What is he best known for?
The fields of study he is best known for:
- Artificial intelligence
- Statistics
- Mathematical analysis
Suvrit Sra mainly investigates Mathematical optimization, Artificial intelligence, Variance reduction, Algorithm and Metric.
His Mathematical optimization study combines topics from a wide range of disciplines, such as Matrix decomposition, Nonnegative matrix and Riemannian geometry.
His study in Artificial intelligence is interdisciplinary in nature, drawing from both Machine learning, Subgradient method, State and Pattern recognition.
His work carried out in the field of Pattern recognition brings together such families of science as Cluster analysis and Expectation–maximization algorithm.
The Metric study combines topics in areas such as Positive-definite matrix, Matrix, Manifold, Divergence and Nearest neighbor search.
His study on Nearest neighbor search also encompasses disciplines like
- Function that connect with fields like Kullback–Leibler divergence, Similarity learning, k-nearest neighbors algorithm, Mahalanobis distance and Feature,
- Covariance matrix that connect with fields like Similarity.
His most cited work include:
- Information-theoretic metric learning (1617 citations)
- Clustering on the Unit Hypersphere using von Mises-Fisher Distributions (634 citations)
- Optimization for Machine Learning (378 citations)
What are the main themes of his work throughout his whole career to date?
Suvrit Sra mostly deals with Mathematical optimization, Applied mathematics, Algorithm, Artificial intelligence and Positive-definite matrix.
His study looks at the intersection of Mathematical optimization and topics like Gradient descent with Stationary point.
His Applied mathematics study deals with Manifold intersecting with Stochastic optimization.
As part of the same scientific family, Suvrit Sra usually focuses on Algorithm, concentrating on Sampling and intersecting with Markov chain, Probabilistic logic and Distribution.
As a member of one scientific family, Suvrit Sra mostly works in the field of Artificial intelligence, focusing on Machine learning and, on occasion, Sample.
As a part of the same scientific family, Suvrit Sra mostly works in the field of Positive-definite matrix, focusing on Metric and, on occasion, Representation.
He most often published in these fields:
- Mathematical optimization (26.14%)
- Applied mathematics (21.59%)
- Algorithm (20.83%)
What were the highlights of his more recent work (between 2017-2021)?
- Applied mathematics (21.59%)
- Artificial intelligence (20.45%)
- Mathematical optimization (26.14%)
In recent papers he was focusing on the following fields of study:
Suvrit Sra spends much of his time researching Applied mathematics, Artificial intelligence, Mathematical optimization, Algorithm and Discrete mathematics.
His Applied mathematics research is multidisciplinary, incorporating elements of Artificial neural network, Sequence, Gradient method and Convex function.
His Artificial intelligence research is multidisciplinary, relying on both Optimization problem, Machine learning and Key.
His biological study spans a wide range of topics, including Manifold, Pooling, Reproducing kernel Hilbert space and Pattern recognition.
He interconnects Cover and Computation in the investigation of issues within Mathematical optimization.
His Discrete mathematics research incorporates elements of Supervised learning, Upper and lower bounds and Series.
Between 2017 and 2021, his most popular works were:
- Why Gradient Clipping Accelerates Training: A Theoretical Justification for Adaptivity (41 citations)
- Small nonlinearities in activation functions create bad local minima in neural networks (36 citations)
- Small ReLU networks are powerful memorizers: a tight analysis of memorization capacity (35 citations)
In his most recent research, the most cited papers focused on:
- Artificial intelligence
- Statistics
- Mathematical analysis
The scientist’s investigation covers issues in Applied mathematics, Artificial intelligence, Deep learning, Stochastic gradient descent and Sequence.
The concepts of his Applied mathematics study are interwoven with issues in Artificial neural network, Sigmoid function, Counterexample, Convex function and Gradient method.
His Artificial intelligence research includes elements of Machine learning, Key and Pattern recognition.
His work in Deep learning addresses subjects such as Saddle point, which are connected to disciplines such as Computation, Hessian matrix and Mathematical optimization.
His work in Stochastic gradient descent addresses issues such as Clipping, which are connected to fields such as Algorithm.
He has researched Condition number in several fields, including Manifold, Metric and Convex optimization.
This overview was generated by a machine learning system which analysed the scientist’s
body of work. If you have any feedback, you can
contact us
here.
