BAR-ILAN UNIVERSITYDEPT OF COMPUTER SCIENCE
.01

ABOUT

PERSONAL DETAILS
Deep Learning and Evolutionary Computation Researcher at Bar-Ilan University. Co-Founder and CTO of Deep Learning Based Company.

BIO

ABOUT

Dr. Eli David is a leading expert in the field of computational intelligence, specializing in deep learning (neural networks) and evolutionary computation. He has published more than thirty papers in leading artificial intelligence journals and conferences, mostly focusing on applications of deep learning and genetic algorithms in various real-world domains. For the past ten years, he has been teaching courses on deep learning and evolutionary computation at Bar-Ilan University, in addition to supervising the research of graduate students in these fields. He has also served in numerous capacities successfully designing, implementing, and leading deep learning based projects in real-world environments.

Dr. David is the developer of Falcon, a grandmaster-level chess playing program, which automatically learns by processing datasets of chess games. The program reached the second place in World Computer Speed Chess Championship 2008 relying solely on machine learning for its performance. He received the Best Paper Award in 2008 Genetic and Evolutionary Computation Conference, the Gold Award in the prestigious "Humies" Awards for Human-Competitive Results in 2014, and recently the Best Paper Award in 2016 International Conference on Artificial Neural Networks.
.02

PUBLICATIONS

PUBLICATIONS LIST
Note: My academic publications appear under the name "Omid E. David" (middle name used as first name, and vice versa). This misnomer persists for backward compatibility.
2016 Sep

DeepChess: End-to-End Deep Neural Network for Automatic Learning in Chess

International Conference on Artificial Neural Networks (ICANN '16)

Winner of Best Paper Award in ICANN '16
Springer LNCS, Vol. 9887, pp. 88-96, Barcelona, Spain, 2016.

Conferences Selected O.E. David and N.S. Netanyahu

DeepChess: End-to-End Deep Neural Network for Automatic Learning in Chess

O.E. David and N.S. Netanyahu Conferences Selected

Winner of Best Paper Award in ICANN ’16

We present an end-to-end learning method for chess, relying on deep neural networks. Without any a priori knowledge, in particular without any knowledge regarding the rules of chess, a deep neural network is trained using a combination of unsupervised pretraining and supervised training. The unsupervised training extracts high level features from a given position, and the supervised training learns to compare two chess positions and select the more favorable one. The training relies entirely on datasets of several million chess games, and no further domain specific knowledge is incorporated. The experiments show that the resulting neural network (referred to as DeepChess) is on a par with state-of-the-art chess playing programs, which have been developed through many years of manual feature selection and tuning. DeepChess is the first end-to-end machine learning-based method that results in a grandmaster-level chess playing performance.

2016 Sep

DeepPainter: Painter Classification Using Deep Convolutional Autoencoders

International Conference on Artificial Neural Networks (ICANN '16)

Springer LNCS, Vol. 9887, pp. 20-28, Barcelona, Spain, September 2016.

Conferences Selected O.E. David and N.S. Netanyahu

DeepPainter: Painter Classification Using Deep Convolutional Autoencoders

O.E. David and N.S. Netanyahu Conferences Selected

In this paper we describe the problem of painter classification, and propose a novel approach based on deep convolutional autoencoder neural networks. While previous approaches relied on image processing and manual feature extraction from paintings, our approach operates on the raw pixel level, without any preprocessing or manual feature extraction. We first train a deep convolutional autoencoder on a dataset of paintings, and subsequently use it to initialize a supervised convolutional neural network for the classification phase. The proposed approach substantially outperforms previous methods, improving the previous state-of-the-art for the 3-painter classification problem from 90.44% accuracy (previous state-of-the-art) to 96.52% accuracy, i.e., a 63% reduction in error rate.

2016 Sep

DNN-Buddies: A Deep Learning-Based Estimation Metric for the Jigsaw Puzzle Problem

International Conference on Artificial Neural Networks (ICANN '16)

Springer LNCS, Vol. 9887, pp. 20-28, Barcelona, Spain, September 2016.

Conferences Selected D. Sholomon, O.E. David, and N.S. Netanyahu

DNN-Buddies: A Deep Learning-Based Estimation Metric for the Jigsaw Puzzle Problem

D. Sholomon, O.E. David, and N.S. Netanyahu Conferences Selected

This paper introduces the first deep neural network-based estimation metric for the jigsaw puzzle problem. Given two puzzle piece edges, the neural network predicts whether or not they should be adjacent in the correct assembly of the puzzle, using nothing but the pixels of each piece. The proposed metric exhibits an extremely high precision even though no manual feature extraction is performed. When incorporated into an existing puzzle solver, the solution’s accuracy increases significantly, achieving thereby a new state-of-the-art standard.

