Arnav Jain

Arnav Kumar Jain, Shivakanth Sujit, Shruti Joshi et al. · deepmind, mila

Learning world models from their sensory inputs enables agents to plan for actions by imagining their future outcomes. World models have previously been shown to improve sample-efficiency in simulated environments with few objects, but have not yet been applied successfully to environments with many objects. In environments with many objects, often only a small number of them are moving or interacting at the same time. In this paper, we investigate integrating this inductive bias of sparse interactions into the latent dynamics of world models trained from pixels. First, we introduce Variational Sparse Gating (VSG), a latent dynamics model that updates its feature dimensions sparsely through stochastic binary gates. Moreover, we propose a simplified architecture Simple Variational Sparse Gating (SVSG) that removes the deterministic pathway of previous models, resulting in a fully stochastic transition function that leverages the VSG mechanism. We evaluate the two model architectures in the BringBackShapes (BBS) environment that features a large number of moving objects and partial observability, demonstrating clear improvements over prior models.

Arnav Kumar Jain, Lucas Lehnert, Irina Rish et al.

Animals have a developed ability to explore that aids them in important tasks such as locating food, exploring for shelter, and finding misplaced items. These exploration skills necessarily track where they have been so that they can plan for finding items with relative efficiency. Contemporary exploration algorithms often learn a less efficient exploration strategy because they either condition only on the current state or simply rely on making random open-loop exploratory moves. In this work, we propose $ηψ$-Learning, a method to learn efficient exploratory policies by conditioning on past episodic experience to make the next exploratory move. Specifically, $ηψ$-Learning learns an exploration policy that maximizes the entropy of the state visitation distribution of a single trajectory. Furthermore, we demonstrate how variants of the predecessor representation and successor representations can be combined to predict the state visitation entropy. Our experiments demonstrate the efficacy of $ηψ$-Learning to strategically explore the environment and maximize the state coverage with limited samples.

Arnav Jain, Jasmer Singh Sanjotra, Harshvardhan Choudhary et al.

In this paper, we propose long short term memory speech enhancement network (LSTMSE-Net), an audio-visual speech enhancement (AVSE) method. This innovative method leverages the complementary nature of visual and audio information to boost the quality of speech signals. Visual features are extracted with VisualFeatNet (VFN), and audio features are processed through an encoder and decoder. The system scales and concatenates visual and audio features, then processes them through a separator network for optimized speech enhancement. The architecture highlights advancements in leveraging multi-modal data and interpolation techniques for robust AVSE challenge systems. The performance of LSTMSE-Net surpasses that of the baseline model from the COG-MHEAR AVSE Challenge 2024 by a margin of 0.06 in scale-invariant signal-to-distortion ratio (SISDR), $0.03$ in short-time objective intelligibility (STOI), and $1.32$ in perceptual evaluation of speech quality (PESQ). The source code of the proposed LSTMSE-Net is available at \url{https://github.com/mtanveer1/AVSEC-3-Challenge}.

Wenjie Xi, Arnav Jain, Li Zhang et al.

Time series classification is an important data mining task that has received a lot of interest in the past two decades. Due to the label scarcity in practice, semi-supervised time series classification with only a few labeled samples has become popular. Recently, Similarity-aware Time Series Classification (SimTSC) is proposed to address this problem by using a graph neural network classification model on the graph generated from pairwise Dynamic Time Warping (DTW) distance of batch data. It shows excellent accuracy and outperforms state-of-the-art deep learning models in several few-label settings. However, since SimTSC relies on pairwise DTW distances, the quadratic complexity of DTW limits its usability to only reasonably sized datasets. To address this challenge, we propose a new efficient semi-supervised time series classification technique, LB-SimTSC, with a new graph construction module. Instead of using DTW, we propose to utilize a lower bound of DTW, LB_Keogh, to approximate the dissimilarity between instances in linear time, while retaining the relative proximity relationships one would have obtained via computing DTW. We construct the pairwise distance matrix using LB_Keogh and build a graph for the graph neural network. We apply this approach to the ten largest datasets from the well-known UCR time series classification archive. The results demonstrate that this approach can be up to 104x faster than SimTSC when constructing the graph on large datasets without significantly decreasing classification accuracy.

