Search

In recent years, there has been a growing emphasis on compressing large pre-trained transformer models for resource-constrained devices. However, traditional pruning methods often leave the embedding layer untouched, leading to model over-parameterization. Additionally, they require extensive compression time with large datasets to maintain performance in pruned models. To address these challenges, we propose VTrans, an iterative pruning framework guided by the Variational Information Bottleneck (VIB) principle. Our method compresses all structural components, including embeddings, attention heads, and layers using VIB-trained masks. This approach retains only essential weights in each layer, ensuring compliance with specified model size or computational constraints. Notably, our method achieves upto 70% more compression than prior state-of-the-art approaches, both task-agnostic and task-specific. We further propose faster variants of our method: Fast-VTrans utilizing only 3% of the data and Faster-VTrans, a time efficient alternative that involves exclusive finetuning of VIB masks, accelerating compression by upto 25 times with minimal performance loss compared to previous methods. Extensive experiments on BERT, ROBERTa, and GPT-2 models substantiate the efficacy of our method. Moreover, our method demonstrates scalability in compressing large models such as LLaMA-2-7B, achieving superior performance compared to previous pruning methods. Additionally, we use attention-based probing to qualitatively assess model redundancy and interpret the efficiency of our approach. Notably, our method considers heads with high attention to special and current tokens in un-pruned model as foremost candidates for pruning while retained heads are observed to attend more to task-critical keywords.

Deep extreme classification (XC) aims to train an encoder architecture and an accompanying classifier architecture to tag a data point with the most relevant subset of labels from a very large universe of labels. XC applications in ranking, recommendation and tagging routinely encounter tail labels for which the amount of training data is exceedingly small. Graph convolutional networks (GCN) present a convenient but computationally expensive way to leverage task metadata and enhance model accuracies in these settings. This paper formally establishes that in several use cases, the steep computational cost of GCNs is entirely avoidable by replacing GCNs with non-GCN architectures. The paper notices that in these settings, it is much more effective to use graph data to regularize encoder training than to implement a GCN. Based on these insights, an alternative paradigm RAMEN is presented to utilize graph metadata in XC settings that offers significant performance boosts with zero increase in inference computational costs. RAMEN scales to datasets with up to 1M labels and offers prediction accuracy up to 15% higher on benchmark datasets than state of the art methods, including those that use graph metadata to train GCNs. RAMEN also offers 10% higher accuracy over the best baseline on a proprietary recommendation dataset sourced from click logs of a popular search engine. Code for RAMEN will be released publicly.

This paper develops the MUFIN technique for extreme classification (XC) tasks with millions of labels where datapoints and labels are endowed with visual and textual descriptors. Applications of MUFIN to product-to-product recommendation and bid query prediction over several millions of products are presented. Contemporary multi-modal methods frequently rely on purely embedding-based methods. On the other hand, XC methods utilize classifier architectures to offer superior accuracies than embedding only methods but mostly focus on text-based categorization tasks. MUFIN bridges this gap by reformulating multi-modal categorization as an XC problem with several millions of labels. This presents the twin challenges of developing multi-modal architectures that can offer embeddings sufficiently expressive to allow accurate categorization over millions of labels; and training and inference routines that scale logarithmically in the number of labels. MUFIN develops an architecture based on cross-modal attention and trains it in a modular fashion using pre-training and positive and negative mining. A novel product-to-product recommendation dataset MM-AmazonTitles-300K containing over 300K products was curated from publicly available amazon.com listings with each product endowed with a title and multiple images. On the all datasets MUFIN offered at least 3% higher accuracy than leading text-based, image-based and multi-modal techniques. Code for MUFIN is available at https://github.com/Extreme-classification/MUFIN