2016 Sep

An Automatic Solver for Very Large Jigsaw Puzzles Using Genetic Algorithms

Genetic Programming and Evolvable Machines

Vol. 17, No. 3, pp. 291-313, September 2016.

Journal Papers Selected D. Sholomon, O.E. David, and N.S. Netanyahu

An Automatic Solver for Very Large Jigsaw Puzzles Using Genetic Algorithms

D. Sholomon, O.E. David, and N.S. Netanyahu Journal Papers Selected

In this paper we propose the first effective genetic algorithm (GA)-based jigsaw puzzle solver. We introduce a novel crossover procedure that merges two ‘‘parent’’ solutions to an improved ‘‘child’’ configuration by detecting, extracting, and combining correctly assembled puzzle segments. The solver proposed exhibits state-of-the-art performance, as far as handling previously attempted puzzles more accurately and efficiently, as well puzzle sizes that have not been attempted before. The extended experimental results provided in this paper include, among others, a thorough inspection of up to 30,745-piece puzzles (compared to previous attempts on 22,755-piece puzzles), using a considerably faster concurrent implementation of the algorithm. Furthermore, we explore the impact of different phases of the novel crossover operator by experimenting with several variants of the GA. Finally, we compare different fitness functions and their effect on the overall results of the GAbased solver

2016 Jan

Deep Learning: an Artificial Brain that Detects and Blocks any Type of Cyber Threat

Deep Learning Summit

San Francisco, January 2016.

Invited Talks O.E. David

Deep Learning: an Artificial Brain that Detects and Blocks any Type of Cyber Threat

O.E. David Invited Talks

Join Dr. Eli David’s presentation on the first application of deep learning to cybersecurity. Deep learning is inspired by the brain’s ability to learn: once a brain learns to identify an object, its identification becomes second nature. Similarly, as a deep learning-based artificial brain learns to detect any type of cyber threat, its prediction capabilities become instinctive. As a result, the most evasive and unknown cyber-attacks are immediately detected and prevented. Dr. David’s presentation will cover the evolution of artificial intelligence, from old rule-based systems to classical machine learning models and state-of-the-art deep learning models. A live demonstration will reveal the giant leap in accurately detecting new malware.

2015 Jul

DeepSign: Deep Learning for Automatic Malware Signature Generation and Classification

IEEE International Joint Conference on Neural Networks (IJCNN '15)

pp. 1-8, Killarney, Ireland, July 2015.

Conferences Selected O.E. David and N.S. Netanyahu

DeepSign: Deep Learning for Automatic Malware Signature Generation and Classification

O.E. David and N.S. Netanyahu Conferences Selected

This paper presents a novel deep learning based method for automatic malware signature generation and classification. The method uses a deep belief network (DBN), implemented with a deep stack of denoising autoencoders, generating an invariant compact representation of the malware behavior. While conventional signature and token based methods for malware detection do not detect a majority of new variants for existing malware, the results presented in this paper show that signatures generated by the DBN allow for an accurate classification of new malware variants. Using a dataset containing hundreds of variants for several major malware families, our method achieves 98.6% classification accuracy using the signatures generated by the DBN. The presented method is completely agnostic to the type of malware behavior that is logged (e.g., API calls and their parameters, registry entries, websites and ports accessed, etc.), and can use any raw input from a sandbox to successfully train the deep neural network which is used to generate malware signatures.

2015 May

The Golem Rises Against its Creator? Deep Learning and the Emergence of True Artificial Intelligence

College of Management Academic Studies

Keynote Speech, Rishon LeZion, Israel, May 2015.

Invited Talks O.E. David

The Golem Rises Against its Creator? Deep Learning and the Emergence of True Artificial Intelligence

O.E. David Invited Talks
2014 Dec

Malware Classification and Clustering with Deep Learning

KABARNIT Annual Conference

Herzlya, Israel, December 2014.

Invited Talks O.E. David

Malware Classification and Clustering with Deep Learning

O.E. David Invited Talks
2014 Dec

Team O.E. David Wins 11th Annual “Humies” Awards

ICGA Journal

Vol. 37, No. 4, pp. 224-225, December 2014.

Reports O.E. David, H.J. van den Herik, M. Koppel, and N.S. Netanyahu

Team O.E. David Wins 11th Annual “Humies” Awards

O.E. David, H.J. van den Herik, M. Koppel, and N.S. Netanyahu Reports
2014 Sep

Genetic Algorithms for Evolving Computer Chess Programs

IEEE Transactions on Evolutionary Computation

Winner of Gold Award in 11th Annual "Humies" Awards for Human-Competitive Results
Vol. 18, No. 5, pp. 779-789, September 2014.

