Can intelligence be learned? At LISP, Prof. Chin and a collection of budding (master's and PhD) computer scientists, engineers, and mathematicians are passionate about exploring, disseminating, and innovating research in machine learning, intelligent decision making systems, and signal processing in order to answer this question.
Hi! I am a research professor in the Dept. of Computer Science & Hariri Institute for Computing at Boston University, My primary appointment is with Draper Laboratory in Cambridge, MA., where I am a Chief Scientist - Decision Systems, and lead various research in the area of Data-to-Decision (D2D). I am also a research affiliate of Dept. of AeroAstro Engineering at MIT, from which Draper Lab started as Prof. Draper’s Laboratory 80 some years ago in 1943, and a visiting research fellow at London Institute of Mathematical Sciences. My life-time job, however, is to serve God and His purpose in my generation, and I attend Antioch Baptist Church located in Cambridge, MA.
Machine Learning, Differential Geometry, Game Theory, Compressive Sensing, Extremal Graph Theory
I am an incoming Computer Science PhD student at Boston University. I am very passionate about space and computer science and I hope to play a part in making space travel safer and more accessible. I am originally from London. I lived there for eleven years before moving to Dubai, where I spent my middle and high school years. After Dubai I moved to Boston, where I completed my undergraduate career at Boston University. I currently live in Boston.
My research interests include machine learning, computer vision, networking, operating systems, and graphics. I am curious about a great many things in computer science and I hope to be able to work on many interesting problems.
I am an avid powerlifter and I thoroughly enjoy all the mental and physical challenges it presents. One of my biggest dreams is to be able to go to space.
I am a master's student in the Computer Science department at Boston University. My interest in machine learning is focused primarily on its vast array of applications, such as in the fields of computer vision, game playing, and social networks.
Machine Learning, Social Network Analysis, Virtual Reality
Academics aside, I'm a guitarist and soccer lover. Also, I spend a majority of my free time mulling over what's going to happen next on Game of Thrones.
I am currently a PhD student in the computer science department at Boston University and a Draper fellow. My research interests focus in machine learning, specifically artificial neural networks. I deal a lot with recurrent neural network approaches to both supervised and unsupervised problems. I earned my B.S in computer science and electrical engineering from Duke University, and my M.S. in computer engineering from N.C. State.
I received my Bachelor of Science from Boston University's Computer Science department in 2014. After I graduated, I collected the data for, designed, and created the website Parkour Theory to help those interested learn and referenece terminology, and share new moves. In 2015, I returned to Boston University because I missed the joy of arduous late night studying. I am now a second year master's student in computer science again at Boston University, and a researcher at Draper. The CS department can't seem to get rid of me.
Currently, I am researching neural network architectures for improving action recognition, natural language understanding for creating labeled datasets, and unsupervised learning for constrained automatic neural network model selection.
I am currently a third-year doctoral candidate at Harvard University (SEAS) in Applied Mathematics and a Draper Laboratory fellow. I work under the joint advisory of Dr. Vahid Tarokh (Harvard) and Dr. Peter Chin (Draper). I hold a BS in Applied Math from UCLA and should be receiving my MS from Harvard (also in Applied Math!) at the end of the Fall 2016 semester. My past research has been fairly diverse, ranging from signal processing using the Empirical Wavelet Transform (UCLA, with Dr. J. Gilles), to distributed resource allocation strategies (Harvard/Draper, with Drs. N. Li & R. Mangoubi). My current and future pursuits concern computational algebraic topology: in a nutshell, identifying/characterizing lower-dimensional algebraic structure within high-dimensional data sets. Please feel free to email me at kathrynheal[at]g.harvard.edu, or visit my website at http://scholar.harvard.edu/heal.
Computational Algebraic Topology, Signal & Image Processing, Machine Learning
I am a first year master's student in Department of Computer Science at Boston University and a Draper Laboratory Fellow under the supervision of Prof. Peter Chin. I received my B.S. in chemical engineering with a minor in Chemistry from MIT. I have 4-5 years of experience working in a Biotech startup firm and have done research on operations and technology evaluation.
Language of Fluency: Matlab, Python, Java
Interested in neural networks (CNN, RNN) to solve label prediction problems. My current projects include extraction of kinematic features to predict the trajectory of moving vehicles.
I am a 2nd year master’s student in biomedical engineering at Brown University. I graduated from Brown in May 2015 in biomedical engineering. My focus is on neuroengineering, specifically signal processing and feature identification & classification for closed loop implantable devices. I’m a Draper fellow on TRANSFORM DBS, which aims to develop a closed loop deep brain stimulation implant for the treatment of neuropsychiatric disorders. My past work has been on identifying the neural features of Parkinson’s disease in the motor cortex.
I was a 4 year varsity athlete on Brown’s swim team, I am scuba certified, and I love Grey’s Anatomy.
