Publications

Home robots operate in diverse and dynamic environments, delivering a range of functions that enhance utility. Many of these functions span extended periods, from weeks to months, typically improving through observations and interactions. Efficient development and validation of these functions necessitate simulations that can run faster than real time. However, many current robot simulators focus on high-fidelity

Low-precision computation has emerged as one of the most effective techniques for accelerating convolutional neural networks and has garnered widespread support on modern hardware. Despite its effectiveness in accelerating convolutional neural networks, low-precision computation has not been commonly applied to fast convolutions, such as the Winograd algorithm, due to numerical issues. In this paper, we

Spoken language understanding (SLU) systems often exhibit suboptimal performance in processing atypical speech, typically caused by neurological conditions and motor impairments. Recent advancements in Text-to-Speech (TTS) synthesis-based augmentation for more fair SLU have struggled to accurately capture the unique vocal characteristics of atypical speakers, largely due to insufficient data. To address

With increasing scale in model and dataset size, the training of deep neural networks becomes a massive computational burden. One approach to speed up the training process is Selective Backprop. For this approach, we perform a forward pass to obtain a loss value for each data point in a minibatch. The backward pass is then restricted to a subset of that minibatch, prioritizing high-loss examples. We build

Most linear experimental design problems assume homogeneous variance, even though heteroskedastic noise is present in many realistic settings. Let a learner have access to a finite set of measurement vectors X ⊂ ℝd that can be probed to receive noisy linear responses of the form y = x⊤θ* + η. Here θ* ∈ ℝd is an unknown parameter vector, and η is independent mean-zero σx2 -strictly-sub-Gaussian noise defined

Activity classification has become a vital feature of wearable health tracking devices. As innovation in this field grows, wearable devices worn on different parts of the body are emerging. To perform activity classification on a new body location, labeled data corresponding to the new locations are generally required, but this is expensive to acquire. In this work, we present an innovative method to leverage

To build scalable models for the challenging real-world tasks, it is important to learn from diverse, multi-modal data in various forms (e.g., videos, text, images). Amongst the existing works, a plethora of them have been focusing on leveraging large but cumbersome cross-modal architectures. Regardless of their effectiveness, larger architectures unavoidably prevent the models from being extended to realworld

The Net Promoter Score (NPS) is often used in customer experience programs for measuring customer loyalty. Increasingly more companies seek to automatically process millions of pieces of customer feedback from social media per month in order to estimate their NPS, leveraging advanced analytics like machine learning (ML) and natural language processing (NLP). Discovering trends and themes in customer interactions

The recent surge in Large Language Model (LLM) related applications has led to a concurrent escalation in expectations for LLMs to accommodate a myriad of personas and encompass a broad spectrum of perspectives. An important first step towards addressing this demand is to align language models with specific personas, be it groups of users or individuals. Towards this goal, we first present a new conceptualization

Hierarchies are common structures used to organize data, such as e-commerce hierarchies associated with product data. With these product hierarchies, we aim to learn hierarchy-aware product text embeddings to improve fine-tuning performance on a variety of downstream e-commerce tasks. Existing methods leverage hierarchies by either aligning the text embeddings to separate hierarchical embeddings or by aligning

Recent works have shown great success in synchronizing lip-movements in a given video with a dubbed audio stream. However, comparison and efficacy of the synchronization capabilities of these methods is still weakly substantiated due to the lack of a generalized and visually-grounded evaluation method. This work proposes a simple and grounded algorithm – PhoVis, that can measure synchronization and the

Robust, low-loss photonic packaging of on-chip nanophotonic circuits is a key enabling technology for the deployment of integrated photon- ics in a variety of classical and quantum technologies including optical communications and quantum communications, sensing, and trans- duction. To date, no process has been established that enables permanent, broadband, and cryogenically compatible coupling with sub-dB

This paper proposes a framework leveraging small samples from different Automatic Speech Recognition (ASR) data sources to predict model performance and facilitate ASR data selection decisions. By utilizing data distribution distance and a mapping technique inspired by neural scaling laws, our framework estimates the model performance for various data mixtures within the disclosed range and extrapolates

Writing radiology reports from medical images requires a high level of domain expertise. It is time-consuming even for trained radiologists and can be error-prone for inexperienced radiologists. It would be appealing to automate this task by leveraging generative AI, which has shown drastic progress in vision and language understanding. In particular, Large Language Models (LLM) have demonstrated impressive

Many control systems rely on a pipeline of machine learning models and handcoded rules to make decisions. However, due to changes in the operating environment, these rules require constant tuning to maintain optimal system performance. Reinforcement learning (RL) can automate the online optimization of rules based on incoming data. However, RL requires extensive training data and exploration, which limits