Journal Papers Selected O.E. David, H.J. van den Herik, M. Koppel, and N.S. Netanyahu

Genetic Algorithms for Evolving Computer Chess Programs

O.E. David, H.J. van den Herik, M. Koppel, and N.S. Netanyahu Journal Papers Selected

Winner of Gold Award in 11th Annual “Humies” Awards for Human-Competitive Results

This paper demonstrates the use of genetic algorithms for evolving: 1) a grandmaster-level evaluation function, and 2) a search mechanism for a chess program, the parameter values of which are initialized randomly. The evaluation function of the program is evolved by learning from databases of (human) grandmaster games. At first, the organisms are evolved to mimic the behavior of human grandmasters, and then these organisms are further improved upon by means of coevolution. The search mechanism is evolved by learning from tactical test suites. Our results show that the evolved program outperforms a two-time world computer chess champion and is at par with the other leading computer chess programs.

2014 Jul

A Generalized Genetic Algorithm-Based Solver for Very Large Jigsaw Puzzles of Complex Types

AAAI Conference on Artificial Intelligence (AAAI '14)

pp. 2839-2845, Quebec City, Canada, July 2014.

Conferences Selected D. Sholomon, O.E. David, and N.S. Netanyahu

A Generalized Genetic Algorithm-Based Solver for Very Large Jigsaw Puzzles of Complex Types

D. Sholomon, O.E. David, and N.S. Netanyahu Conferences Selected

In this paper we introduce new types of square-piece jigsaw puzzles, where in addition to the unknown location and orientation of each piece, a piece might also need to be flipped. These puzzles, which are associated with a number of real world problems, are considerably harder, from a computational standpoint. Specifically, we present a novel generalized genetic algorithm (GA)- based solver that can handle puzzle pieces of unknown location and orientation (Type 2 puzzles) and (two-sided) puzzle pieces of unknown location, orientation, and face (Type 4 puzzles). To the best of our knowledge, our solver provides a new state-of-the-art, solving previously attempted puzzles faster and far more accurately, handling puzzle sizes that have never been attempted before, and assembling the newly introduced two-sided puzzles automatically and effectively. This paper also presents, among other results, the most extensive set of experimental results, compiled as of yet, on Type 2 puzzles.

2014 Jul

Genetic Algorithms and Deep Learning for Automatic Painter Classification

ACM Genetic and Evolutionary Computation Conference (GECCO '14)

pp. 1143-1150, Vancouver, Canada, July 2014.

Conferences E. Levy, O.E. David, and N.S. Netanyahu

Genetic Algorithms and Deep Learning for Automatic Painter Classification

E. Levy, O.E. David, and N.S. Netanyahu Conferences

In this paper we describe the problem of painter classification, and propose a novel hybrid approach incorporating genetic algorithms (GA) and deep restricted Boltzmann machines (RBM). Given a painting, we extract features using both generic image processing (IP) functions (e.g., fractal dimension, Fourier spectra coefficients, texture coefficients, etc.) and unsupervised deep learning (using deep RBMs). We subsequently compare several supervised learning techniques for classification using the extracted features as input. The results show that the weighted nearest neighbor (WNN) method, for which the weights are evolved using GA, outperforms both a support vector machine (SVM) classifier and a standard nearest neighbor classifier, achieving over 90% classification accuracy for the 3-painter problem (an improvement of over 10% relatively to previous results due to standard feature extraction only).

2014 Jul

Genetic Algorithm-Based Solver for Very Large Multiple Jigsaw Puzzles of Unknown Dimensions and Piece Orientation

ACM Genetic and Evolutionary Computation Conference (GECCO '14)

pp. 1191-1198, Vancouver, Canada, July 2014.

Conferences D. Sholomon, O.E. David, and N.S. Netanyahu

Genetic Algorithm-Based Solver for Very Large Multiple Jigsaw Puzzles of Unknown Dimensions and Piece Orientation

D. Sholomon, O.E. David, and N.S. Netanyahu Conferences

In this paper we propose the first genetic algorithm (GA)- based solver for jigsaw puzzles of unknown puzzle dimensions and unknown piece location and orientation. Our solver uses a novel crossover technique, and sets a new state-of-the-art in terms of the puzzle sizes solved and the accuracy obtained. The results are significantly improved, even when compared to previous solvers assuming known puzzle dimensions. Moreover, the solver successfully contends with a mixed bag of multiple puzzle pieces, assembling simultaneously all puzzles.

2014 Jul

Genetic Algorithms for Evolving Deep Neural Networks

ACM Genetic and Evolutionary Computation Conference (GECCO '14)

pp. 1451-1452, Vancouver, Canada, July 2014.

Conferences O.E. David and I. Greental

Genetic Algorithms for Evolving Deep Neural Networks

O.E. David and I. Greental Conferences

In recent years, deep learning methods applying unsupervised learning to train deep layers of neural networks have achieved remarkable results in numerous fields. In the past, many genetic algorithms based methods have been successfully applied to training neural networks. In this paper, we extend previous work and propose a GA-assisted method for deep learning. Our experimental results indicate that this GA-assisted approach improves the performance of a deep autoencoder, producing a sparser neural network.