Ravi Kailasam Rajendran
I hold a Bachelor’s degree in Engineering majoring in Electronics and Communication from Madras Institute of Technology, Anna University, Chennai. Currently, I am working on my thesis, which is mainly based on the study of topological and statistical analysis of the behavior classifiers for tracking algorithms focusing on the improvisation by integrating the multiple hypothesis tracking with target behaviors.
My research areas include computer vision and machine learning focusing on object recognition and target tracking. I also like creating AI models for the strategic games and graphical models for image processing.
My favorite author is Sidney Sheldon and my hobbies are listening to string instruments and watching TV-serials and I like F.R.I.E.N.D.S and The Big Bang Theory.
I earned my B.S in Mathematics from Southeast University in Nanjing. Currently, I am a master student in the computer science department at Boston University. My research interests focus on machine learning, graphical models and statistical learning. Recently, I'm working on unsupervised social network analysis problems and outlier detection in large scale data. I also love open source technologies (such as R lang) and would like to be one of the selfless geeks in open source societies.
Grid computing techniques in linear algebra and optimization, network science and combinatorial optimazations, support vector machines, unsupervied learning with neural networks, unsupervised and supervised outlier detection, social network analysis
I like traveling and hope to travel all over the world. I also enjoy hiking (mountains, canyons, caves, costlines and deserts), photography, museums, Chinese and Japanese animes, instrumental music.
Xiao (Kieran) Wang
I am a PhD student in System Engineering.
My research interests are in the areas of machine learning, control theory and operations research.
I'm a PhD student in the LISP research group in CS department at Boston University. I'm supervised by professor Peter Chin. I received my B.A. in mathematics and computer science from Boston University in May 2016. Currently, I'm working on programming language processing.
My research interests lie in machine learning, computer vision, game theory, robotics.
Yiteng (Ivan) Xu
Hi, my name is Yiteng (Ivan) Xu. I'm a second year master's student. I am interested in Boltzmann Machines and Deep Belief Networks.
Contagion in Networks, Graph Isomorphism, Unsupervised Learning
My hobby is everything about airplanes.
I am a second year master's student in Computer Science. I am interested in machine learning and computer vision.
Machine Learning, Computer Vision, and Data Mining
My hobbies include swimming, basketball and programming! Although, I am not good at any one of them.
Machine learning is a subfield of computer science that evolved from the study of pattern recognition and computational learning theory in artificial intelligence. In 1959, Arthur Samuel defined machine learning as a "Field of study that gives computers the ability to learn without being explicitly programmed". Machine learning explores the study and construction of algorithms that can learn from and make predictions on data. Such algorithms operate by building a model from example inputs in order to make data-driven predictions or decisions expressed as outputs, rather than following strictly static program instructions.
Game theory is "the study of mathematical models of conflict and cooperation between intelligent rational decision-makers." Game theory is mainly used in economics, political science, and psychology, as well as logic, computer science, biology and poker. Originally, it addressed zero-sum games, in which one person's gains result in losses for the other participants. Today, game theory applies to a wide range of behavioral relations, and is now an umbrella term for the science of logical decision making in humans, animals, and computers.
Compressed sensing (also known as compressive sensing, compressive sampling, or sparse sampling) is a signal processing technique for efficiently acquiring and reconstructing a signal, by finding solutions to underdetermined linear systems. This is based on the principle that, through optimization, the sparsity of a signal can be exploited to recover it from far fewer samples than required by the Shannon-Nyquist sampling theorem. There are two conditions under which recovery is possible. The first one is sparsity which requires the signal to be sparse in some domain. The second one is incoherence which is applied through the isometric property which is sufficient for sparse signals.
Algorithmic topology, or computational topology, is a subfield of topology with an overlap with areas of computer science, in particular computational geometry and computational complexity theory. A primary concern of algorithmic topology, as its name suggests, is to develop efficient algorithms for solving topological problems using topological methods from computable topology to solve algorithmic problems from other fields.
EXTREMAL GRAPH THEORY
Extremal graph theory is a branch of the mathematical field of graph theory. Extremal graph theory studies extremal (maximal or minimal) graphs which satisfy a certain property. Extremality can be taken with respect to different graph invariants, such as order, size or girth. More abstractly, it studies how global properties of a graph influence local substructures of the graph.
Computer security, also known as cybersecurity or IT security, is the protection of information systems from theft or damage to the hardware, the software, and to the information on them, as well as from disruption or misdirection of the services they provide. It includes controlling physical access to the hardware, as well as protecting against harm that may come via network access, data and code injection, and due to malpractice by operators, whether intentional, accidental, or due to them being tricked into deviating from secure procedures.
