Dipankar Das

Subhabrata Dutta, Jeevesh Juneja, Dipankar Das et al.

Identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining. The intrinsic complexity of these tasks demands powerful learning models. While pretrained Transformer-based Language Models (LM) have been shown to provide state-of-the-art results over different NLP tasks, the scarcity of manually annotated data and the highly domain-dependent nature of argumentation restrict the capabilities of such models. In this work, we propose a novel transfer learning strategy to overcome these challenges. We utilize argumentation-rich social discussions from the ChangeMyView subreddit as a source of unsupervised, argumentative discourse-aware knowledge by finetuning pretrained LMs on a selectively masked language modeling task. Furthermore, we introduce a novel prompt-based strategy for inter-component relation prediction that compliments our proposed finetuning method while leveraging on the discourse context. Exhaustive experiments show the generalization capability of our method on these two tasks over within-domain as well as out-of-domain datasets, outperforming several existing and employed strong baselines.

Prantik Guha, Rudra Dhar, Dipankar Das

In this paper we describe a system submitted to the INLG 2022 Generation Challenge (GenChal) on Quality Evaluation of the Low-Resource Synthetically Generated Code-Mixed Hinglish Text. We implement a Bi-LSTM-based neural network model to predict the Average rating score and Disagreement score of the synthetic Hinglish dataset. In our models, we used word embeddings for English and Hindi data, and one hot encodings for Hinglish data. We achieved a F1 score of 0.11, and mean squared error of 6.0 in the average rating score prediction task. In the task of Disagreement score prediction, we achieve a F1 score of 0.18, and mean squared error of 5.0.

Dipankar Das, Deepak Ranjan Nayak

Glaucoma is a common eye disease that leads to irreversible blindness unless timely detected. Hence, glaucoma detection at an early stage is of utmost importance for a better treatment plan and ultimately saving the vision. The recent literature has shown the prominence of CNN-based methods to detect glaucoma from retinal fundus images. However, such methods mainly focus on solving binary classification tasks and have not been thoroughly explored for the detection of different glaucoma stages, which is relatively challenging due to minute lesion size variations and high inter-class similarities. This paper proposes a global self-attention based network called GS-Net for efficient multi-stage glaucoma classification. We introduce a global self-attention module (GSAM) consisting of two parallel attention modules, a channel attention module (CAM) and a spatial attention module (SAM), to learn global feature dependencies across channel and spatial dimensions. The GSAM encourages extracting more discriminative and class-specific features from the fundus images. The experimental results on a publicly available dataset demonstrate that our GS-Net outperforms state-of-the-art methods. Also, the GSAM achieves competitive performance against popular attention modules.

Md. Fahad Ullah Utsho, Mohd. Ruhul Ameen, Akif Islam et al.

Large Language Models (LLMs) are increasingly described as possessing strong reasoning capabilities, supported by high performance on mathematical, logical, and planning benchmarks. However, most existing evaluations rely on aggregate accuracy over fixed datasets, obscuring how reasoning behavior evolves as task complexity increases. In this work, we introduce a controlled benchmarking framework to systematically evaluate the robustness of reasoning in Large Reasoning Models (LRMs) under progressively increasing problem complexity. We construct a suite of nine classical reasoning tasks: Boolean Satisfiability, Cryptarithmetic, Graph Coloring, River Crossing, Tower of Hanoi, Water Jug, Checker Jumping, Sudoku, and Rubik's Cube, each parameterized to precisely control complexity while preserving underlying semantics. Using deterministic validators, we evaluate multiple open and proprietary LRMs across low, intermediate, and high complexity regimes, ensuring that only fully valid solutions are accepted. Our results reveal a consistent phase transition like behavior: models achieve high accuracy at low complexity but degrade sharply beyond task specific complexity thresholds. We formalize this phenomenon as reasoning collapse. Across tasks, we observe substantial accuracy declines, often exceeding 50%, accompanied by inconsistent reasoning traces, constraint violations, loss of state tracking, and confidently incorrect outputs. Increased reasoning length does not reliably improve correctness, and gains in one problem family do not generalize to others. These findings highlight the need for evaluation methodologies that move beyond static benchmarks and explicitly measure reasoning robustness under controlled complexity.

Dharma Teja Vooturi, Dhiraj Kalamkar, Dipankar Das et al.

Pretraining Large Language Models (LLMs) from scratch requires massive amount of compute. Aurora super computer is an ExaScale machine with 127,488 Intel PVC (Ponte Vechio) GPU tiles. In this work, we showcase LLM pretraining on Aurora at the scale of 1000s of GPU tiles. Towards this effort, we developed Optimus, an inhouse training library with support for standard large model training techniques. Using Optimus, we first pretrained Mula-1B, a 1 Billion dense model and Mula-7B-A1B, a 7 Billion Mixture of Experts (MoE) model from scratch on 3072 GPU tiles for the full 4 trillion tokens of the OLMoE-mix-0924 dataset. We then demonstrated model scaling by pretraining three large MoE models Mula-20B-A2B, Mula-100B-A7B, and Mula-220B-A10B till 100 Billion tokens on the same dataset. On our largest model Mula-220B-A10B, we pushed the compute scaling from 384 to 12288 GPU tiles and observed scaling efficiency of around 90% at 12288 GPU tiles. We significantly improved the runtime performance of MoE models using custom GPU kernels for expert computation, and a novel EP-Aware sharded optimizer resulting in training speedups up to 1.71x. As part of the Optimus library, we also developed a robust set of reliability and fault tolerant features to improve training stability and continuity at scale.