2014 Jul

Genetic Algorithms for Evolving Computer Chess Programs

ACM Genetic and Evolutionary Computation Conference (GECCO '14)

"Hot Off the Press" track, Vancouver, Canada, July 2014.

Invited Talks O.E. David

Genetic Algorithms for Evolving Computer Chess Programs

O.E. David Invited Talks
2013 Jul

A Hybrid Genetic Approach for Stereo Matching

ACM Genetic and Evolutionary Computation Conference (GECCO '13)

pp. 1325-1332, Amsterdam, The Netherlands, July 2013.

Conferences E. Kiperwasser, O.E. David, and N.S. Netanyahu

A Hybrid Genetic Approach for Stereo Matching

E. Kiperwasser, O.E. David, and N.S. Netanyahu Conferences

In this paper we present a genetic algorithm (GA)-based approach for the stereo matching problem. More precisely, the approach presented is a combination of a simple dynamic programming algorithm, commonly used for stereo matching, with a practical GA-based optimization scheme. The performance of our scheme was evaluated on standard test data of the Middlebury benchmark. Specifically, the number of incorrect disparities on these data decreases by approximately 20% in comparison to the original approach (without the use of a GA).

2013 Jun

A Genetic Algorithm-Based Solver for Very Large Jigsaw Puzzles

IEEE Conference on Computer Vision and Pattern Recognition (CVPR '13)

pp. 1767--1774, Portland, OR, June 2013.

Conferences Selected D. Sholomon, O.E. David, and N.S. Netanyahu

A Genetic Algorithm-Based Solver for Very Large Jigsaw Puzzles

D. Sholomon, O.E. David, and N.S. Netanyahu Conferences Selected

In this paper we propose the first effective automated, genetic algorithm (GA)-based jigsaw puzzle solver. We introduce a novel procedure of merging two “parent” solutions to an improved ”child” solution by detecting, extracting, and combining correctly assembled puzzle segments. The solver proposed exhibits state-of-the-art performance solving previously attempted puzzles faster and far more accurately, and also puzzles of size never before attempted. Other contributions include the creation of a benchmark of large images, previously unavailable. We share the data sets and all of our results for future testing and comparative evaluation of jigsaw puzzle solvers.

2013 Jun

Painter Classification Using Genetic Algorithms

IEEE Congress on Evolutionary Computation (CEC '13)

pp. 3027-3034, Cancun, Mexico, June 2013.

Conferences E. Levy, O.E. David, and N.S. Netanyahu

Painter Classification Using Genetic Algorithms

E. Levy, O.E. David, and N.S. Netanyahu Conferences

This paper describes the problem of painter classification. We propose solving the problem by using genetic algorithms, which yields very promising results. The proposed methodology combines dimensionality reduction (via image preprocessing) and evolutionary computation techniques, by representing preprocessed data as a chromosome for a genetic algorithm (GA). The preprocessing of our scheme incorporates a diverse set of complex features (e.g., fractal dimension, Fourier spectra coefficients, and texture). The training phase of the GA employs a weighted nearest neighbor (NN) algorithm. We provide initial promising results for the 2- and 3-class cases, which offer significant improvement in comparison to a standard nearest neighbor classifier.

2013 Jan

In Vivo Quantification of Clot Formation in Extracorporeal Circuits

Studies in Health Technology and Informatics (MMVR20)

Vol. 184, pp. 148-150, January 2013.

Conferences R. Gerrah and O.E. David

In Vivo Quantification of Clot Formation in Extracorporeal Circuits

R. Gerrah and O.E. David Conferences

Clot formation is a common complication in extracorporeal circuits. In this paper we describe a novel method for clot formation analysis using image processing. We assembled a closed extracorporeal circuit and circulated blood at varying speeds. Blood filters were placed in downstream of the flow, and clotting agents were added to the circuit. Digital images of the filter were subsequently taken, and image analysis was applied to calculate the density of the clot. Our results show a significant correlation between the cumulative size of the clots, the density measure of the clot based on image analysis, and flow duration in the system.

2012 Sep

Device and Method for Performing Endoluminal Proximal Anastomosis

United States Patent

US8262681 B1, September 2012.

Patents R. Gerrah and O.E. David

Device and Method for Performing Endoluminal Proximal Anastomosis

R. Gerrah and O.E. David Patents

The present invention is of a method for performing an anastomosis and in particular, to such a method in which an end-to-side vessel endoluminal proximal anastomosis is performed in a minimally invasive CABG procedure.

2012 Mar

Genetic Programming for Classification of Moving Objects

VULCAN Annual Conference

Herzlya, Israel, March 2012.