Siddharth Joshi, Arnav Jain, Ali Payani et al.

Contrastive Language-Image Pre-training (CLIP) on large-scale image-caption datasets learns representations that can achieve remarkable zero-shot generalization. However, such models require a massive amount of pre-training data. Improving the quality of the pre-training data has been shown to be much more effective in improving CLIP's performance than increasing its volume. Nevertheless, finding small subsets of training data that provably generalize the best has remained an open question. In this work, we propose the first theoretically rigorous data selection method for CLIP. We show that subsets that closely preserve the cross-covariance of the images and captions of the full data provably achieve a superior generalization performance. Our extensive experiments on ConceptualCaptions3M and ConceptualCaptions12M demonstrate that subsets found by \method\ achieve over 2.7x and 1.4x the accuracy of the next best baseline on ImageNet and its shifted versions. Moreover, we show that our subsets obtain 1.5x the average accuracy across 11 downstream datasets, of the next best baseline. The code is available at: https://github.com/BigML-CS-UCLA/clipcov-data-efficient-clip.

Arnav Kumar Jain, Gonzalo Gonzalez-Pumariega, Wayne Chen et al.

We address the problem of code generation from multi-turn execution feedback. Existing methods either generate code without feedback or use complex, hierarchical reinforcement learning to optimize multi-turn rewards. We propose a simple yet scalable approach, $μ$Code, that solves multi-turn code generation using only single-step rewards. Our key insight is that code generation is a one-step recoverable MDP, where the correct code can be recovered from any intermediate code state in a single turn. $μ$Code iteratively trains both a generator to provide code solutions conditioned on multi-turn execution feedback and a verifier to score the newly generated code. Experimental evaluations show that our approach achieves significant improvements over the state-of-the-art baselines. We provide analysis of the design choices of the reward models and policy, and show the efficacy of $μ$Code at utilizing the execution feedback. Our code is available at https://github.com/portal-cornell/muCode.

William Yang, Noah Bergam, Arnav Jain et al.

In this paper, we consider the extent to which the transformer-based Dense Passage Retrieval (DPR) algorithm, developed by (Karpukhin et. al. 2020), can be optimized without further pre-training. Our method involves two particular insights: we apply the DPR context encoder at various phrase lengths (e.g. one-sentence versus five-sentence segments), and we take a confidence-calibrated ensemble prediction over all of these different segmentations. This somewhat exhaustive approach achieves start-of-the-art results on benchmark datasets such as Google NQ and SQuAD. We also apply our method to domain-specific datasets, and the results suggest how different granularities are optimal for different domains

Arnav Kumar Jain, Harley Wiltzer, Jesse Farebrother et al.

In inverse reinforcement learning (IRL), an agent seeks to replicate expert demonstrations through interactions with the environment. Traditionally, IRL is treated as an adversarial game, where an adversary searches over reward models, and a learner optimizes the reward through repeated RL procedures. This game-solving approach is both computationally expensive and difficult to stabilize. In this work, we propose a novel approach to IRL by direct policy optimization: exploiting a linear factorization of the return as the inner product of successor features and a reward vector, we design an IRL algorithm by policy gradient descent on the gap between the learner and expert features. Our non-adversarial method does not require learning a reward function and can be solved seamlessly with existing actor-critic RL algorithms. Remarkably, our approach works in state-only settings without expert action labels, a setting which behavior cloning (BC) cannot solve. Empirical results demonstrate that our method learns from as few as a single expert demonstration and achieves improved performance on various control tasks.

Shahid Shafi Dar, Bharat Kaurav, Arnav Jain et al.