Arka Dutta, Agrik Majumdar, Sombrata Biswas et al.

This paper proposes a comprehensive framework for the generation of covert advertisements within Conversational AI systems, along with robust techniques for their detection. It explores how subtle promotional content can be crafted within AI-generated responses and introduces methods to identify and mitigate such covert advertising strategies. For generation (Sub-Task~1), we propose a novel framework that leverages user context and query intent to produce contextually relevant advertisements. We employ advanced prompting strategies and curate paired training data to fine-tune a large language model (LLM) for enhanced stealthiness. For detection (Sub-Task~2), we explore two effective strategies: a fine-tuned CrossEncoder (\texttt{all-mpnet-base-v2}) for direct classification, and a prompt-based reformulation using a fine-tuned \texttt{DeBERTa-v3-base} model. Both approaches rely solely on the response text, ensuring practicality for real-world deployment. Experimental results show high effectiveness in both tasks, achieving a precision of 1.0 and recall of 0.71 for ad generation, and F1-scores ranging from 0.99 to 1.00 for ad detection. These results underscore the potential of our methods to balance persuasive communication with transparency in conversational AI.

Sourya Dipta Das, Ayan Basak, Soumil Mandal et al.

Research on adversarial attacks are becoming widely popular in the recent years. One of the unexplored areas where prior research is lacking is the effect of adversarial attacks on code-mixed data. Therefore, in the present work, we have explained the first generalized framework on text perturbation to attack code-mixed classification models in a black-box setting. We rely on various perturbation techniques that preserve the semantic structures of the sentences and also obscure the attacks from the perception of a human user. The present methodology leverages the importance of a token to decide where to attack by employing various perturbation strategies. We test our strategies on various sentiment classification models trained on Bengali-English and Hindi-English code-mixed datasets, and reduce their F1-scores by nearly 51 % and 53 % respectively, which can be further reduced if a larger number of tokens are perturbed in a given sentence.

Subhabrata Dutta, Shravika Mittal, Dipankar Das et al.

The behaviour of information cascades (such as retweets) has been modelled extensively. While point process-based generative models have long been in use for estimating cascade growths, deep learning has greatly enhanced diverse feature integration. We observe two significant temporal signals in cascade data that have not been emphasized or reported to our knowledge. First, the popularity of the cascade root is known to influence cascade size strongly; but the effect falls off rapidly with time. Second, there is a measurable positive correlation between the novelty of the root content (with respect to a streaming external corpus) and the relative size of the resulting cascade. Responding to these observations, we propose GammaCas, a new cascade growth model as a parametric function of time, which combines deep influence signals from content (e.g., tweet text), network features (e.g., followers of the root user), and exogenous event sources (e.g., online news). Specifically, our model processes these signals through a customized recurrent network, whose states then provide the parameters of the cascade rate function, which is integrated over time to predict the cascade size. The network parameters are trained end-to-end using observed cascades. GammaCas outperforms seven recent and diverse baselines significantly on a large-scale dataset of retweet cascades coupled with time-aligned online news -- it beats the best baseline with an 18.98% increase in terms of Kendall's $τ$ correlation and $35.63$ reduction in Mean Absolute Percentage Error. Extensive ablation and case studies unearth interesting insights regarding retweet cascade dynamics.

Sainik Kumar Mahata, Dipankar Das, Sivaji Bandyopadhyay

Sentiment analysis has been an active area of research in the past two decades and recently, with the advent of social media, there has been an increasing demand for sentiment analysis on social media texts. Since the social media texts are not in one language and are largely code-mixed in nature, the traditional sentiment classification models fail to produce acceptable results. This paper tries to solve this very research problem and uses bi-directional LSTMs along with language tagging, to facilitate sentiment tagging of code-mixed Tamil texts that have been extracted from social media. The presented algorithm, when evaluated on the test data, garnered precision, recall, and F1 scores of 0.59, 0.66, and 0.58 respectively.

Tathagata Raha, Sainik Kumar Mahata, Dipankar Das et al.

Code-mixed texts are widespread nowadays due to the advent of social media. Since these texts combine two languages to formulate a sentence, it gives rise to various research problems related to Natural Language Processing. In this paper, we try to excavate one such problem, namely, Parts of Speech tagging of code-mixed texts. We have built a system that can POS tag English-Bengali code-mixed data where the Bengali words were written in Roman script. Our approach initially involves the collection and cleaning of English-Bengali code-mixed tweets. These tweets were used as a development dataset for building our system. The proposed system is a modular approach that starts by tagging individual tokens with their respective languages and then passes them to different POS taggers, designed for different languages (English and Bengali, in our case). Tags given by the two systems are later joined together and the final result is then mapped to a universal POS tag set. Our system was checked using 100 manually POS tagged code-mixed sentences and it returned an accuracy of 75.29%

Sainik Kumar Mahata, Amrita Chandra, Dipankar Das et al.