Invited Talks O.E. David

Genetic Programming for Classification of Moving Objects

O.E. David Invited Talks
2011 Oct

Genetic Algorithms for Automatic Object Movement Classification

International Conference on Convergence and Hybrid Information Technology (ICHIT '11)

Springer LNCS, Vol. 6935, pp. 258-265, Daejeon, Korea, October 2011.

Conferences O.E. David, N.S. Netanyahu, and Y. Rosenberg

Genetic Algorithms for Automatic Object Movement Classification

O.E. David, N.S. Netanyahu, and Y. Rosenberg Conferences

This paper presents an integrated approach, combining a state-of-the-art commercial object detection system and genetic algorithms (GA)-based learning for automatic object classification. Specifically, the approach is based on applying weighted nearest neighbor classification to feature vectors extracted from the detected objects, where the weights are evolved due to GA-based learning. Our results demonstrate that this GA-based approach is considerably superior to other standard classification methods.

2011 Mar

Genetic Algorithms and Weighted Nearest Neighbor for Automatic Object Classification

VULCAN Annual Conference

Herzlya, Israel, March 2011.

Invited Talks O.E. David

Genetic Algorithms and Weighted Nearest Neighbor for Automatic Object Classification

O.E. David Invited Talks
2011 Mar

Expert-Driven Genetic Algorithms for Simulating Evaluation Functions

Genetic Programming and Evolvable Machines

Vol. 12, No. 1, pp. 5-22, March 2011.

Journal Papers Selected O.E. David, M. Koppel, and N.S. Netanyahu

Expert-Driven Genetic Algorithms for Simulating Evaluation Functions

O.E. David, M. Koppel, and N.S. Netanyahu Journal Papers Selected

In this paper we demonstrate how genetic algorithms can be used to reverse engineer an evaluation function’s parameters for computer chess. Our results show that using an appropriate expert (or mentor), we can evolve a program that is on par with top tournament-playing chess programs, outperforming a two time World Computer Chess Champion. This performance gain is achieved by evolving a program that mimics the behavior of a superior expert. The resulting evaluation function of the evolved program consists of a much smaller number of parameters than the expert’s. The extended experimental results provided in this paper include a report on our successful participation in the 2008 World Computer Chess Championship. In principle, our expert-driven approach could be used in a wide range of problems for which appropriate experts are available.

2010 Aug

Device for Harvesting a Blood Vessel

United States Patent

2010/0198241 A1, August 2010.

Patents R. Gerrah and O.E. David

Device for Harvesting a Blood Vessel

R. Gerrah and O.E. David Patents

The present invention relates to an apparatus, device and a method for harvesting blood vessels, and in particular, to such an apparatus, device and method in which the internal mammary artery (IMA) is harvested for coronary artery bypass graft (CABG) surgery using a minimally invasive approach or a conventional procedure.

2010 Jul

Genetic Algorithms for Automatic Classification of Moving Objects

ACM Genetic and Evolutionary Computation Conference (GECCO '10)

pp. 2069-2070, Portland, OR, July 2010.

Conferences O.E. David, N.S. Netanyahu, Y. Rosenberg, and M. Shimoni

Genetic Algorithms for Automatic Classification of Moving Objects

O.E. David, N.S. Netanyahu, Y. Rosenberg, and M. Shimoni Conferences

This paper presents an integrated approach, combining a state-of-the-art commercial object detection system and genetic algorithms (GA)-based learning for automatic object classification. Specifically, the approach is based on applying weighted nearest neighbor classification to feature vectors extracted from the detected objects, where the weights are evolved due to GA-based learning. Our results demonstrate that this GA-based approach is considerably superior to other standard classification methods.

2010 Jun

Genetic Algorithms for Automatic Search Tuning

ICGA Journal

Vol. 33, No. 2, pp. 67-79, June 2010.

Journal Papers O.E. David, M. Koppel, and N.S. Netanyahu

Genetic Algorithms for Automatic Search Tuning

O.E. David, M. Koppel, and N.S. Netanyahu Journal Papers

In this paper we introduce a novel method for automatically tuning the search parameters of a chess program using genetic algorithms (GA). Our results show that a large set of parameter values can be learned automatically, such that the resulting performance is comparable with that of manually tuned parameters of top tournament-playing chess programs.

2010 Jun

Optimizing Selective Search in Chess

International Conference on Machine Learning

Workshop on Machine Learning and Games (ICML-MLG), Haifa, Israel, June 2010.

Conferences O.E. David, M. Koppel, and N.S. Netanyahu

Optimizing Selective Search in Chess

O.E. David, M. Koppel, and N.S. Netanyahu Conferences

In this paper we introduce a novel method for automatically tuning the search parameters of a chess program using genetic algorithms. Our results show that a large set of parameter values can be learned automatically, such that the resulting performance is comparable with that of manually tuned parameters of top tournament-playing chess programs.

2010 Apr

Genetic Algorithms for Automatic Object Classification

VULCAN Annual Conference

Herzlya, Israel, April 2010.

