation fields as continuous coordinate-to-displacement mappings, yet their per-pair optimization makes runtime efficiency and robustness to initialization essential. We introduce KAN-IDIR and RandKAN-IDIR, the first Kolmogorov--Arnold network (KAN)-based INR framework for pairwise-optimized, resolution-independent DIR, designed to improve seed stability and resource efficiency without dataset-level training. KANs use learnable activation functions that are well suited to continuous, physically structured deformation fields. RandKAN-IDIR further reduces cost through randomized basis sampling, preserving registration quality with fewer basis functions. We evaluate the methods on lung CT, brain MRI, and cardiac MRI datasets against pairwise INR approaches, dataset-trained deep models, and classical baselines. KAN-IDIR and RandKAN-IDIR achieve the highest accuracy among INR-based methods, with low computational overhead and superior stability across random initializations. RandKAN-IDIR slightly outperforms adaptive basis selection variants while avoiding their additional training-time complexity. This makes the approach practical for reproducible clinical research use. Source code is available at https://github.com/anac0der/KAN-IDIR.

cs.CV↗

MGRegBench: A Novel Benchmark Dataset with Anatomical Landmarks for Mammography Image Registration

Robust mammography registration is essential for clinically relevant applications like tracking disease progression in breast tissue. However, progress has been limited by the absence of transparent public datasets and reproducible standardized benchmarks. Existing studies are often not directly comparable, as they use private data and inconsistent evaluation frameworks. To address this, we present MGRegBench, a patient-disjoint, leakage-controlled evaluation protocol for mammography registration, comprising over 5,000 image pairs, each with a breast segmentation mask, and 100 pairs with manually annotated anatomical landmarks, plus standardized train/evaluation splits and ready-to-run baselines. Using this resource, we benchmark diverse registration methods -- including classical (ANTs), learning-based (VoxelMorph, TransMorph), implicit neural representation (IDIR), a mammography-specific approach, and a recent deep learning method MammoRegNet, with implementations adapted to this modality, and validate generalization on the independent SDM-MCs dataset. Our contributions are: (1) the first public dataset of this scale with manual landmarks and masks for mammography registration; (2) a transparent, leakage-controlled benchmark enabling the first like-for-like comparison of diverse classical and machine learning-based methods; (3) external validation on SDM-MCs to test whether the main trend transfers beyond MGRegBench; and (4) an extensive analysis of deep learning-based registration. We publicly release our code and data to establish a foundational resource for fair, reproducible, and clinically relevant comparisons and catalyze future research in AI-driven medical imaging.

cs.CV↗

BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack

In recent years, the input context sizes of large language models (LLMs) have increased dramatically. However, existing evaluation methods have not kept pace, failing to comprehensively assess the efficiency of models in handling long contexts. To bridge this gap, we introduce the BABILong benchmark, designed to test language models' ability to reason across facts distributed in extremely long documents. BABILong includes a diverse set of 20 reasoning tasks, including fact chaining, simple induction, deduction, counting, and handling lists/sets. These tasks are challenging on their own, and even more demanding when the required facts are scattered across long natural text. Our evaluations show that popular LLMs effectively utilize only 10-20\% of the context and their performance declines sharply with increased reasoning complexity. Among alternatives to in-context reasoning, Retrieval-Augmented Generation methods achieve a modest 60\% accuracy on single-fact question answering, independent of context length. Among context extension methods, the highest performance is demonstrated by recurrent memory transformers after fine-tuning, enabling the processing of lengths up to 50 million tokens. The BABILong benchmark is extendable to any length to support the evaluation of new upcoming models with increased capabilities, and we provide splits up to 10 million token lengths.

cs.CL↗

In Search of Needles in a 11M Haystack: Recurrent Memory Finds What LLMs Miss

This paper addresses the challenge of processing long documents using generative transformer models. To evaluate different approaches, we introduce BABILong, a new benchmark designed to assess model capabilities in extracting and processing distributed facts within extensive texts. Our evaluation, which includes benchmarks for GPT-4 and RAG, reveals that common methods are effective only for sequences up to $10^4$ elements. In contrast, fine-tuning GPT-2 with recurrent memory augmentations enables it to handle tasks involving up to $11\times 10^6$ elements. This achievement marks a substantial leap, as it is by far the longest input processed by any neural network model to date, demonstrating a significant improvement in the processing capabilities for long sequences.

cs.CL↗

Insights From the NeurIPS 2021 NetHack Challenge

In this report, we summarize the takeaways from the first NeurIPS 2021 NetHack Challenge. Participants were tasked with developing a program or agent that can win (i.e., 'ascend' in) the popular dungeon-crawler game of NetHack by interacting with the NetHack Learning Environment (NLE), a scalable, procedurally generated, and challenging Gym environment for reinforcement learning (RL). The challenge showcased community-driven progress in AI with many diverse approaches significantly beating the previously best results on NetHack. Furthermore, it served as a direct comparison between neural (e.g., deep RL) and symbolic AI, as well as hybrid systems, demonstrating that on NetHack symbolic bots currently outperform deep RL by a large margin. Lastly, no agent got close to winning the game, illustrating NetHack's suitability as a long-term benchmark for AI research.

cs.LG↗

Aligning an optical interferometer with beam divergence control and continuous action space

Reinforcement learning is finding its way to real-world problem application, transferring from simulated environments to physical setups. In this work, we implement vision-based alignment of an optical Mach-Zehnder interferometer with a confocal telescope in one arm, which controls the diameter and divergence of the corresponding beam. We use a continuous action space; exponential scaling enables us to handle actions within a range of over two orders of magnitude. Our agent trains only in a simulated environment with domain randomizations. In an experimental evaluation, the agent significantly outperforms an existing solution and a human expert.

cs.RO↗

Adaptation of Quadruped Robot Locomotion with Meta-Learning

Animals have remarkable abilities to adapt locomotion to different terrains and tasks. However, robots trained by means of reinforcement learning are typically able to solve only a single task and a transferred policy is usually inferior to that trained from scratch. In this work, we demonstrate that meta-reinforcement learning can be used to successfully train a robot capable to solve a wide range of locomotion tasks. The performance of the meta-trained robot is similar to that of a robot that is trained on a single task.

cs.RO↗

Interferobot: aligning an optical interferometer by a reinforcement learning agent

Limitations in acquiring training data restrict potential applications of deep reinforcement learning (RL) methods to the training of real-world robots. Here we train an RL agent to align a Mach-Zehnder interferometer, which is an essential part of many optical experiments, based on images of interference fringes acquired by a monocular camera. The agent is trained in a simulated environment, without any hand-coded features or a priori information about the physics, and subsequently transferred to a physical interferometer. Thanks to a set of domain randomizations simulating uncertainties in physical measurements, the agent successfully aligns this interferometer without any fine tuning, achieving a performance level of a human expert.

cs.RO↗