In the current work, we explore the enrichment in the machine translation output when the training parallel corpus is augmented with the introduction of sentiment analysis. The paper discusses the preparation of the same sentiment tagged English-Bengali parallel corpus. The preparation of raw parallel corpus, sentiment analysis of the sentences and the training of a Character Based Neural Machine Translation model using the same has been discussed extensively in this paper. The output of the translation model has been compared with a base-line translation model using automated metrics such as BLEU and TER as well as manually.

Avishek Garain, Sainik Kumar Mahata, Dipankar Das

Code-mixing is a phenomenon which arises mainly in multilingual societies. Multilingual people, who are well versed in their native languages and also English speakers, tend to code-mix using English-based phonetic typing and the insertion of anglicisms in their main language. This linguistic phenomenon poses a great challenge to conventional NLP domains such as Sentiment Analysis, Machine Translation, and Text Summarization, to name a few. In this work, we focus on working out a plausible solution to the domain of Code-Mixed Sentiment Analysis. This work was done as participation in the SemEval-2020 Sentimix Task, where we focused on the sentiment analysis of English-Hindi code-mixed sentences. our username for the submission was "sainik.mahata" and team name was "JUNLP". We used feature extraction algorithms in conjunction with traditional machine learning algorithms such as SVR and Grid Search in an attempt to solve the task. Our approach garnered an f1-score of 66.2\% when tested using metrics prepared by the organizers of the task.

Prashant Kapil, Asif Ekbal, Dipankar Das

The phenomenal growth on the internet has helped in empowering individual's expressions, but the misuse of freedom of expression has also led to the increase of various cyber crimes and anti-social activities. Hate speech is one such issue that needs to be addressed very seriously as otherwise, this could pose threats to the integrity of the social fabrics. In this paper, we proposed deep learning approaches utilizing various embeddings for detecting various types of hate speeches in social media. Detecting hate speech from a large volume of text, especially tweets which contains limited contextual information also poses several practical challenges. Moreover, the varieties in user-generated data and the presence of various forms of hate speech makes it very challenging to identify the degree and intention of the message. Our experiments on three publicly available datasets of different domains shows a significant improvement in accuracy and F1-score.

Sainik Kumar Mahata, Soumil Mandal, Dipankar Das et al.

The use of multilingualism in the new generation is widespread in the form of code-mixed data on social media, and therefore a robust translation system is required for catering to the monolingual users, as well as for easier comprehension by language processing models. In this work, we present a translation framework that uses a translation-transliteration strategy for translating code-mixed data into their equivalent monolingual instances. For converting the output to a more fluent form, it is reordered using a target language model. The most important advantage of the proposed framework is that it does not require a code-mixed to monolingual parallel corpus at any point. On testing the framework, it achieved BLEU and TER scores of 16.47 and 55.45, respectively. Since the proposed framework deals with various sub-modules, we dive deeper into the importance of each of them, analyze the errors and finally, discuss some improvement strategies.

Abhisek Kundu, Alex Heinecke, Dhiraj Kalamkar et al.

We propose K-TanH, a novel, highly accurate, hardware efficient approximation of popular activation function TanH for Deep Learning. K-TanH consists of parameterized low-precision integer operations, such as, shift and add/subtract (no floating point operation needed) where parameters are stored in very small look-up tables that can fit in CPU registers. K-TanH can work on various numerical formats, such as, Float32 and BFloat16. High quality approximations to other activation functions, e.g., Sigmoid, Swish and GELU, can be derived from K-TanH. Our AVX512 implementation of K-TanH demonstrates $>5\times$ speed up over Intel SVML, and it is consistently superior in efficiency over other approximations that use floating point arithmetic. Finally, we achieve state-of-the-art Bleu score and convergence results for training language translation model GNMT on WMT16 data sets with approximate TanH obtained via K-TanH on BFloat16 inputs.

Sudarshan Srinivasan, Pradeep Janedula, Saurabh Dhoble et al.

Low-precision is the first order knob for achieving higher Artificial Intelligence Operations (AI-TOPS). However the algorithmic space for sub-8-bit precision compute is diverse, with disruptive changes happening frequently, making FPGAs a natural choice for Deep Neural Network inference, In this work we present an FPGA-based accelerator for CNN inference acceleration. We use {\it INT-8-2} compute (with {\it 8 bit} activation and {2 bit} weights) which is recently showing promise in the literature, and which no known ASIC, CPU or GPU natively supports today. Using a novel Adaptive Logic Module (ALM) based design, as a departure from traditional DSP based designs, we are able to achieve high performance measurement of 5 AI-TOPS for {\it Arria10} and project a performance of 76 AI-TOPS at 0.7 TOPS/W for {\it Stratix10}. This exceeds known CPU, GPU performance and comes close to best known ASIC (TPU) numbers, while retaining the versatility of the FPGA platform for other applications.