Invited Talks O.E. David

Genetic Algorithms for Automatic Object Classification

O.E. David Invited Talks
2009 Jul

Simulating Human Grandmasters: Evolution and Coevolution of Evaluation Functions

ACM Genetic and Evolutionary Computation Conference (GECCO '09)

pp. 1483-1489, Montreal, Canada, July 2009.

Conferences Selected O.E. David, H.J. van den Herik, M. Koppel, and N.S. Netanyahu

Simulating Human Grandmasters: Evolution and Coevolution of Evaluation Functions

O.E. David, H.J. van den Herik, M. Koppel, and N.S. Netanyahu Conferences Selected

This paper demonstrates the use of genetic algorithms for evolving a grandmaster-level evaluation function for a chess program. This is achieved by combining supervised and unsupervised learning. In the supervised learning phase the organisms are evolved to mimic the behavior of human grandmasters, and in the unsupervised learning phase these evolved organisms are further improved upon by means of coevolution.

While past attempts succeeded in creating a grandmaster-level program by mimicking the behavior of existing computer chess programs, this paper presents the first successful attempt at evolving a state-of-the-art evaluation function by learning only from databases of games played by humans. Our results demonstrate that the evolved program outperforms a two-time World Computer Chess Champion.

2009 Jun

Genetic Algorithms Based Learning for Evolving Intelligent Organisms

Ph.D. Thesis

Advisors: Nathan Netanyahu and Moshe Koppel. Department of Computer Science, Bar-Ilan University, Israel, June 2009.

Theses O.E. David

Genetic Algorithms Based Learning for Evolving Intelligent Organisms

O.E. David Theses

The following is a review of the thesis by Dap Hartmann:

Mimicking the Black Box: Genetically evolving evaluation functions and search algorithms, ICGA Journal, Vol. 33, No. 1, pp. 42-43, March 2010.

2009 Jan

Expert-Driven Parameter Tuning by Means of Natural Selection

Department of Industrial Engineering Department of Computer Science Colloquium

Tel Aviv University, Israel, January 2009.

Invited Talks O.E. David

Expert-Driven Parameter Tuning by Means of Natural Selection

O.E. David Invited Talks
2008 Oct

Extended Null-Move Reductions

International Conference on Computers and Games (CG 08)

Springer LNCS, Vol. 5131, pp. 205-216, Beijing, China, October 2008.

Conferences O.E. David and N.S. Netanyahu

Extended Null-Move Reductions

O.E. David and N.S. Netanyahu Conferences

In this paper we reviw the conventional versions of nullmove pruning, and present our enhancements which allow for a deeper search with greater accuracy. While the conventional versions of nullmove pruning use reduction values of R ≤ 3, we use an aggressive reduction value of R = 4 within a verified adaptive configuration which maximizes the benefit from the more aggressive pruning, while limiting its tactical liabilities. Our experimental results using our grandmasterlevel chess program, Falcon, show that our null-move reductions (NMR) outperform the conventional methods, with the tactical benefits of the deeper search dominating the deficiencies. Moreover, unlike standard null-move pruning, which fails badly in zugzwang positions, NMR is impervious to zugzwangs. Finally, the implementation of NMR in any program already using null-move pruning requires a modification of only a few lines of code.

2008 Sep

The 16th World Computer-Chess Championship

ICGA Journal

Vol. 31, No. 3, pp. 166-171, September 2008.

Reports O.E. David
2008 Jul

Genetic Algorithms for Mentor-Assisted Evaluation Function Optimization

ACM Genetic and Evolutionary Computation Conference (GECCO '08)

Winner of Best Paper Award in GECCO '08
pp. 1469-1475, Atlanta, GA, July 2008.

Conferences Selected O.E. David, M. Koppel, and N.S. Netanyahu

Genetic Algorithms for Mentor-Assisted Evaluation Function Optimization

O.E. David, M. Koppel, and N.S. Netanyahu Conferences Selected

Winner of Best Paper Award in GECCO ’08

In this paper we demonstrate how genetic algorithms can be used to reverse engineer an evaluation function’s parameters for computer chess. Our results show that using an appropriate mentor, we can evolve a program that is on par with top tournament-playing chess programs, outperforming a two-time World Computer Chess Champion. This performance gain is achieved by evolving a program with a smaller number of parameters in its evaluation function to mimic the behavior of a superior mentor which uses a more extensive evaluation function. In principle, our mentor-assisted approach could be used in a wide range of problems for which appropriate mentors are available.

2008 Jul

Genetic Algorithms and Chess: From Novice to Grandmaster in 60 Seconds.

Department of Computer Science Colloquium

Bar-Ilan University, Israel, July 2008.

Invited Talks O.E. David

Genetic Algorithms and Chess: From Novice to Grandmaster in 60 Seconds.