In an era of escalating climate change, urban flooding has emerged as a critical challenge for sustainable cities, threatening lives, infrastructure, and ecosystems. Traditional flood detection methods are constrained by their reliance on unimodal data and static rule-based systems, which fail to capture the dynamic, non-linear relationships inherent in flood events. Furthermore, existing attention mechanisms and ensemble learning approaches exhibit limitations in hierarchical refinement, cross-modal feature integration, and adaptability to noisy or unstructured environments, resulting in suboptimal flood classification performance. To address these challenges, we present XFloodNet, a novel framework that redefines urban flood classification through advanced deep-learning techniques. XFloodNet integrates three novel components: (1) a Hierarchical Cross-Modal Gated Attention mechanism that dynamically aligns visual and textual features, enabling precise multi-granularity interactions and resolving contextual ambiguities; (2) a Heterogeneous Convolutional Adaptive Multi-Scale Attention module, which leverages frequency-enhanced channel attention and frequency-modulated spatial attention to extract and prioritize discriminative flood-related features across spectral and spatial domains; and (3) a Cascading Convolutional Transformer Feature Refinement technique that harmonizes hierarchical features through adaptive scaling and cascading operations, ensuring robust and noise-resistant flood detection. We evaluate our proposed method on three benchmark datasets, such as Chennai Floods, Rhine18 Floods, and Harz17 Floods, XFloodNet achieves state-of-the-art F1-scores of 93.33%, 82.24%, and 88.60%, respectively, surpassing existing methods by significant margins.

Hadia Mohmmed Osman Ahmed Samil, Annabelle Martin, Arnav Kumar Jain et al.

Locust infestation of some regions in the world, including Africa, Asia and Middle East has become a concerning issue that can affect the health and the lives of millions of people. In this respect, there have been attempts to resolve or reduce the severity of this problem via detection and monitoring of locust breeding areas using satellites and sensors, or the use of chemicals to prevent the formation of swarms. However, such methods have not been able to suppress the emergence and the collective behaviour of locusts. The ability to predict the location of the locust swarms prior to their formation, on the other hand, can help people get prepared and tackle the infestation issue more effectively. Here, we use machine learning to predict the location of locust swarms using the available data published by the Food and Agriculture Organization of the United Nations. The data includes the location of the observed swarms as well as environmental information, including soil moisture and the density of vegetation. The obtained results show that our proposed model can successfully, and with reasonable precision, predict the location of locust swarms, as well as their likely level of damage using a notion of density.

Avisek Lahiri, Arnav Kumar Jain, Prabir Kumar Biswas

Contemporary deep learning based inpainting algorithms are mainly based on a hybrid dual stage training policy of supervised reconstruction loss followed by an unsupervised adversarial critic loss. However, there is a dearth of literature for a fully unsupervised GAN based inpainting framework. The primary aversion towards the latter genre is due to its prohibitively slow iterative optimization requirement during inference to find a matching noise prior for a masked image. In this paper, we show that priors matter in GAN: we learn a data driven parametric network to predict a matching prior for a given image. This converts an iterative paradigm to a single feed forward inference pipeline with a massive 1500X speedup and simultaneous improvement in reconstruction quality. We show that an additional structural prior imposed on GAN model results in higher fidelity outputs. To extend our model for sequence inpainting, we propose a recurrent net based grouped noise prior learning. To our knowledge, this is the first demonstration of an unsupervised GAN based sequence inpainting. A further improvement in sequence inpainting is achieved with an additional subsequence consistency loss. These contributions improve the spatio-temporal characteristics of reconstructed sequences. Extensive experiments conducted on SVHN, Standford Cars, CelebA and CelebA-HQ image datasets, synthetic sequences and ViDTIMIT video datasets reveal that we consistently improve upon previous unsupervised baseline and also achieve comparable performances(sometimes also better) to hybrid benchmarks.

Avisek Lahiri, Arnav Kumar Jain, Divyasri Nadendla et al.

In this paper, we propose to improve the inference speed and visual quality of contemporary baseline of Generative Adversarial Networks (GAN) based unsupervised semantic inpainting. This is made possible with better initialization of the core iterative optimization involved in the framework. To our best knowledge, this is also the first attempt of GAN based video inpainting with consideration to temporal cues. On single image inpainting, we achieve about 4.5-5$\times$ speedup and 80$\times$ on videos compared to baseline. Simultaneously, our method has better spatial and temporal reconstruction qualities as found on three image and one video dataset.

Avisek Lahiri, Arnav Jain, Divyasri Nadendla et al.

