Ankit Gupta

Ankit Gupta, Albert Gu, Jonathan Berant · deepmind, ibm-research

Modeling long range dependencies in sequential data is a fundamental step towards attaining human-level performance in many modalities such as text, vision, audio and video. While attention-based models are a popular and effective choice in modeling short-range interactions, their performance on tasks requiring long range reasoning has been largely inadequate. In an exciting result, Gu et al. (ICLR 2022) proposed the $\textit{Structured State Space}$ (S4) architecture delivering large gains over state-of-the-art models on several long-range tasks across various modalities. The core proposition of S4 is the parameterization of state matrices via a diagonal plus low rank structure, allowing efficient computation. In this work, we show that one can match the performance of S4 even without the low rank correction and thus assuming the state matrices to be diagonal. Our $\textit{Diagonal State Space}$ (DSS) model matches the performance of S4 on Long Range Arena tasks, speech classification on Speech Commands dataset, while being conceptually simpler and straightforward to implement.

Guy Dar, Mor Geva, Ankit Gupta et al. · deepmind, ibm-research

Understanding Transformer-based models has attracted significant attention, as they lie at the heart of recent technological advances across machine learning. While most interpretability methods rely on running models over inputs, recent work has shown that a zero-pass approach, where parameters are interpreted directly without a forward/backward pass is feasible for some Transformer parameters, and for two-layer attention networks. In this work, we present a theoretical analysis where all parameters of a trained Transformer are interpreted by projecting them into the embedding space, that is, the space of vocabulary items they operate on. We derive a simple theoretical framework to support our arguments and provide ample evidence for its validity. First, an empirical analysis showing that parameters of both pretrained and fine-tuned models can be interpreted in embedding space. Second, we present two applications of our framework: (a) aligning the parameters of different models that share a vocabulary, and (b) constructing a classifier without training by ``translating'' the parameters of a fine-tuned classifier to parameters of a different model that was only pretrained. Overall, our findings open the door to interpretation methods that, at least in part, abstract away from model specifics and operate in the embedding space only.

Ankit Gupta, Ida-Maria Sintorn

Attention Branch Networks (ABNs) have been shown to simultaneously provide visual explanation and improve the performance of deep convolutional neural networks (CNNs). In this work, we introduce Multi-Scale Attention Branch Networks (MSABN), which enhance the resolution of the generated attention maps, and improve the performance. We evaluate MSABN on benchmark image recognition and fine-grained recognition datasets where we observe MSABN outperforms ABN and baseline models. We also introduce a new data augmentation strategy utilizing the attention maps to incorporate human knowledge in the form of bounding box annotations of the objects of interest. We show that even with a limited number of edited samples, a significant performance gain can be achieved with this strategy.

Ankit Gupta, George Saon, Brian Kingsbury · ibm-research

The emergence of industrial-scale speech recognition (ASR) models such as Whisper and USM, trained on 1M hours of weakly labelled and 12M hours of audio only proprietary data respectively, has led to a stronger need for large scale public ASR corpora and competitive open source pipelines. Unlike the said models, large language models are typically based on Transformer decoders, and it remains unclear if decoder-only models trained on public data alone can deliver competitive performance. In this work, we investigate factors such as choice of training datasets and modeling components necessary for obtaining the best performance using public English ASR corpora alone. Our Decoder-Only Transformer for ASR (DOTA) model comprehensively outperforms the encoder-decoder open source replication of Whisper (OWSM) on nearly all English ASR benchmarks and outperforms Whisper large-v3 on 7 out of 15 test sets. We release our codebase and model checkpoints under permissive license.

Linyong Nan, Dragomir Radev, Rui Zhang et al.

We present DART, an open domain structured DAta Record to Text generation dataset with over 82k instances (DARTs). Data-to-Text annotations can be a costly process, especially when dealing with tables which are the major source of structured data and contain nontrivial structures. To this end, we propose a procedure of extracting semantic triples from tables that encodes their structures by exploiting the semantic dependencies among table headers and the table title. Our dataset construction framework effectively merged heterogeneous sources from open domain semantic parsing and dialogue-act-based meaning representation tasks by utilizing techniques such as: tree ontology annotation, question-answer pair to declarative sentence conversion, and predicate unification, all with minimum post-editing. We present systematic evaluation on DART as well as new state-of-the-art results on WebNLG 2017 to show that DART (1) poses new challenges to existing data-to-text datasets and (2) facilitates out-of-domain generalization. Our data and code can be found at https://github.com/Yale-LILY/dart.