O.E. David Invited Talks
2006 Sep

Blockage Detection in King and Pawn Endings

International Conference on Computers and Games (CG '04)

Springer LNCS, Vol. 3846, pp. 187-201, Ramat-Gan, Israel, September 2006.

Conferences O.E. David, A. Felner, and N.S. Netanyahu

Blockage Detection in King and Pawn Endings

O.E. David, A. Felner, and N.S. Netanyahu Conferences

The conventional search methods using static evaluation at leaf nodes are liable to missing the fact that no win goal exists in certain positions. Blockage positions, in which neither side can penetrate into the opponent’s camp, are a prominent example of such positions. Deep search cannot detect the existence of a blockage, since its judgment is based solely on static evaluation, without taking the goals into consideration. In this paper we introduce a blockage detection method, which manages to detect a large set of blockage positions in pawn endgames, with practically no additional overhead. By examining different aspects of the pawn structure, it checks whether the pawns form a blockage which prevents the king from penetrating into the opponent’s camp. It then checks several criteria to find out whether the blockage is permanent.

2005 May

Search and Knowledge-Based Innovations in Computer Chess

Master's Thesis

Advisor: Nathan Netanyahu. Department of Computer Science, Bar-Ilan University, Israel, May 2005.

Theses O.E. David

Search and Knowledge-Based Innovations in Computer Chess

O.E. David Theses
2004 Sep

Blockage Detection in Pawn Endgames

ICGA Journal

Vol. 27, No. 3, pp. 150-158, September 2004.

Journal Papers O.E. David, A. Felner, and N.S. Netanyahu

Blockage Detection in Pawn Endgames

O.E. David, A. Felner, and N.S. Netanyahu Journal Papers

In this article we introduce a blockage-detection method, which manages to detect a large set of blockage positions in pawn endgames, with practically no additional overhead. By examining different elements of the pawn structure, the method checks whether the Pawns form a blockage which prevents the King from penetrating into the opponent’s camp. It then checks several criteria to find out whether the blockage is permanent.

2004 Aug

12th World Computer-Chess Championship

BNVKI, Belgium-Netherlands Association for Artificial Intelligence

Vol. 21, No. 4, pp. 84-85, August 2004.

Reports O.E. David
2003 Dec

11th World Computer-Chess Championship

ICGA Journal

Vol. 26, No. 4, pp. 252-258, December 2003.

Reports O.E. David
2002 Dec

Deep (Re)Search in Computer Chess

Department of Computer Science Colloquium

Bar-Ilan University, Israel, December 2002.

Invited Talks O.E. David

Deep (Re)Search in Computer Chess

O.E. David Invited Talks
2002 Oct

Report on the Symposium Man vs. Machine: The Experiment

ICGA Journal

Vol. 25, No. 4, pp. 250-251, December 2002.

Reports O.E. David and N.S. Netanyahu

Report on the Symposium Man vs. Machine: The Experiment

O.E. David and N.S. Netanyahu Reports
2002 Sep

Verified Null-Move Pruning

ICGA Journal

Vol. 25, No. 3, pp. 153-161, September 2002.

Journal Papers O.E. David and N.S. Netanyahu

Verified Null-Move Pruning

O.E. David and N.S. Netanyahu Journal Papers

In this article we review standard null-move pruning and introduce our extended version of it, which we call verified null-move pruning. In verified null-move pruning, whenever the shallow null-move search indicates a fail-high, instead of cutting off the search from the current node, the search is continued with reduced depth. Our experiments with verified null-move pruning show that on average, it constructs a smaller search tree with greater tactical strength in comparison to standard null-move pruning. Moreover, unlike standard null-move pruning, which fails badly in zugzwang positions, verified null-move pruning manages to detect most zugzwangs and in such cases conducts a re-search to obtain the correct result. In addition, verified null-move pruning is very easy to implement, and any standard null-move pruning program can use verified null-move pruning by modifying only a few lines of code.

Our experiments with verified null-move pruning show that on average, it constructs a smaller search tree with greater tactical strength in comparison to standard null-move pruning. Moreover, unlike standard null-move pruning, which fails badly in zugzwang positions, verified null-move pruning manages to detect most zugzwangs and in such cases conducts a re-search to obtain the correct result. In addition, verified null-move pruning is very easy to implement, and any standard null-move pruning program can use verified null-move pruning by modifying only a few lines of code.

.03

RESUME

ACADEMIC AND PROFESSIONAL POSITIONS
  • Present

    Co-Founder and CTO

    Deep Learning Based Company

  • 2009
    Present

    Researcher

    Bar-Ilan University

    Researcher in deep learning and evolutionary computation (genetic algorithms and genetic programming), in addition to supervising the research of graduate students in these fields.
  • 2003
    Present

    Lecturer

    Bar-Ilan University

    Lecturer in Evolutionary Algorithms and Bio-Inspired Algorithms courses, at department of Computer Science. The two courses cover evolutionary algorithms and neural networks (focusing on deep learning).