Subhabrata Dutta, Dipankar Das, Tanmoy Chakraborty

Online discussions are valuable resources to study user behaviour on a diverse set of topics. Unlike previous studies which model a discussion in a static manner, in the present study, we model it as a time-varying process and solve two inter-related problems -- predict which user groups will get engaged with an ongoing discussion, and forecast the growth rate of a discussion in terms of the number of comments. We propose RGNet (Relativistic Gravitational Nerwork), a novel algorithm that uses Einstein Field Equations of gravity to model online discussions as `cloud of dust' hovering over a user spacetime manifold, attracting users of different groups at different rates over time. We also propose GUVec, a global user embedding method for an online discussion, which is used by RGNet to predict temporal user engagement. RGNet leverages different textual and network-based features to learn the dust distribution for discussions. We employ four baselines -- first two using LSTM architecture, third one using Newtonian model of gravity, and fourth one using a logistic regression adopted from a previous work on engagement prediction. Experiments on Reddit dataset show that RGNet achieves 0.72 Micro F1 score and 6.01% average error for temporal engagement prediction of user groups and growth rate forecasting, respectively, outperforming all the baselines significantly. We further employ RGNet to predict non-temporal engagement -- whether users will comment to a given post or not. RGNet achieves 0.62 AUC for this task, outperforming existing baseline by 8.77% AUC.

Sainik Kumar Mahata, Avishek Garain, Adityar Rayala et al.

In the current work, we present a description of the system submitted to WMT 2019 News Translation Shared task. The system was created to translate news text from Lithuanian to English. To accomplish the given task, our system used a Word Embedding based Neural Machine Translation model to post edit the outputs generated by a Statistical Machine Translation model. The current paper documents the architecture of our model, descriptions of the various modules and the results produced using the same. Our system garnered a BLEU score of 17.6.

Sainik Kumar Mahata, Dipankar Das, Sivaji Bandyopadhyay

In the current work, we present a description of the system submitted to WMT 2018 News Translation Shared task. The system was created to translate news text from Finnish to English. The system used a Character Based Neural Machine Translation model to accomplish the given task. The current paper documents the preprocessing steps, the description of the submitted system and the results produced using the same. Our system garnered a BLEU score of 12.9.

Naveen Mellempudi, Sudarshan Srinivasan, Dipankar Das et al.

Reduced precision computation for deep neural networks is one of the key areas addressing the widening compute gap driven by an exponential growth in model size. In recent years, deep learning training has largely migrated to 16-bit precision, with significant gains in performance and energy efficiency. However, attempts to train DNNs at 8-bit precision have met with significant challenges because of the higher precision and dynamic range requirements of back-propagation. In this paper, we propose a method to train deep neural networks using 8-bit floating point representation for weights, activations, errors, and gradients. In addition to reducing compute precision, we also reduced the precision requirements for the master copy of weights from 32-bit to 16-bit. We demonstrate state-of-the-art accuracy across multiple data sets (imagenet-1K, WMT16) and a broader set of workloads (Resnet-18/34/50, GNMT, Transformer) than previously reported. We propose an enhanced loss scaling method to augment the reduced subnormal range of 8-bit floating point for improved error propagation. We also examine the impact of quantization noise on generalization and propose a stochastic rounding technique to address gradient noise. As a result of applying all these techniques, we report slightly higher validation accuracy compared to full precision baseline.

Dhiraj Kalamkar, Dheevatsa Mudigere, Naveen Mellempudi et al.

This paper presents the first comprehensive empirical study demonstrating the efficacy of the Brain Floating Point (BFLOAT16) half-precision format for Deep Learning training across image classification, speech recognition, language modeling, generative networks and industrial recommendation systems. BFLOAT16 is attractive for Deep Learning training for two reasons: the range of values it can represent is the same as that of IEEE 754 floating-point format (FP32) and conversion to/from FP32 is simple. Maintaining the same range as FP32 is important to ensure that no hyper-parameter tuning is required for convergence; e.g., IEEE 754 compliant half-precision floating point (FP16) requires hyper-parameter tuning. In this paper, we discuss the flow of tensors and various key operations in mixed precision training, and delve into details of operations, such as the rounding modes for converting FP32 tensors to BFLOAT16. We have implemented a method to emulate BFLOAT16 operations in Tensorflow, Caffe2, IntelCaffe, and Neon for our experiments. Our results show that deep learning training using BFLOAT16 tensors achieves the same state-of-the-art (SOTA) results across domains as FP32 tensors in the same number of iterations and with no changes to hyper-parameters.

Sainik Kumar Mahata, Soumil Mandal, Dipankar Das et al.