In this paper we present several architectural and optimization recipes for generative adversarial network(GAN) based facial semantic inpainting. Current benchmark models are susceptible to initial solutions of non-convex optimization criterion of GAN based inpainting. We present an end-to-end trainable parametric network to deterministically start from good initial solutions leading to more photo realistic reconstructions with significant optimization speed up. For the first time, we show how to efficiently extend GAN based single image inpainter models to sequences by a)learning to initialize a temporal window of solutions with a recurrent neural network and b)imposing a temporal smoothness loss(during iterative optimization) to respect the redundancy in temporal dimension of a sequence. We conduct comprehensive empirical evaluations on CelebA images and pseudo sequences followed by real life videos of VidTIMIT dataset. The proposed method significantly outperforms current GAN based state-of-the-art in terms of reconstruction quality with a simultaneous speedup of over 15$\times$. We also show that our proposed model is better in preserving facial identity in a sequence even without explicitly using any face recognition module during training.

Avisek Lahiri, Arnav Jain, Prabir Kumar Biswas et al.

Contemporary benchmark methods for image inpainting are based on deep generative models and specifically leverage adversarial loss for yielding realistic reconstructions. However, these models cannot be directly applied on image/video sequences because of an intrinsic drawback- the reconstructions might be independently realistic, but, when visualized as a sequence, often lacks fidelity to the original uncorrupted sequence. The fundamental reason is that these methods try to find the best matching latent space representation near to natural image manifold without any explicit distance based loss. In this paper, we present a semantically conditioned Generative Adversarial Network (GAN) for sequence inpainting. The conditional information constrains the GAN to map a latent representation to a point in image manifold respecting the underlying pose and semantics of the scene. To the best of our knowledge, this is the first work which simultaneously addresses consistency and correctness of generative model based inpainting. We show that our generative model learns to disentangle pose and appearance information; this independence is exploited by our model to generate highly consistent reconstructions. The conditional information also aids the generator network in GAN to produce sharper images compared to the original GAN formulation. This helps in achieving more appealing inpainting performance. Though generic, our algorithm was targeted for inpainting on faces. When applied on CelebA and Youtube Faces datasets, the proposed method results in a significant improvement over the current benchmark, both in terms of quantitative evaluation (Peak Signal to Noise Ratio) and human visual scoring over diversified combinations of resolutions and deformations.

Arnav Kumar Jain, Abhinav Agarwalla, Kumar Krishna Agrawal et al.

In this paper, we introduce Key-Value Memory Networks to a multimodal setting and a novel key-addressing mechanism to deal with sequence-to-sequence models. The proposed model naturally decomposes the problem of video captioning into vision and language segments, dealing with them as key-value pairs. More specifically, we learn a semantic embedding (v) corresponding to each frame (k) in the video, thereby creating (k, v) memory slots. We propose to find the next step attention weights conditioned on the previous attention distributions for the key-value memory slots in the memory addressing schema. Exploiting this flexibility of the framework, we additionally capture spatial dependencies while mapping from the visual to semantic embedding. Experiments done on the Youtube2Text dataset demonstrate usefulness of recurrent key-addressing, while achieving competitive scores on BLEU@4, METEOR metrics against state-of-the-art models.

Abhinav Agarwalla, Arnav Kumar Jain, KV Manohar et al.

We integrate learning and motion planning for soccer playing differential drive robots using Bayesian optimisation. Trajectories generated using end-slope cubic Bezier splines are first optimised globally through Bayesian optimisation for a set of candidate points with obstacles. The optimised trajectories along with robot and obstacle positions and velocities are stored in a database. The closest planning situation is identified from the database using k-Nearest Neighbour approach. It is further optimised online through reuse of prior information from previously optimised trajectory. Our approach reduces computation time of trajectory optimisation considerably. Velocity profiling generates velocities consistent with robot kinodynamoic constraints, and avoids collision and slipping. Extensive testing is done on developed simulator, as well as on physical differential drive robots. Our method shows marked improvements in mitigating tracking error, and reducing traversal and computational time over competing techniques under the constraints of performing tasks in real time.