Edge computing aims to enable edge devices, such as IoT devices, to process data locally instead of relying on the cloud. However, deep learning techniques like computer vision and natural language processing can be computationally expensive and memory-intensive. Creating manual architectures specialized for each device is infeasible due to their varying memory and computational constraints. To address these concerns, we automate the construction of task-specific deep learning architectures optimized for device constraints through Neural Architecture Search (NAS). We present DCA-NAS, a principled method of fast neural network architecture search that incorporates edge-device constraints such as model size and floating-point operations. It incorporates weight sharing and channel bottleneck techniques to speed up the search time. Based on our experiments, we see that DCA-NAS outperforms manual architectures for similar sized models and is comparable to popular mobile architectures on various image classification datasets like CIFAR-10, CIFAR-100, and Imagenet-1k. Experiments with search spaces -- DARTS and NAS-Bench-201 show the generalization capabilities of DCA-NAS. On further evaluating our approach on Hardware-NAS-Bench, device-specific architectures with low inference latency and state-of-the-art performance were discovered., Comment: Accepted to 10th International Conference on Pattern Recognition and Machine Intelligence (PReMI) 2023

We test the hypothesis that discourse predictability influences Hindi syntactic choice. While prior work has shown that a number of factors (e.g., information status, dependency length, and syntactic surprisal) influence Hindi word order preferences, the role of discourse predictability is underexplored in the literature. Inspired by prior work on syntactic priming, we investigate how the words and syntactic structures in a sentence influence the word order of the following sentences. Specifically, we extract sentences from the Hindi-Urdu Treebank corpus (HUTB), permute the preverbal constituents of those sentences, and build a classifier to predict which sentences actually occurred in the corpus against artificially generated distractors. The classifier uses a number of discourse-based features and cognitive features to make its predictions, including dependency length, surprisal, and information status. We find that information status and LSTM-based discourse predictability influence word order choices, especially for non-canonical object-fronted orders. We conclude by situating our results within the broader syntactic priming literature., Comment: Accepted to EMNLP 2022

Word order choices during sentence production can be primed by preceding sentences. In this work, we test the DUAL MECHANISM hypothesis that priming is driven by multiple different sources. Using a Hindi corpus of text productions, we model lexical priming with an n-gram cache model and we capture more abstract syntactic priming with an adaptive neural language model. We permute the preverbal constituents of corpus sentences, and then use a logistic regression model to predict which sentences actually occurred in the corpus against artificially generated meaning-equivalent variants. Our results indicate that lexical priming and lexically-independent syntactic priming affect complementary sets of verb classes. By showing that different priming influences are separable from one another, our results support the hypothesis that multiple different cognitive mechanisms underlie priming., Comment: Accepted to AACL 2022

Extreme Classification (XC) seeks to tag data points with the most relevant subset of labels from an extremely large label set. Performing deep XC with dense, learnt representations for data points and labels has attracted much attention due to its superiority over earlier XC methods that used sparse, hand-crafted features. Negative mining techniques have emerged as a critical component of all deep XC methods that allow them to scale to millions of labels. However, despite recent advances, training deep XC models with large encoder architectures such as transformers remains challenging. This paper identifies that memory overheads of popular negative mining techniques often force mini-batch sizes to remain small and slow training down. In response, this paper introduces NGAME, a light-weight mini-batch creation technique that offers provably accurate in-batch negative samples. This allows training with larger mini-batches offering significantly faster convergence and higher accuracies than existing negative sampling techniques. NGAME was found to be up to 16% more accurate than state-of-the-art methods on a wide array of benchmark datasets for extreme classification, as well as 3% more accurate at retrieving search engine queries in response to a user webpage visit to show personalized ads. In live A/B tests on a popular search engine, NGAME yielded up to 23% gains in click-through-rates.

Scalability and accuracy are well recognized challenges in deep extreme multi-label learning where the objective is to train architectures for automatically annotating a data point with the most relevant subset of labels from an extremely large label set. This paper develops the DeepXML framework that addresses these challenges by decomposing the deep extreme multi-label task into four simpler sub-tasks each of which can be trained accurately and efficiently. Choosing different components for the four sub-tasks allows DeepXML to generate a family of algorithms with varying trade-offs between accuracy and scalability. In particular, DeepXML yields the Astec algorithm that could be 2-12% more accurate and 5-30x faster to train than leading deep extreme classifiers on publically available short text datasets. Astec could also efficiently train on Bing short text datasets containing up to 62 million labels while making predictions for billions of users and data points per day on commodity hardware. This allowed Astec to be deployed on the Bing search engine for a number of short text applications ranging from matching user queries to advertiser bid phrases to showing personalized ads where it yielded significant gains in click-through-rates, coverage, revenue and other online metrics over state-of-the-art techniques currently in production. DeepXML's code is available at https://github.com/Extreme-classification/deepxml