In the present article, we identified the qualitative differences between Statistical Machine Translation (SMT) and Neural Machine Translation (NMT) outputs. We have tried to answer two important questions: 1. Does NMT perform equivalently well with respect to SMT and 2. Does it add extra flavor in improving the quality of MT output by employing simple sentences as training units. In order to obtain insights, we have developed three core models viz., SMT model based on Moses toolkit, followed by character and word level NMT models. All of the systems use English-Hindi and English-Bengali language pairs containing simple sentences as well as sentences of other complexity. In order to preserve the translations semantics with respect to the target words of a sentence, we have employed soft-attention into our word level NMT model. We have further evaluated all the systems with respect to the scenarios where they succeed and fail. Finally, the quality of translation has been validated using BLEU and TER metrics along with manual parameters like fluency, adequacy etc. We observed that NMT outperforms SMT in case of simple sentences whereas SMT outperforms in case of all types of sentence.

Apoorv Vyas, Nataraj Jammalamadaka, Xia Zhu et al.

As deep learning methods form a critical part in commercially important applications such as autonomous driving and medical diagnostics, it is important to reliably detect out-of-distribution (OOD) inputs while employing these algorithms. In this work, we propose an OOD detection algorithm which comprises of an ensemble of classifiers. We train each classifier in a self-supervised manner by leaving out a random subset of training data as OOD data and the rest as in-distribution (ID) data. We propose a novel margin-based loss over the softmax output which seeks to maintain at least a margin $m$ between the average entropy of the OOD and in-distribution samples. In conjunction with the standard cross-entropy loss, we minimize the novel loss to train an ensemble of classifiers. We also propose a novel method to combine the outputs of the ensemble of classifiers to obtain OOD detection score and class prediction. Overall, our method convincingly outperforms Hendrycks et al.[7] and the current state-of-the-art ODIN[13] on several OOD detection benchmarks.

Subhabrata Dutta, Tanmoy Chakraborty, Dipankar Das

We propose a novel attention based hierarchical LSTM model to classify discourse act sequences in social media conversations, aimed at mining data from online discussion using textual meanings beyond sentence level. The very uniqueness of the task is the complete categorization of possible pragmatic roles in informal textual discussions, contrary to extraction of question-answers, stance detection or sarcasm identification which are very much role specific tasks. Early attempt was made on a Reddit discussion dataset. We train our model on the same data, and present test results on two different datasets, one from Reddit and one from Facebook. Our proposed model outperformed the previous one in terms of domain independence; without using platform-dependent structural features, our hierarchical LSTM with word relevance attention mechanism achieved F1-scores of 71\% and 66\% respectively to predict discourse roles of comments in Reddit and Facebook discussions. Efficiency of recurrent and convolutional architectures in order to learn discursive representation on the same task has been presented and analyzed, with different word and comment embedding schemes. Our attention mechanism enables us to inquire into relevance ordering of text segments according to their roles in discourse. We present a human annotator experiment to unveil important observations about modeling and data annotation. Equipped with our text-based discourse identification model, we inquire into how heterogeneous non-textual features like location, time, leaning of information etc. play their roles in charaterizing online discussions on Facebook.

Braja Gopal Patra, Dipankar Das, Amitava Das

Sentiment analysis is essential in many real-world applications such as stance detection, review analysis, recommendation system, and so on. Sentiment analysis becomes more difficult when the data is noisy and collected from social media. India is a multilingual country; people use more than one languages to communicate within themselves. The switching in between the languages is called code-switching or code-mixing, depending upon the type of mixing. This paper presents overview of the shared task on sentiment analysis of code-mixed data pairs of Hindi-English and Bengali-English collected from the different social media platform. The paper describes the task, dataset, evaluation, baseline and participant's systems.

Soumil Mandal, Sainik Kumar Mahata, Dipankar Das

Analysis of informative contents and sentiments of social users has been attempted quite intensively in the recent past. Most of the systems are usable only for monolingual data and fails or gives poor results when used on data with code-mixing property. To gather attention and encourage researchers to work on this crisis, we prepared gold standard Bengali-English code-mixed data with language and polarity tag for sentiment analysis purposes. In this paper, we discuss the systems we prepared to collect and filter raw Twitter data. In order to reduce manual work while annotation, hybrid systems combining rule based and supervised models were developed for both language and sentiment tagging. The final corpus was annotated by a group of annotators following a few guidelines. The gold standard corpus thus obtained has impressive inter-annotator agreement obtained in terms of Kappa values. Various metrics like Code-Mixed Index (CMI), Code-Mixed Factor (CF) along with various aspects (language and emotion) also qualitatively polled the code-mixed and sentiment properties of the corpus.

Soumil Mandal, Sourya Dipta Das, Dipankar Das

Language identification of social media text still remains a challenging task due to properties like code-mixing and inconsistent phonetic transliterations. In this paper, we present a supervised learning approach for language identification at the word level of low resource Bengali-English code-mixed data taken from social media. We employ two methods of word encoding, namely character based and root phone based to train our deep LSTM models. Utilizing these two models we created two ensemble models using stacking and threshold technique which gave 91.78% and 92.35% accuracies respectively on our testing data.

Dipankar Das, Naveen Mellempudi, Dheevatsa Mudigere et al.

