Subhradeep Kayal

Erasmus MC, Radiology & Medical Informatics, Graduate Student

Aalto University, Information and Computer Science, Graduate Student

National Institute of Technology Rourkela, Dept. of Electrical Engineering, Undergraduate

Elsevier, Content Enrichment, Senior Data Scientist

Followers

142

Following

Public Views

I am a machine learning and analytics professional with several years of relevant work experience in data analytics, algorithm development and overall product development.

Professional Experience:
I have worked for both startups and enterprises, and have been involved in a wide variety of projects. I'm familiar with both the engineering concepts related to analytics, namely SQL (MySQL) and NoSQL databases (Cassandra, MongoDB, Amazon Redshift, HBase), big data concepts (Hadoop Map-Reduce, Spark), as well as knowledge of various machine learning and data mining algorithms (Regressions, Support Vector Machines, Tree based algorithms, Neural Networks, Ensembling, Recommendation algorithms, Frequent pattern mining). I am used to coding in Java, Scala, Python, and R, and to an extent C++, and well-versed with version control systems like git and svn.
Given my experience in working with different sized teams, I am used to both working alone on my own deliverable and collaborating with teams for a common project.

Academic Experience:
Besides my professional background, I am currently a part-time researcher at Erasmus MC in Rotterdam, exploring machine learning based biomedical image analysis, after having finished my MSc degree in Machine Learning from Finland.
Till date, I have also published several research papers and patents.

less

Chen Yang

Hefei University of technology

Sench Galiedon

Hangzhou Dianzi University

Denis Newman-Griffis

University of Pittsburgh

Hakan Ferhatosmanoglu

Computer Science & Information Technology (CS & IT) Computer Science Conference Proceedings (CSCP)

Melison Dylan

Peking University

Ra S

University of Oxford

Samuel J Bell

University of Cambridge

Amit Kumar Jaiswal

University College London

Fatima Alkhawaldeh

InterestsView All (14)

Uploads

Papers by Subhradeep Kayal

EigenSent: Spectral sentence embeddings using higher-order Dynamic Mode Decomposition

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019

Distributed representation of words, or word embeddings, have motivated methods for calculating s... more Distributed representation of words, or word embeddings, have motivated methods for calculating semantic representations of word sequences such as phrases, sentences and paragraphs. Most of the existing methods to do so either use algorithms to learn such representations, or improve on calculating weighted averages of the word vectors. In this work, we experiment with spectral methods of signal representation and summarization as mechanisms for constructing such word-sequence embeddings in an unsupervised fashion. In particular, we explore an algorithm rooted in fluid-dynamics, known as higher-order Dynamic Mode Decomposition, which is designed to capture the eigenfrequencies, and hence the fundamental transition dynamics, of periodic and quasi-periodic systems. It is empirically observed that this approach, which we call EigenSent, can summarize transitions in a sequence of words and generate an embedding that can represent well the sequence itself. To the best of the authors’ knowledge, this is the first application of a spectral decomposition and signal summarization technique on text, to create sentence embeddings. We test the efficacy of this algorithm in creating sentence embeddings on three public datasets, where it performs appreciably well. Moreover it is also shown that, due to the positive combination of their complementary properties, concatenating the embeddings generated
by EigenSent with simple word vector averaging achieves state-of-the-art results.

Download

Improved Hierarchical Clustering for Face Images in Videos: Integrating positional and temporal information with HAC

Efficient techniques for face clustering, along with speech and text recognition, can provide the... more

Download

Face clustering in videos: GMM-based hierarchical clustering using Spatio-Temporal data

In recent years, an increase in multimedia data generation and efficient forms of storage have gi... more In recent years, an increase in multimedia data generation and efficient forms of storage have given rise to needs like quick browsing, efficient summarization and techniques for information retrieval. Face Clustering, together with other technologies such as speech recognition, can effectively solve these problems. Applications such as video indexing, major cast detection and video summarization greatly benefit from the development of accurate face clustering algorithms. Since videos represent a temporally ordered collection of faces, it is only natural to use the knowledge of the temporal ordering of these faces, in conjunction with the spatial features extracted from them, to obtain optimal clusterings. This paper is aimed at developing a novel clustering algorithm, by modifying the highly successful hierarchical agglomerative clustering (HAC) process, so that it includes an effective initialization mechanism, via an initial temporal clustering and Gaussian Mixture Model based cluster splitting, and introduces a temporal aspect during cluster combination, in addition to the spatial distances. Experiments show that it significantly outperforms HAC while being equally flexible.

Face verification experiments on the LFW database with simple features, metrics and classifiers

Face verification is fast becoming a dual platform for testing face recognition algorithms. Unlik... more Face verification is fast becoming a dual platform for testing face recognition algorithms. Unlike recognition, verification is non-intrusive and more practical for security applications, where it is more important to look for match or mismatch between faces, rather than the exact identity. This paper aims at providing comprehensive overwiev af the verification accuracies that can be achieved through the use of simple techniques and out-of-the-box methods. Experiments are performed with features, metrics and learning methods from different categories, i.e, global an local features, generative and discriminative models, and results are appropriately reported. The main contribution of this paper is to provide a starting point for future researchers in the field of face recognition and verification.

Face Clustering Experiments on News Video Images

With the advent of the information age and an explosion in the availability of multimedia data, s... more With the advent of the information age and an explosion in the availability of multimedia data, successful clustering of faces is absolutely essential for many applications such as video indexing and summarization, major cast detection, and even unsupervised face-database generation. This paper compares with each other, the combinations of existing approaches of dimensionality reduction and different methods of clustering on face images detected from a TV news broadcast clip. The real value of the paper is in providing future researchers an introductory overview of possible methods and recommend a method as a good starting point for face clustering from videos.

Download

Experiments on the LFW database using curvelet transforms and a random forest kNN cascade

Successful recognition of faces from unconstrained complex images is absolutely essential for man... more Successful recognition of faces from unconstrained complex images is absolutely essential for many biometrics and surveillance applications. This paper aims at exploring the use of the curvelet transform as a method of facial feature extraction and the use of the random forests as a successful classifier. The real value of this paper is its suggested use of a cascade of the random forest classifier with a nearest neighbour verifier. In this framework, the wrapping based curvelet transform is used to extract features, which are then used to train a random forest classifier. A kNN classifier (termed here as a 'verifier') is cascaded with the random forest to further correct any wrong decisions made by the random forest. On a subset of the Labeled Faces in the Wild dataset, this method performs well with an average percentage recognition of 82%.

Download