Extreme multi-label classification (XML) involves tagging a data point with its most relevant subset of labels from an extremely large label set, with several applications such as product-to-product recommendation with millions of products. Although leading XML algorithms scale to millions of labels, they largely ignore label meta-data such as textual descriptions of the labels. On the other hand, classical techniques that can utilize label metadata via representation learning using deep networks struggle in extreme settings. This paper develops the DECAF algorithm that addresses these challenges by learning models enriched by label metadata that jointly learn model parameters and feature representations using deep networks and offer accurate classification at the scale of millions of labels. DECAF makes specific contributions to model architecture design, initialization, and training, enabling it to offer up to 2-6% more accurate prediction than leading extreme classifiers on publicly available benchmark product-to-product recommendation datasets, such as LF-AmazonTitles-1.3M. At the same time, DECAF was found to be up to 22x faster at inference than leading deep extreme classifiers, which makes it suitable for real-time applications that require predictions within a few milliseconds. The code for DECAF is available at the following URL https://github.com/Extreme-classification/DECAF.

Deep extreme classification (XC) seeks to train deep architectures that can tag a data point with its most relevant subset of labels from an extremely large label set. The core utility of XC comes from predicting labels that are rarely seen during training. Such rare labels hold the key to personalized recommendations that can delight and surprise a user. However, the large number of rare labels and small amount of training data per rare label offer significant statistical and computational challenges. State-of-the-art deep XC methods attempt to remedy this by incorporating textual descriptions of labels but do not adequately address the problem. This paper presents ECLARE, a scalable deep learning architecture that incorporates not only label text, but also label correlations, to offer accurate real-time predictions within a few milliseconds. Core contributions of ECLARE include a frugal architecture and scalable techniques to train deep models along with label correlation graphs at the scale of millions of labels. In particular, ECLARE offers predictions that are 2 to 14% more accurate on both publicly available benchmark datasets as well as proprietary datasets for a related products recommendation task sourced from the Bing search engine. Code for ECLARE is available at https://github.com/Extreme-classification/ECLARE.

Previous work suggests that RNNs trained on natural language corpora can capture number agreement well for simple sentences but perform less well when sentences contain agreement attractors: intervening nouns between the verb and the main subject with grammatical number opposite to the latter. This suggests these models may not learn the actual syntax of agreement, but rather infer shallower heuristics such as `agree with the recent noun'. In this work, we investigate RNN models with varying inductive biases trained on selectively chosen `hard' agreement instances, i.e., sentences with at least one agreement attractor. For these the verb number cannot be predicted using a simple linear heuristic, and hence they might help provide the model additional cues for hierarchical syntax. If RNNs can learn the underlying agreement rules when trained on such hard instances, then they should generalize well to other sentences, including simpler ones. However, we observe that several RNN types, including the ONLSTM which has a soft structural inductive bias, surprisingly fail to perform well on sentences without attractors when trained solely on sentences with attractors. We analyze how these selectively trained RNNs compare to the baseline (training on a natural distribution of agreement attractors) along the dimensions of number agreement accuracy, representational similarity, and performance across different syntactic constructions. Our findings suggest that RNNs trained on our hard agreement instances still do not capture the underlying syntax of agreement, but rather tend to overfit the training distribution in a way which leads them to perform poorly on `easy' out-of-distribution instances. Thus, while RNNs are powerful models which can pick up non-trivial dependency patterns, inducing them to do so at the level of syntax rather than surface remains a challenge., Comment: 15 pages, 3 figures, 13 Tables (including Appendix); Non Archival Extended Abstract Accepted in SciL 2021 - https://scholarworks.umass.edu/scil/vol4/iss1/38/

In the last few years, compression of deep neural networks has become an important strand of machine learning and computer vision research. Deep models require sizeable computational complexity and storage, when used for instance for Human Action Recognition (HAR) from videos, making them unsuitable to be deployed on edge devices. In this paper, we address this issue and propose a method to effectively compress Recurrent Neural Networks (RNNs) such as Gated Recurrent Units (GRUs) and Long-Short-Term-Memory Units (LSTMs) that are used for HAR. We use a Variational Information Bottleneck (VIB) theory-based pruning approach to limit the information flow through the sequential cells of RNNs to a small subset. Further, we combine our pruning method with a specific group-lasso regularization technique that significantly improves compression. The proposed techniques reduce model parameters and memory footprint from latent representations, with little or no reduction in the validation accuracy while increasing the inference speed several-fold. We perform experiments on the three widely used Action Recognition datasets, viz. UCF11, HMDB51, and UCF101, to validate our approach. It is shown that our method achieves over 70 times greater compression than the nearest competitor with comparable accuracy for the task of action recognition on UCF11., Comment: Accepted at IEEE WACV 2021