The state-of-the-art (SOTA) for mixed precision training is dominated by variants of low precision floating point operations, and in particular, FP16 accumulating into FP32 Micikevicius et al. (2017). On the other hand, while a lot of research has also happened in the domain of low and mixed-precision Integer training, these works either present results for non-SOTA networks (for instance only AlexNet for ImageNet-1K), or relatively small datasets (like CIFAR-10). In this work, we train state-of-the-art visual understanding neural networks on the ImageNet-1K dataset, with Integer operations on General Purpose (GP) hardware. In particular, we focus on Integer Fused-Multiply-and-Accumulate (FMA) operations which take two pairs of INT16 operands and accumulate results into an INT32 output.We propose a shared exponent representation of tensors and develop a Dynamic Fixed Point (DFP) scheme suitable for common neural network operations. The nuances of developing an efficient integer convolution kernel is examined, including methods to handle overflow of the INT32 accumulator. We implement CNN training for ResNet-50, GoogLeNet-v1, VGG-16 and AlexNet; and these networks achieve or exceed SOTA accuracy within the same number of iterations as their FP32 counterparts without any change in hyper-parameters and with a 1.8X improvement in end-to-end training throughput. To the best of our knowledge these results represent the first INT16 training results on GP hardware for ImageNet-1K dataset using SOTA CNNs and achieve highest reported accuracy using half-precision

Srinivas Sridharan, Karthikeyan Vaidyanathan, Dhiraj Kalamkar et al.

The exponential growth in use of large deep neural networks has accelerated the need for training these deep neural networks in hours or even minutes. This can only be achieved through scalable and efficient distributed training, since a single node/card cannot satisfy the compute, memory, and I/O requirements of today's state-of-the-art deep neural networks. However, scaling synchronous Stochastic Gradient Descent (SGD) is still a challenging problem and requires continued research/development. This entails innovations spanning algorithms, frameworks, communication libraries, and system design. In this paper, we describe the philosophy, design, and implementation of Intel Machine Learning Scalability Library (MLSL) and present proof-points demonstrating scaling DL training on 100s to 1000s of nodes across Cloud and HPC systems.

Soumil Mandal, Dipankar Das

Multilingual speakers often switch between languages to express themselves on social communication platforms. Sometimes, the original script of the language is preserved, while using a common script for all the languages is quite popular as well due to convenience. On such occasions, multiple languages are being mixed with different rules of grammar, using the same script which makes it a challenging task for natural language processing even in case of accurate sentiment identification. In this paper, we report results of various experiments carried out on movie reviews dataset having this code-mixing property of two languages, English and Bengali, both typed in Roman script. We have tested various machine learning algorithms trained only on English features on our code-mixed data and have achieved the maximum accuracy of 59.00% using Naive Bayes (NB) model. We have also tested various models trained on code-mixed data, as well as English features and the highest accuracy of 72.50% was obtained by a Support Vector Machine (SVM) model. Finally, we have analyzed the misclassified snippets and have discussed the challenges needed to be resolved for better accuracy.

Anirban Santara, Abhishek Naik, Balaraman Ravindran et al.

Imitation learning algorithms learn viable policies by imitating an expert's behavior when reward signals are not available. Generative Adversarial Imitation Learning (GAIL) is a state-of-the-art algorithm for learning policies when the expert's behavior is available as a fixed set of trajectories. We evaluate in terms of the expert's cost function and observe that the distribution of trajectory-costs is often more heavy-tailed for GAIL-agents than the expert at a number of benchmark continuous-control tasks. Thus, high-cost trajectories, corresponding to tail-end events of catastrophic failure, are more likely to be encountered by the GAIL-agents than the expert. This makes the reliability of GAIL-agents questionable when it comes to deployment in risk-sensitive applications like robotic surgery and autonomous driving. In this work, we aim to minimize the occurrence of tail-end events by minimizing tail risk within the GAIL framework. We quantify tail risk by the Conditional-Value-at-Risk (CVaR) of trajectories and develop the Risk-Averse Imitation Learning (RAIL) algorithm. We observe that the policies learned with RAIL show lower tail-end risk than those of vanilla GAIL. Thus the proposed RAIL algorithm appears as a potent alternative to GAIL for improved reliability in risk-sensitive applications.

Abhisek Kundu, Kunal Banerjee, Naveen Mellempudi et al.

Sub-8-bit representation of DNNs incur some discernible loss of accuracy despite rigorous (re)training at low-precision. Such loss of accuracy essentially makes them equivalent to a much shallower counterpart, diminishing the power of being deep networks. To address this problem of accuracy drop we introduce the notion of \textit{residual networks} where we add more low-precision edges to sensitive branches of the sub-8-bit network to compensate for the lost accuracy. Further, we present a perturbation theory to identify such sensitive edges. Aided by such an elegant trade-off between accuracy and compute, the 8-2 model (8-bit activations, ternary weights), enhanced by ternary residual edges, turns out to be sophisticated enough to achieve very high accuracy ($\sim 1\%$ drop from our FP-32 baseline), despite $\sim 1.6\times$ reduction in model size, $\sim 26\times$ reduction in number of multiplications, and potentially $\sim 2\times$ power-performance gain comparing to 8-8 representation, on the state-of-the-art deep network ResNet-101 pre-trained on ImageNet dataset. Moreover, depending on the varying accuracy requirements in a dynamic environment, the deployed low-precision model can be upgraded/downgraded on-the-fly by partially enabling/disabling residual connections. For example, disabling the least important residual connections in the above enhanced network, the accuracy drop is $\sim 2\%$ (from FP32), despite $\sim 1.9\times$ reduction in model size, $\sim 32\times$ reduction in number of multiplications, and potentially $\sim 2.3\times$ power-performance gain comparing to 8-8 representation. Finally, all the ternary connections are sparse in nature, and the ternary residual conversion can be done in a resource-constraint setting with no low-precision (re)training.

