Sanjeev arora born january 1968 is an indian american theoretical computer scientist who is. The past few years have seen a dramatic increase in the performance of recognition systems thanks to the introduction of deep. Algorithms and complexity dover books on computer science. What are the best nonintroductory books for deep learning. Provable bounds for learning some deep representations. Cos597g fall 2018 theoretical foundations of deep learning.
Princeton prof details mysteries of machine learning at. For several years now i am most interested in developing new theory for machine learning including deep learning. I am a member of the groups in theoretical computer science and theoretical machine learning. Computational complexity see my book on this topic, probabilistically checkable proofs pcps. Toward theoretical understanding of deep learning sanjeev arora. As part of the 201718 theoretical machine learning lecture series at ias, visiting professor in the school of mathematics sanjeev arora. It is based upon a novel idea of observing correlations among features and using these to infer the underlying edge structure via a global graph recovery.
List of computer science publications by sanjeev arora. We survey progress in recent years toward developing a theory of deep learning. Sanjeev arora 407 cs building 6092583869 arora at the domain name cs. Sanjeev arora princeton university and institute for. This paper suggests that, sometimes, increasing depth can speed up optimization.
Implicit acceleration by overparameterization sanjeev arora1 2 nadav cohen2 elad hazan1 3 abstract conventional wisdom in deep learning states that increasing depth improves expressiveness but complicates optimization. The next evolution in artificial intelligence may be a matter of dispensing with all the probabilistic tricks of deep learning. Goodreads members who liked computational complexity also liked. An exponential learning rate schedule for deep learning. Sanjeev arora, princeton university, new jersey this text gives a clear exposition of important algorithmic problems in unsupervised machine learning including nonnegative matrix factorization, topic modeling, tensor decomposition, matrix completion, compressed sensing, and mixture model learning. Sanjeev arora, aditya bhaskara, rong ge, tengyu ma submitted on 23 oct 20. Paper 1 by agrarwal et al and paper 2 by carmon et al. Fitzmorris professor of computer science at princeton university, and his research interests include computational complexity theory, uses of randomness in computation. Matus telgarskys deep learning course is possibly the most. Resources for deep reinforcement learning yuxi li medium.
Facebooks ai guru lecun imagines ais next frontier. Find books like computational complexity from the worlds largest community of readers. He is a coauthor with boaz barak of the book computational complexity. Fitzmorris professor of computer science at princeton university. The list of speakers is really fantastic, the likes of which included yoshua bengio, sanjeev arora, jimmy ba, hugo larochelle, sanja fidler, been kim and so on. Some provable bounds for deep learning sanjeev arora. A modern approach 1 by sanjeev arora, boaz barak isbn.
Sanjeev arora born january 1968 is an indian american theoretical computer scientist who is best known for his work on probabilistically checkable proofs and, in particular, the pcp theorem. This is a collection of resources for deep reinforcement learning, including the following sections. Sanjeev arora september 30, 20 deep learning, a modern version of neural nets, is increasingly seen as a promising way to implement ai tasks such as speech recognition and image recognition. We study the implicit regularization of gradient descent. Sanjeev arora, a princeton university computer science professor, gives a lecture about the theoretical understanding of deep learning in science center hall d. Efforts to understand the generalization mystery in deep learning have led to the belief that gradientbased optimization induces a form of implicit regularization, a bias towards models of low complexity.
25 873 1161 900 554 259 403 362 949 935 1042 131 1473 1120 1385 1312 480 1284 628 541 376 579 1227 393 267 35 381 2 1149 969 1082 275 1236