    Previously TA in Image Processing, Operating Systems, and Algorithms, at departments of Computer Science, Mathematics, and Engineering.
  • 2015
    Present

    Researcher

    Tel Aviv University

    Researcher in deep learning.
  • 2007
    2015

    Consultant in machine learning and deep learning

    Served in numerous capacities successfully designing, implementing, and leading machine learning and deep learning based projects in real-world environments.
EDUCATION
  • 2005
    2009

    Computer Science - Ph.D.

    Bar-Ilan University

    Thesis title: Genetic Algorithms Based Learning for Evolving Intelligent Organisms
    Advisors: Prof. Nathan Netanyahu and Prof. Moshe Koppel
  • 2003
    2005

    Computer Science - M.Sc.

    Bar-Ilan University

    Thesis title: Search and Knowledge-Based Innovations in Computer Chess
    Advisor: Prof. Nathan Netanyahu
  • 2000
    2003

    Computer Science - B.Sc.

    Bar-Ilan University

HONORS AND AWARDS
  • 2016
    Barcelona

    Best Paper Award

    25th International Conference on Artificial Neural Networks (ICANN)

    For the paper "DeepChess: End-to-End Deep Neural Network for Automatic Learning in Chess".

    First end-to-end learning method that results in a grandmaster-level chess program without any a priori knowledge, using a combination of unsupervised and supervised deep learning methods.
  • 2014
    Vancouver

    Gold "Humies" Award

    11th Annual "Humies" Awards for Human-Competitive Results

    www.genetic-programming.org/combined.php#humie_year_2014

    See report at: content.iospress.com/articles/icga-journal/icg37407
  • 2008
    Beijing

    Second place in World Computer Speed Chess Championship

    ICGA - International Computer Games Association

    Second place in World Computer Speed Chess Championship 2008 using a genetically evolved version of FALCON chess program (Beijing, 2008).
  • 2008
    Atlanta

    Best Paper Award

    Genetic and Evolutionary Computation Conference (GECCO)

    For the paper "Genetic Algorithms for Mentor-Assisted Evaluation Function Optimization".

    First successful application of machine learning using genetic algorithms that evolves a chess program's evaluation function from scratch to reach grandmaster-level performance.
.04

TEACHING

CURRENT
  • 2006
    Present

    Bio-Intelligent Algorithms (Deep Learning)

    Department of Computer Science - Bar-Ilan University

    Advanced course in neural networks and deep learning.
  • 2006
    Present

    Evolutionary Algorithms

    Department of Computer Science - Bar-Ilan University

    Advanced course in evolutionary computation (genetic algorithms and genetic programming).
TEACHING HISTORY
  • 2004
    2010

    Image Processing

    Departments of Computer Science and Mathematics - Bar-Ilan University

    Covering core concepts in image processing and computer vision.
  • 2003
    2007

    Algorithms

    Department of Computer Science - Bar-Ilan University

    Introductory course on algorithms.
  • 2004
    2009

    Operating Systems

    Departments of Computer Science and Engineering - Bar-Ilan University

    Extensive course on the inner workings of operating systems.
.05

PROJECTS

Computer Chess
  • 2003
    2012

    Falcon

    FALCON is the first grandmaster-level chess playing program which learns all its evaluation function and search parameters using machine learning. These values are learned using a combination of genetic algorithms based methods, which are described in several of my publications.

    The main concepts are also described in the following Galileo and Ynet article:
    אבולוציה בשחור-לבן

    FALCON successfully participated in three World Computer Chess Championships:
    - 2nd place, World Computer Speed Chess Championship 2008, Beijing, China.
    - 6th place, World Computer Chess Championship 2008, Beijing, China.
    - 3rd place, World Computer Speed Chess Championship 2004, Ramat-Gan, Israel.
    - 7th place, World Computer Chess Championship 2004, Ramat-Gan, Israel.
    - 10th place, World Computer Chess Championship 2003, Graz, Austria.

    Additionally, FALCON has served as a testbed for several major research innovations, which resulted in a Best Paper Award in 2008 Genetic and Evolutionary Computation Conference, the Gold Award in the prestigious "Humies" Awards for Human-Competitive Results in 2014, and recently the Best Paper Award in 2016 International Conference on Artificial Neural Networks.
  • 2002
    2004

    Genesis

    The first scientific chess engine from Israel, used for algorithmic research of search trees. Freely available online.
.06

MEDIA

Selected Recent Media Mentions and Reports
  • May 06
    2016

    Cool Vendors in Digital Workplace Security, 2016

    Gartner

    Deep Instinct selected as "Cool Vendor", one of the most prestigious awards in cybersecurity!

    www.gartner.com/doc/3309518
.07

CONTACT