Souvick Ghosh, Dipankar Das, Tanmoy Chakraborty

Whenever human beings interact with each other, they exchange or express opinions, emotions, and sentiments. These opinions can be expressed in text, speech or images. Analysis of these sentiments is one of the popular research areas of present day researchers. Sentiment analysis, also known as opinion mining tries to identify or classify these sentiments or opinions into two broad categories - positive and negative. In recent years, the scientific community has taken a lot of interest in analyzing sentiment in textual data available in various social media platforms. Much work has been done on social media conversations, blog posts, newspaper articles and various narrative texts. However, when it comes to identifying emotions from scientific papers, researchers have faced some difficulties due to the implicit and hidden nature of opinion. By default, citation instances are considered inherently positive in emotion. Popular ranking and indexing paradigms often neglect the opinion present while citing. In this paper, we have tried to achieve three objectives. First, we try to identify the major sentiment in the citation text and assign a score to the instance. We have used a statistical classifier for this purpose. Secondly, we have proposed a new index (we shall refer to it hereafter as M-index) which takes into account both the quantitative and qualitative factors while scoring a paper. Thirdly, we developed a ranking of research papers based on the M-index. We also try to explain how the M-index impacts the ranking of scientific papers.

Souvick Ghosh, Satanu Ghosh, Dipankar Das

Sentiment analysis is the Natural Language Processing (NLP) task dealing with the detection and classification of sentiments in texts. While some tasks deal with identifying the presence of sentiment in the text (Subjectivity analysis), other tasks aim at determining the polarity of the text categorizing them as positive, negative and neutral. Whenever there is a presence of sentiment in the text, it has a source (people, group of people or any entity) and the sentiment is directed towards some entity, object, event or person. Sentiment analysis tasks aim to determine the subject, the target and the polarity or valence of the sentiment. In our work, we try to automatically extract sentiment (positive or negative) from Facebook posts using a machine learning approach.While some works have been done in code-mixed social media data and in sentiment analysis separately, our work is the first attempt (as of now) which aims at performing sentiment analysis of code-mixed social media text. We have used extensive pre-processing to remove noise from raw text. Multilayer Perceptron model has been used to determine the polarity of the sentiment. We have also developed the corpus for this task by manually labeling Facebook posts with their associated sentiments.

Souvick Ghosh, Satanu Ghosh, Dipankar Das

An evaluation metric is an absolute necessity for measuring the performance of any system and complexity of any data. In this paper, we have discussed how to determine the level of complexity of code-mixed social media texts that are growing rapidly due to multilingual interference. In general, texts written in multiple languages are often hard to comprehend and analyze. At the same time, in order to meet the demands of analysis, it is also necessary to determine the complexity of a particular document or a text segment. Thus, in the present paper, we have discussed the existing metrics for determining the code-mixing complexity of a corpus, their advantages, and shortcomings as well as proposed several improvements on the existing metrics. The new index better reflects the variety and complexity of a multilingual document. Also, the index can be applied to a sentence and seamlessly extended to a paragraph or an entire document. We have employed two existing code-mixed corpora to suit the requirements of our study.

Naveen Mellempudi, Abhisek Kundu, Dheevatsa Mudigere et al.

We propose a novel fine-grained quantization (FGQ) method to ternarize pre-trained full precision models, while also constraining activations to 8 and 4-bits. Using this method, we demonstrate a minimal loss in classification accuracy on state-of-the-art topologies without additional training. We provide an improved theoretical formulation that forms the basis for a higher quality solution using FGQ. Our method involves ternarizing the original weight tensor in groups of $N$ weights. Using $N=4$, we achieve Top-1 accuracy within $3.7\%$ and $4.2\%$ of the baseline full precision result for Resnet-101 and Resnet-50 respectively, while eliminating $75\%$ of all multiplications. These results enable a full 8/4-bit inference pipeline, with best-reported accuracy using ternary weights on ImageNet dataset, with a potential of $9\times$ improvement in performance. Also, for smaller networks like AlexNet, FGQ achieves state-of-the-art results. We further study the impact of group size on both performance and accuracy. With a group size of $N=64$, we eliminate $\approx99\%$ of the multiplications; however, this introduces a noticeable drop in accuracy, which necessitates fine tuning the parameters at lower precision. We address this by fine-tuning Resnet-50 with 8-bit activations and ternary weights at $N=64$, improving the Top-1 accuracy to within $4\%$ of the full precision result with $<30\%$ additional training overhead. Our final quantized model can run on a full 8-bit compute pipeline using 2-bit weights and has the potential of up to $15\times$ improvement in performance compared to baseline full-precision models.

