Pranav Sharma

Pranav Sharma, Bowen Huang, Umesh Vaidya et al.

In this paper, we propose linear operator theoretic framework involving Koopman operator for the data-driven identification of power system dynamics. We explicitly account for noise in the time series measurement data and propose robust approach for data-driven approximation of Koopman operator for the identification of nonlinear power system dynamics. The identified model is used for the prediction of state trajectories in the power system. The application of the framework is illustrated using an IEEE nine bus test system.

Pranav Sharma, Jigyasa Singh Katrolia, Jason Rambach et al.

Depth is a very important modality in computer vision, typically used as complementary information to RGB, provided by RGB-D cameras. In this work, we show that it is possible to obtain the same level of accuracy as RGB-D cameras on a semantic segmentation task using infrared (IR) and depth images from a single Time-of-Flight (ToF) camera. In order to fuse the IR and depth modalities of the ToF camera, we introduce a method utilizing depth-specific convolutions in a multi-task learning framework. In our evaluation on an in-car segmentation dataset, we demonstrate the competitiveness of our method against the more costly RGB-D approaches.

Yiren Liu, Pranav Sharma, Mehul Jitendra Oswal et al.

Generating interdisciplinary research ideas requires diverse domain expertise, but access to timely feedback is often limited by the availability of experts. In this paper, we introduce PersonaFlow, a novel system designed to provide multiple perspectives by using LLMs to simulate domain-specific experts. Our user studies showed that the new design 1) increased the perceived relevance and creativity of ideated research directions, and 2) promoted users' critical thinking activities (e.g., interpretation, analysis, evaluation, inference, and self-regulation), without increasing their perceived cognitive load. Moreover, users' ability to customize expert profiles significantly improved their sense of agency, which can potentially mitigate their over-reliance on AI. This work contributes to the design of intelligent systems that augment creativity and collaboration, and provides design implications of using customizable AI-simulated personas in domains within and beyond research ideation.

Pranav Sharma

The goal of serving and delighting customers in a personal and near human like manner is very high on automation agendas of most Enterprises. Last few years, have seen huge progress in Natural Language Processing domain which has led to deployments of conversational agents in many enterprises. Most of the current industrial deployments tend to use Monolithic Single Agent designs that model the entire knowledge and skill of the Domain. While this approach is one of the fastest to market, the monolithic design makes it very hard to scale beyond a point. There are also challenges in seamlessly leveraging many tools offered by sub fields of Natural Language Processing and Information Retrieval in a single solution. The sub fields that can be leveraged to provide relevant information are, Question and Answer system, Abstractive Summarization, Semantic Search, Knowledge Graph etc. Current deployments also tend to be very dependent on the underlying Conversational AI platform (open source or commercial) , which is a challenge as this is a fast evolving space and no one platform can be considered future proof even in medium term of 3-4 years. Lately,there is also work done to build multi agent solutions that tend to leverage a concept of master agent. While this has shown promise, this approach still makes the master agent in itself difficult to scale. To address these challenges, we introduce LPar, a distributed multi agent platform for large scale industrial deployment of polyglot, diverse and inter-operable agents. The asynchronous design of LPar supports dynamically expandable domain. We also introduce multiple strategies available in the LPar system to elect the most suitable agent to service a customer query.

Bhaskar Singh, Shobhit Banga, Pranav Sharma

Full-duplex spoken dialogue systems can model natural conversational behaviours such as interruptions, overlaps, and backchannels, yet such systems remain largely unexplored for Indian languages. We present the first open, reproducible full-duplex spoken dialogue system for Hindi by adapting Moshi, a state-of-the-art duplex speech architecture, using a custom Hindi tokeniser and training on 26,000 hours of real spontaneous conversations collected from 14,695 speakers with separate speaker channels, enabling direct learning of turn-taking and overlap patterns from natural interactions. To support Hindi text generation, we replace the original English tokeniser and reinitialise text-vocabulary-dependent parameters while retaining the pre-trained audio components. We propose a two-stage training recipe -- large-scale pre-training followed by fine-tuning on 1,000 hours of conversational data. Evaluation through the prompted dialogue continuation paradigm with both automatic metrics and human judgments demonstrates that the resulting model generates natural and meaningful full-duplex conversational behaviour in Hindi. This work serves as a first step toward real-time duplex spoken dialogue systems for Hindi and other Indian languages.

Bhargav Ghanekar, Salman Siddique Khan, Pranav Sharma et al.

Passive, compact, single-shot 3D sensing is useful in many application areas such as microscopy, medical imaging, surgical navigation, and autonomous driving where form factor, time, and power constraints can exist. Obtaining RGB-D scene information over a short imaging distance, in an ultra-compact form factor, and in a passive, snapshot manner is challenging. Dual-pixel (DP) sensors are a potential solution to achieve the same. DP sensors collect light rays from two different halves of the lens in two interleaved pixel arrays, thus capturing two slightly different views of the scene, like a stereo camera system. However, imaging with a DP sensor implies that the defocus blur size is directly proportional to the disparity seen between the views. This creates a trade-off between disparity estimation vs. deblurring accuracy. To improve this trade-off effect, we propose CADS (Coded Aperture Dual-Pixel Sensing), in which we use a coded aperture in the imaging lens along with a DP sensor. In our approach, we jointly learn an optimal coded pattern and the reconstruction algorithm in an end-to-end optimization setting. Our resulting CADS imaging system demonstrates improvement of >1.5dB PSNR in all-in-focus (AIF) estimates and 5-6% in depth estimation quality over naive DP sensing for a wide range of aperture settings. Furthermore, we build the proposed CADS prototypes for DSLR photography settings and in an endoscope and a dermoscope form factor. Our novel coded dual-pixel sensing approach demonstrates accurate RGB-D reconstruction results in simulations and real-world experiments in a passive, snapshot, and compact manner.

