We introduce a framework to analyse interpretability in deep learning, by drawing on a formal notion of model semantics from the philosophy of science. We argue that interpretability is only one aspect of a models semantics and illustrate our framework with examples from biomedicine.
We demonstrate agnostic QoT probing for datacenter exchange in a metro network via receiver-side ASE loading. Knowing BER telemetry and the progressive ASEload, the device estimates GSNR, enabling IPoWDM operations and digital-twin calibration.
We report the first field study of the phase and polarization dynamics of deployed antiresonant hollow core fiber cable in a data center interconnect for real-world vibration sensing,revealing enhanced phase sensitivity and significantly faster polarization angular rate compared with standard single mode fibers.
We propose a method to compensate the phase offset between samples from different tributaries in time-interleaved phase OTDR using nested phase reference channels. We demonstrate our method for a four-span bidirectional link with high-loss loopback.
We propose a mobile orbital domain-based hierarchical routing scheme which addresses the challenges posed by constant satellite movement and the resulting dynamicnetwork topology, thus significantly improving the routing scalability and efficiency in satellite networks.
As AI data centers grow, the fiber-optic networks that connect massive computing clusters become critical infrastructure. This talk explores how Distributed Fiber Optic Sensing (DFOS) can turn communication cables into real-time sensors that detect physical threats and improve cyber-physical resilience.
Many pioneers of modern artificial intelligence and machine learning spent part of their careers at NEC research labs in the United States. Researchers such as Yann LeCun, Vladimir Vapnik, Léon Bottou, Corinna Cortes, and others contributed foundational ideas in deep learning, statistical learning theory, speech recognition, and computer vision.
Vision Transformers (ViTs) have achieved state-of-the-art performance in offline video action detection, but their reliance on processing fixed-size clips with full spatio-temporal attention makes them computationally expensive and ill-suited for real-time streaming applications due to massive computational redundancy. This paper introduces a novel framework to adapt these powerful offline models into efficient, online student models through knowledge distillation. We propose two causal, streaming-friendly attention architectures that replace the full self-attention mechanism: (1) a lightweight Temporal Shift Attention that integrates past context with minimal overhead, and (2) a Decomposed Spatial-Temporal Attention that combines intra-frame spatial attention with an LSTM for temporal modeling. Both architectures utilize caching to eliminate redundant operations on a frame-by-frame basis. To maximize knowledge transfer, we introduce an uncertainty-guided distillation process, which formulates the training as a multi-task learning problem. Our resulting models demonstrate significant efficiency gains, achieving up to a4x improvement in latency and throughput compared to the original offline teacher while ensuring state-of-the-art performance for online methods. Our work provides a practical and effective methodology for deploying high-accuracy transformer models in latency-sensitive, real-world video analysis systems.
Vision transformer-based models bring significant improvements for image segmentation tasks. Although these architectures offer powerful capabilities irrespective of specific segmentation tasks, their use of computational resources can be taxing on deployed devices. One way to overcome this challenge is by adapting the computation level to the specific needs of the input image rather than the current one-size-fits-all approach. To this end, we introduce ECO-M2F or EffiCient TransfOrmer Encoders for Mask2Former-style models. Noting that the encoder module of M2F-style models incurs high resource-intensive computations, ECO-M2F provides a strategy to self-select the number of hidden layers in the encoder, conditioned on the input image. To enable this self-selection ability for providing a balance between performance and computational efficiency, we present a three-step recipe. The first step is to train the parent architecture to enable early exiting from the encoder. The second step is to create a derived dataset of the ideal number of encoder layers required for each training example. The third step is to use the aforementioned derived dataset to train a gating network that predicts the number of encoder layers to be used, conditioned on the input image. Additionally, to change the computational-accuracy tradeoff, only steps two and three need to be repeated, which significantly reduces retraining time. Experiments on the public datasets show that the proposed approach reduces expected encoder computational cost while maintaining performance, adapts to various user compute resources, is flexible in architecture configurations, and can be extended beyond the segmentation task to object detection.
NEC Laboratories America celebrates Women’s History Month and International Women’s Day by recognizing the women researchers, engineers, and interns whose work in artificial intelligence, optical networking, cybersecurity, and data science is helping shape the future of technology.