Naveen Mellempudi, Abhisek Kundu, Dipankar Das et al.

We propose a cluster-based quantization method to convert pre-trained full precision weights into ternary weights with minimal impact on the accuracy. In addition, we also constrain the activations to 8-bits thus enabling sub 8-bit full integer inference pipeline. Our method uses smaller clusters of N filters with a common scaling factor to minimize the quantization loss, while also maximizing the number of ternary operations. We show that with a cluster size of N=4 on Resnet-101, can achieve 71.8% TOP-1 accuracy, within 6% of the best full precision results while replacing ~85% of all multiplications with 8-bit accumulations. Using the same method with 4-bit weights achieves 76.3% TOP-1 accuracy which within 2% of the full precision result. We also study the impact of the size of the cluster on both performance and accuracy, larger cluster sizes N=64 can replace ~98% of the multiplications with ternary operations but introduces significant drop in accuracy which necessitates fine tuning the parameters with retraining the network at lower precision. To address this we have also trained low-precision Resnet-50 with 8-bit activations and ternary weights by pre-initializing the network with full precision weights and achieve 68.9% TOP-1 accuracy within 4 additional epochs. Our final quantized model can run on a full 8-bit compute pipeline, with a potential 16x improvement in performance compared to baseline full-precision models.

Promita Maitra, Souvick Ghosh, Dipankar Das

Authorship attribution, being an important problem in many areas in-cluding information retrieval, computational linguistics, law and journalism etc., has been identified as a subject of increasingly research interest in the re-cent years. In case of Author Identification task in PAN at CLEF 2015, the main focus was given on cross-genre and cross-topic author verification tasks. We have used several word-based and style-based features to identify the dif-ferences between the known and unknown problems of one given set and label the unknown ones accordingly using a Random Forest based classifier.

Satanu Ghosh, Souvick Ghosh, Dipankar Das

This paper describes our approach on Query Word Labeling as an attempt in the shared task on Mixed Script Information Retrieval at Forum for Information Retrieval Evaluation (FIRE) 2015. The query is written in Roman script and the words were in English or transliterated from Indian regional languages. A total of eight Indian languages were present in addition to English. We also identified the Named Entities and special symbols as part of our task. A CRF based machine learning framework was used for labeling the individual words with their corresponding language labels. We used a dictionary based approach for language identification. We also took into account the context of the word while identifying the language. Our system demonstrated an overall accuracy of 75.5% for token level language identification. The strict F-measure scores for the identification of token level language labels for Bengali, English and Hindi are 0.7486, 0.892 and 0.7972 respectively. The overall weighted F-measure of our system was 0.7498.

Dipankar Das, Sasikanth Avancha, Dheevatsa Mudigere et al.

We design and implement a distributed multinode synchronous SGD algorithm, without altering hyper parameters, or compressing data, or altering algorithmic behavior. We perform a detailed analysis of scaling, and identify optimal design points for different networks. We demonstrate scaling of CNNs on 100s of nodes, and present what we believe to be record training throughputs. A 512 minibatch VGG-A CNN training run is scaled 90X on 128 nodes. Also 256 minibatch VGG-A and OverFeat-FAST networks are scaled 53X and 42X respectively on a 64 node cluster. We also demonstrate the generality of our approach via best-in-class 6.5X scaling for a 7-layer DNN on 16 nodes. Thereafter we attempt to democratize deep-learning by training on an Ethernet based AWS cluster and show ~14X scaling on 16 nodes.

Tanmoy Chakraborty, Dipankar Das, Sivaji Bandyopadhyay

One of the key issues in both natural language understanding and generation is the appropriate processing of Multiword Expressions (MWEs). MWEs pose a huge problem to the precise language processing due to their idiosyncratic nature and diversity in lexical, syntactical and semantic properties. The semantics of a MWE cannot be expressed after combining the semantics of its constituents. Therefore, the formalism of semantic clustering is often viewed as an instrument for extracting MWEs especially for resource constraint languages like Bengali. The present semantic clustering approach contributes to locate clusters of the synonymous noun tokens present in the document. These clusters in turn help measure the similarity between the constituent words of a potentially candidate phrase using a vector space model and judge the suitability of this phrase to be a MWE. In this experiment, we apply the semantic clustering approach for noun-noun bigram MWEs, though it can be extended to any types of MWEs. In parallel, the well known statistical models, namely Point-wise Mutual Information (PMI), Log Likelihood Ratio (LLR), Significance function are also employed to extract MWEs from the Bengali corpus. The comparative evaluation shows that the semantic clustering approach outperforms all other competing statistical models. As a by-product of this experiment, we have started developing a standard lexicon in Bengali that serves as a productive Bengali linguistic thesaurus.