Uri Shaham, Elad Segal, Maor Ivgi et al.

NLP benchmarks have largely focused on short texts, such as sentences and paragraphs, even though long texts comprise a considerable amount of natural language in the wild. We introduce SCROLLS, a suite of tasks that require reasoning over long texts. We examine existing long-text datasets, and handpick ones where the text is naturally long, while prioritizing tasks that involve synthesizing information across the input. SCROLLS contains summarization, question answering, and natural language inference tasks, covering multiple domains, including literature, science, business, and entertainment. Initial baselines, including Longformer Encoder-Decoder, indicate that there is ample room for improvement on SCROLLS. We make all datasets available in a unified text-to-text format and host a live leaderboard to facilitate research on model architecture and pretraining methods.

Ayush Hariharan, Ankit Gupta, Trisha Pal

As machine learning and cybersecurity continue to explode in the context of the digital ecosystem, the complexity of cybersecurity data combined with complicated and evasive machine learning algorithms leads to vast difficulties in designing an end to end system for intelligent, automatic anomaly classification. On the other hand, traditional systems use elementary statistics techniques and are often inaccurate, leading to weak centralized data analysis platforms. In this paper, we propose a novel system that addresses these two problems, titled CAMLPAD, for Cybersecurity Autonomous Machine Learning Platform for Anomaly Detection. The CAMLPAD systems streamlined, holistic approach begins with retrieving a multitude of different species of cybersecurity data in real time using elasticsearch, then running several machine learning algorithms, namely Isolation Forest, Histogram Based Outlier Score (HBOS), Cluster Based Local Outlier Factor (CBLOF), and K Means Clustering, to process the data. Next, the calculated anomalies are visualized using Kibana and are assigned an outlier score, which serves as an indicator for whether an alert should be sent to the system administrator that there are potential anomalies in the network. After comprehensive testing of our platform in a simulated environment, the CAMLPAD system achieved an adjusted rand score of 95 percent, exhibiting the reliable accuracy and precision of the system. All in all, the CAMLPAD system provides an accurate, streamlined approach to real time cybersecurity anomaly detection, delivering a novel solution that has the potential to revolutionize the cybersecurity sector.

Ankit Gupta, Elliott Ruebush

According to the United Nations World Water Assessment Programme, every day, 2 million tons of sewage and industrial and agricultural waste are discharged into the worlds water. In order to address this pervasive issue of increasing water pollution, while ensuring that the global population has an efficient, accurate, and low cost method to assess whether the water they drink is contaminated, we propose AquaSight, a novel mobile application that utilizes deep learning methods, specifically Convolutional Neural Networks, for automated water impurity detection. After comprehensive training with a dataset of 105 images representing varying magnitudes of contamination, the deep learning algorithm achieved a 96 percent accuracy and loss of 0.108. Furthermore, the machine learning model uses efficient analysis of the turbidity and transparency levels of water to estimate a particular sample of waters level of contamination. When deployed, the AquaSight system will provide an efficient way for individuals to secure an estimation of water quality, alerting local and national government to take action and potentially saving millions of lives worldwide.

Ankit Gupta, Alexander M. Rush

We consider the task of detecting regulatory elements in the human genome directly from raw DNA. Past work has focused on small snippets of DNA, making it difficult to model long-distance dependencies that arise from DNA's 3-dimensional conformation. In order to study long-distance dependencies, we develop and release a novel dataset for a larger-context modeling task. Using this new data set we model long-distance interactions using dilated convolutional neural networks, and compare them to standard convolutions and recurrent neural networks. We show that dilated convolutions are effective at modeling the locations of regulatory markers in the human genome, such as transcription factor binding sites, histone modifications, and DNAse hypersensitivity sites.