Long short-term memory (LSTM) networks and their variants are capable of encapsulating long-range dependencies, which is evident from their performance on a variety of linguistic tasks. On the other hand, simple recurrent networks (SRNs), which appear more biologically grounded in terms of synaptic connections, have generally been less successful at capturing long-range dependencies as well as the loci of grammatical errors in an unsupervised setting. In this paper, we seek to develop models that bridge the gap between biological plausibility and linguistic competence. We propose a new architecture, the Decay RNN, which incorporates the decaying nature of neuronal activations and models the excitatory and inhibitory connections in a population of neurons. Besides its biological inspiration, our model also shows competitive performance relative to LSTMs on subject-verb agreement, sentence grammaticality, and language modeling tasks. These results provide some pointers towards probing the nature of the inductive biases required for RNN architectures to model linguistic phenomena successfully., Comment: 11 pages, 5 figures (including appendix); to appear at ACL SRW 2020

Visual scene understanding often requires the processing of human-object interactions. Here we seek to explore if and how well Deep Neural Network (DNN) models capture features similar to the brain's representation of humans, objects, and their interactions. We investigate brain regions which process human-, object-, or interaction-specific information, and establish correspondences between them and DNN features. Our results suggest that we can infer the selectivity of these regions to particular visual stimuli using DNN representations. We also map features from the DNN to the regions, thus linking the DNN representations to those found in specific parts of the visual cortex. In particular, our results suggest that a typical DNN representation contains encoding of compositional information for human-object interactions which goes beyond a linear combination of the encodings for the two components, thus suggesting that DNNs may be able to model this important property of biological vision., Comment: 4 pages, 2 figures; presented at CCN 2019

The study of virality and information diffusion online is a topic gaining traction rapidly in the computational social sciences. Computer vision and social network analysis research have also focused on understanding the impact of content and information diffusion in making content viral, with prior approaches not performing significantly well as other traditional classification tasks. In this paper, we present a novel pairwise reformulation of the virality prediction problem as an attribute prediction task and develop a novel algorithm to model image virality on online media using a pairwise neural network. Our model provides significant insights into the features that are responsible for promoting virality and surpasses the existing state-of-the-art by a 12% average improvement in prediction. We also investigate the effect of external category supervision on relative attribute prediction and observe an increase in prediction accuracy for the same across several attribute learning datasets., Comment: 9 pages, Accepted as a full paper at the ACM Multimedia Conference (MM) 2017

We compare several ConvNets with different depth and regularization techniques with multi-unit macaque IT cortex recordings and assess the impact of the same on representational similarity with the primate visual cortex. We find that with increasing depth and validation performance, ConvNet features are closer to cortical IT representations., Comment: 4 pages, short abstract, Accepted to the Workshop on Biological and Artificial Vision, ECCV, 2016

The idea of 'date' and 'party' hubs has been influential in the study of protein-protein interaction networks. Date hubs display low co-expression with their partners, whilst party hubs have high co-expression. It was proposed that party hubs are local coordinators whereas date hubs are global connectors. Here we show that the reported importance of date hubs to network connectivity can in fact be attributed to a tiny subset of them. Crucially, these few, extremely central, hubs do not display particularly low expression correlation, undermining the idea of a link between this quantity and hub function. The date/party distinction was originally motivated by an approximately bimodal distribution of hub co-expression; we show that this feature is not always robust to methodological changes. Additionally, topological properties of hubs do not in general correlate with co-expression. Thus, we suggest that a date/party dichotomy is not meaningful and it might be more useful to conceive of roles for protein-protein interactions rather than individual proteins. We find significant correlations between interaction centrality and the functional similarity of the interacting proteins., Comment: 27 pages, 5 main figures, 4 supplementary figures