The low-rank matrix recovery problem consists of reconstructing an unknown low-rank matrix from a few linear measurements, possibly corrupted by noise. One of the most popular method in low-rank matrix recovery is based on nuclear-norm minimization, which seeks to simultaneously estimate the most significant singular values of the target low-rank matrix by adding a penalizing term on its nuclear norm. In this paper, we introduce a new method that re- quires substantially fewer measurements needed for exact matrix recovery compared to nuclear norm minimization. The proposed optimization program utilizes a sparsity promoting regularization in the form of the entropy function of the singular values. Numerical experiments on synthetic and real data demonstrates that the proposed method outperforms stage-of-the-art nuclear norm minimization algorithms.
Recently, CMOS image sensors have attracted more and more attention from the applications of navigation, monitoring and search-and-rescue operations. Specially, CMOS image sensors mounted on insects need to be fast, adaptive to the environment and power efficiency. To simultaneously satisfy both requirements of reconstruction quality and low power consumption, we propose a compressed sensing block-wise exposure control algorithm using optical flow estimation. This framework has been demonstrated to further improve recovery performance (> 25 dB) with high compression ratio (>= 10 : 1), which also provides a promising method for real-time CMOS implementation.
In this paper, we study Locally Compressed Sensing for images, where sampling process is allowed to be performed on arbitrary local regions of the images. We propose a fast and efficient reconstruction algorithm which utilizes local structures of images. Several numerical experiments on real images demonstrates that our algorithm yields better reconstruction quality than existing techniques at much lower computational complexity and memory requirement.
In this paper, we develop Targeted Dot Product Representation (TarDPR), a DPR-based feature selection and combination framework for friend recommendation in online social networks (OSNs). Our approach modifies conventional DPR techniques and makes itself applicable to OSNs by focusing on computing a consistent representation while minimizing unnecessary suggestions made outside these interested regions. A notable property of TarDPR is its ability to effectively incorporate different types of social features and produce new meaningful features that help competitive approaches to significantly improve their recommendation quality. We derive an iterative algorithm for TarDPR that is supported by mathematical analysis, and is efficient on large social traces. To certify the usability of our approach, we conduct empirical experiments on real social traces including Facebook and Foursquare social networks. The competitive experimental results show that TarDPR achieves up to 15% improvement in comparison with other competitive methods. These results consequently confirm the efficacy of our suggested framework.
A fully autonomous intracranial device is built to continually record neural activities in different parts of the brain, process these sampled signals, decode features that correlate to behaviors and neuropsychiatric states, and use these features to deliver brain stimulation in a closed-loop fashion. In this paper, we describe the sampling and stimulation aspects of such a device. We first describe the signal processing algorithms of two unsupervised spike sorting methods. Next, we describe the LFP time-frequency analysis and feature derivation from the two spike sorting methods. Spike sorting includes a novel approach to constructing a dictionary learning algorithm in a Compressed Sensing (CS) framework. We present a joint prediction scheme to determine the class of neural spikes in the dictionary learning framework; and, the second approach is a modified OSort algorithm which is implemented in a distributed system optimized for power efficiency. Furthermore, sorted spikes and time-frequency analysis of LFP signals can be used to generate derived features (including cross-frequency coupling, spike-field coupling). We then show how these derived features can be used in the design and development of novel decode and closed-loop control algorithms that are optimized to apply deep brain stimulation based on a patient's neuropsychiatric state. For the control algorithm, we define the state vector as representative of a patient's impulsivity, avoidance, inhibition, etc. Controller parameters are optimized to apply stimulation based on the state vector's current state as well as its historical values. The overall algorithm and software design for our implantable neural recording and stimulation system uses an innovative, adaptable, and reprogrammable architecture that enables advancement of the state-of-the-art in closed-loop neural control while also meeting the challenges of system power constraints and concurrent development with ongoing scientific research design- d to define brain network connectivity and neural network dynamics that vary at the individual patient level and vary over time.
We demonstrate a photonic system for pseudorandom sampling of multi-tone sparse radio-frequency (RF) signals in an 11.95-GHz bandwidth using < 1% of the measurements required for Nyquist sampling. Pseudorandom binary sequence (PRBS) patterns are modulated onto highly chirped laser pulses, encoding the patterns onto the optical spectra. The pulses are partially compressed to increase the effective sampling rate by 2.07×, modulated with the RF signal, and fully compressed yielding optical integration of the PRBS-RF inner product prior to photodetection. This yields a 266× reduction in the required electronic sampling rate. We introduce a joint-sparsity-based matching-pursuit reconstruction via bagging to achieve accurate recovery of tones at arbitrary frequencies relative to the reconstruction basis.