Pranav Sharma, Shivank Garg, Durga Toshniwal

Recent advances in image generation models have led to models that produce synthetic images that are increasingly difficult for standard AI detectors to identify, even though they often remain distinguishable by humans. To identify this discrepancy, we introduce \textbf{Mirage}, a curated dataset comprising a diverse range of AI-generated images exhibiting visible artifacts, where current state-of-the-art detection methods largely fail. Furthermore, we investigate whether Large Vision-Language Models (LVLMs), which are increasingly employed as substitutes for human judgment in various tasks, can be leveraged for explainable AI image detection. Our experiments on both Mirage and existing benchmark datasets demonstrate that while LVLMs are highly effective at detecting AI-generated images with visible artifacts, their performance declines when confronted with images lacking such cues.

Mehdi Rezagholizadeh, Aref Jafari, Puneeth Salad et al.

With ever growing scale of neural models, knowledge distillation (KD) attracts more attention as a prominent tool for neural model compression. However, there are counter intuitive observations in the literature showing some challenging limitations of KD. A case in point is that the best performing checkpoint of the teacher might not necessarily be the best teacher for training the student in KD. Therefore, one important question would be how to find the best checkpoint of the teacher for distillation? Searching through the checkpoints of the teacher would be a very tedious and computationally expensive process, which we refer to as the \textit{checkpoint-search problem}. Moreover, another observation is that larger teachers might not necessarily be better teachers in KD which is referred to as the \textit{capacity-gap} problem. To address these challenging problems, in this work, we introduce our progressive knowledge distillation (Pro-KD) technique which defines a smoother training path for the student by following the training footprints of the teacher instead of solely relying on distilling from a single mature fully-trained teacher. We demonstrate that our technique is quite effective in mitigating the capacity-gap problem and the checkpoint search problem. We evaluate our technique using a comprehensive set of experiments on different tasks such as image classification (CIFAR-10 and CIFAR-100), natural language understanding tasks of the GLUE benchmark, and question answering (SQuAD 1.1 and 2.0) using BERT-based models and consistently got superior results over state-of-the-art techniques.

Tianda Li, Ahmad Rashid, Aref Jafari et al.

Knowledge Distillation (KD) is a model compression algorithm that helps transfer the knowledge of a large neural network into a smaller one. Even though KD has shown promise on a wide range of Natural Language Processing (NLP) applications, little is understood about how one KD algorithm compares to another and whether these approaches can be complimentary to each other. In this work, we evaluate various KD algorithms on in-domain, out-of-domain and adversarial testing. We propose a framework to assess the adversarial robustness of multiple KD algorithms. Moreover, we introduce a new KD algorithm, Combined-KD, which takes advantage of two promising approaches (better training scheme and more efficient data augmentation). Our extensive experimental results show that Combined-KD achieves state-of-the-art results on the GLUE benchmark, out-of-domain generalization, and adversarial robustness compared to competitive methods.

Aref Jafari, Mehdi Rezagholizadeh, Pranav Sharma et al.

Significant memory and computational requirements of large deep neural networks restrict their application on edge devices. Knowledge distillation (KD) is a prominent model compression technique for deep neural networks in which the knowledge of a trained large teacher model is transferred to a smaller student model. The success of knowledge distillation is mainly attributed to its training objective function, which exploits the soft-target information (also known as "dark knowledge") besides the given regular hard labels in a training set. However, it is shown in the literature that the larger the gap between the teacher and the student networks, the more difficult is their training using knowledge distillation. To address this shortcoming, we propose an improved knowledge distillation method (called Annealing-KD) by feeding the rich information provided by the teacher's soft-targets incrementally and more efficiently. Our Annealing-KD technique is based on a gradual transition over annealed soft-targets generated by the teacher at different temperatures in an iterative process, and therefore, the student is trained to follow the annealed teacher output in a step-by-step manner. This paper includes theoretical and empirical evidence as well as practical experiments to support the effectiveness of our Annealing-KD method. We did a comprehensive set of experiments on different tasks such as image classification (CIFAR-10 and 100) and NLP language inference with BERT-based models on the GLUE benchmark and consistently got superior results.

Subhrajit Sinha, Pranav Sharma, Umesh Vaidya et al.

In this paper, we present a novel approach to identify the generators and states responsible for the small-signal stability of power networks. To this end, the newly developed notion of information transfer between the states of a dynamical system is used. In particular, using the concept of information transfer, which characterizes influence between the various states and a linear combination of states of a dynamical system, we identify the generators and states which are responsible for causing instability of the power network. While characterizing influence from state to state, information transfer can also describe influence from state to modes thereby generalizing the well-known notion of participation factor while at the same time overcoming some of the limitations of the participation factor. The developed framework is applied to study the three bus system identifying various cause of instabilities in the system. The simulation study is extended to IEEE 39 bus system.