Search

In this article, we introduce a novel problem of audio-visual autism behavior recognition, which includes social behavior recognition, an essential aspect previously omitted in AI-assisted autism screening research. We define the task at hand as one that is audio-visual autism behavior recognition, which uses audio and visual cues, including any speech present in the audio, to recognize autism-related behaviors. To facilitate this new research direction, we collected an audio-visual autism spectrum dataset (AV-ASD), currently the largest video dataset for autism screening using a behavioral approach. It covers an extensive range of autism-associated behaviors, including those related to social communication and interaction. To pave the way for further research on this new problem, we intensively explored leveraging foundation models and multimodal large language models across different modalities. Our experiments on the AV-ASD dataset demonstrate that integrating audio, visual, and speech modalities significantly enhances the performance in autism behavior recognition. Additionally, we explored the use of a post-hoc to ad-hoc pipeline in a multimodal large language model to investigate its potential to augment the model's explanatory capability during autism behavior recognition. We will release our dataset, code, and pre-trained models.

State-of-the-art hybrid perovskites have demonstrated excellent functionality in photovoltaics and light-emitting applications, and have emerged as a promising candidate for exciton-polariton (polariton) optoelectronics. In the strong coupling regime, polariton formation and Bose-Einstein condensation (BEC) have been demonstrated at room-temperature in several perovskite formulations. Thermodynamically, low-threshold BEC requires efficient scattering to $k_{||}$ = 0, and many applications demand precise control of polariton interactions. Thus far, the primary mechanisms by which polaritons relax in perovskites remains unclear. In this work, we perform temperature-dependent measurements of polaritons in low-dimensional hybrid perovskites with high light-matter coupling strengths ($\hbar \Omega_{Rabi}$ = 260$\pm$5 meV). By embedding the perovskite active layer in a wedged cavity, we are able to tune the Hopfield coefficients and decouple the primary polariton relaxation mechanisms in this material for the first time. We observe the thermal activation of a bottleneck regime, and reveal that this effect can be overcome by harnessing intrinsic scattering mechanisms arising from the interplay between the different excitonic species, such as biexciton-assisted polariton relaxation pathways, and isoenergetic intracavity pumping. We demonstrate the dependence of the bottleneck suppression on cavity detuning, and are able to achieve efficient relaxation to $k_{||}$ = 0 even at cryogenic temperatures. This new understanding contributes to the design of ultra-low-threshold BEC and condensate control by engineering polariton dispersions resonant with efficient relaxation pathways, leveraging intrinsic material scattering mechanisms for next-generation polariton optoelectronics., Comment: *Corresponding authors: danedeq@mit.edu, bulovic@mit.edu

Chemically prepared colloidal semiconductor quantum dots have long been proposed as scalable and color-tunable single emitters in quantum optics, but they have typically suffered from prohibitively incoherent emission. We now demonstrate that individual colloidal lead halide perovskite quantum dots (PQDs) display highly efficient single photon emission with optical coherence times as long as 80 ps, an appreciable fraction of their 210 ps radiative lifetimes. These measurements suggest that PQDs should be explored as building blocks in sources of indistinguishable single photons and entangled photon pairs. Our results present a starting point for the rational design of lead halide perovskite-based quantum emitters with fast emission, wide spectral-tunability, scalable production, and which benefit from the hybrid-integration with nano-photonic components that has been demonstrated for colloidal materials., Comment: Main text - 20 pages, 4 figures. Supplementary Material - 11 pages, 8 figures

It was predicted long ago that ultra-thin metallic films must exhibit unusual optical properties for radiation frequencies from rf to infrared domain. A film would remain highly reflective even when it is orders of magnitude thinner than a skin depth at any frequency. Only when it is a few nanometers thick (depending on material but not on the frequency), its reflectivity and transmittivity get equal, while its absorption peaks at 50%. It has been confirmed experimentally and new directions and applications were proposed. We review the EM theory of the phenomenon and recent developments in the field, and present some new results., Comment: 13 pages; 5 figs; 1 table

Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU)---deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs (caches, out-of-order execution, multithreading, multiprocessing, prefetching, ...) that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95% of our datacenters' NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X - 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X - 80X higher. Moreover, using the GPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU., Comment: 17 pages, 11 figures, 8 tables. To appear at the 44th International Symposium on Computer Architecture (ISCA), Toronto, Canada, June 24-28, 2017

A methodology for binary classification of EEG records which correspond to different mental states is proposed. This model-free methodology is based on our theory of the $\epsilon$-complexity of continuous functions which is extended here (see Appendix) to the case of vector functions. This extension permits us to handle multichannel EEG recordings. The essence of the methodology is to use the $\epsilon$-complexity coefficients as features to classify (using well known classifiers) different types of vector functions representing EEG-records corresponding to different types of mental states. We apply our methodology to the problem of classification of multichannel EEG-records related to a group of healthy adolescents and a group of adolescents with schizophrenia. We found that our methodology permits accurate classification of the data in the four-dimensional feather space of the $\epsilon$-complexity coefficients., Comment: 14 pages, 2 figures