We experimentally demonstrate a photonic RF sampling system that utilizes chirp processing of ultrafast laser pulses to achieve all-optical high-rate pseudorandom patterning and inner product integration for compressed sensing measurement. We successfully acquire multi-tone sparse radio frequency (RF) signals at arbitrary offsets from the reconstruction basis frequencies in an 11.95 GHz bandwidth utilizing less than 1% of the measurements traditionally required for Nyquist sampling. Pseudorandom binary sequence (PRBS) patterns are modulated onto time-stretched optical pulses, encoding them onto the optical spectra at a rate of one unique pattern per pulse. Thus patterned, the pulses are then partially-compressed, increasing the system's effective sampling rate by 2.07×, well beyond the electronic modulation rate. The partially-compressed patterned pulses are then modulated again with the RF signal under test and fully compressed to perform optical integration of the PRBS-RF inner product before output photodetection and digitization. This achieves a reduction in required electronic sampling rate by two orders of magnitude.
We demonstrate ultrahigh-speed pseudorandom structured illumination imaging of high-speed microscopic flows with real-time all-optical data compression. We present up to 19.8 Gigapixel/sec continuous acquisition rates using a 720-MHz ADC sampling rate.
To meet the growing demand of wireless and power efficient neural recordings systems, we demonstrate an unsupervised dictionary learning algorithm in Compressed Sensing (CS) framework which can be implemented in VLSI systems. Without prior label information of neural spikes, we extend our previous work to unsupervised learning and construct a dictionary with discriminative structures for spike sorting. To further improve the reconstruction and classification performance, we proposed a joint prediction to determine the class of neural spikes in dictionary learning. When the neural spikes is compressed 50 times, our approach can achieve an average gain of 2 dB and 15 percentage units over state-of-the-art of CS approaches in terms of the reconstruction quality and classification accuracy respectively.
We demonstrate 119.2-GSample/s compressive microwave frequency detection using spectrally-encoded ultrafast laser pulses. We sense sparse tones over> 35-GHz instantaneous bandwidth with 2-MHz accuracy using< 300 consecutive compressive measurements acquired at a 400-MHz rate.
We demonstrate ultrahigh-speed spectral shaping of broadband laser pulses to create structured illumination of a high-speed flow for compressive microscopy. We achieve up to 39.6 gigapixel/s continuous imaging rates using a 720-MHz ADC sampling rate.
We demonstrate an imaging system employing continuous high-rate photonically-enabled compressed sensing (CHiRP-CS) to enable efficient microscopic imaging of rapidly moving objects with only a few percent of the samples traditionally required for Nyquist sampling. Ultrahigh-rate spectral shaping is achieved through chirp processing of broadband laser pulses and permits ultrafast structured illumination of the object flow. Image reconstructions of high-speed microscopic flows are demonstrated at effective rates up to 39.6 Gigapixel/sec from a 720-MHz sampling rate.
Nonnegative Matrix Factorization (NMF), defined as factorizing a nonnegative matrix into two nonnegative factor matrices, is a particularly important problem in machine learning. Unfortunately, it is also ill-posed and NP-hard. We propose a fast, robust, and provably correct algorithm, namely Gradient Vertex Pursuit (GVP), for solving a well-defined instance of the problem which results in a unique solution: there exists a polytope, whose vertices consist of a few columns of the original matrix, covering the entire set of remaining columns. Our algorithm is greedy: it detects, at each iteration, a correct vertex until the entire polytope is identified. We evaluate the proposed algorithm on both synthetic and real hyperspectral data, and show its superior performance compared with other state-of-the-art greedy pursuit algorithms.
We present a combined hardware/software architecture to perform Markov Chain Monte Carlo sampling on probabilistic graphical models in a brain-inspired, energy-aware manner. By combining massively-parallel neuromorphic hardware architecture (SpiNNaker) with algorithms we've have developed for the event-based framework employed in SpiNNaker, we achieve large speedups when performing inference as compared to a traditional PC. We present results from two sampling approaches both well suited to the SpiNNaker architecture. Neural sampling, the first of the two approaches relies directly on simulating networks of spiking neurons while the second, Gibb's sampling is more flexible but still takes advantage of the hardware's event-handling capabilities.
We present a chirp processing technique for encoding pseudorandom patterns onto the spectra of broadband optical pulses for compressed sensing (CS) measurement. We demonstrate applications to characterization of ultrawideband sparse radio frequency (RF) signals and to very high-speed continuous microscopic flow imaging. In both domains, the optical sampling technique permits accurate recovery of the signals under test from only a few percent of the measurements required for conventional Nyquist sampling, significantly relaxing the required analog-to-digital conversion bandwidth and amount of data storage.
In this paper, we present and analyze a simple and robust spectral algorithm for the stochastic block model with $ k $ blocks, for any $ k $ fixed. Our algorithm works with graphs having constant edge density, under an optimal condition on the gap between the density inside a block and the density between the blocks. As a co-product, we settle an open question posed by Abbe et. al. concerning censor block models.
If you're a master's or PhD student join us! You can find us in room MCS 289 or join us in the Hariri Institute for our machine learning summer reading group.