From: VideoLectures on Fri, Jan 14 2011 3:05 AM

Stochastic Gradient Boosted Decision Trees (GBDT) is one of the most widely used learning algorithms in machine learning today. It is adaptable, easy to interpret, and produces highly accurate models. However, most implementations today are computationally expensive and require all training data...

From: VideoLectures on Fri, Jan 14 2011 3:04 AM

Problems involving maximization of submodular functions arise in many applications, such as combinatorial auctions and coverage optimization in wireless networks. Submodular maximization can be also thought of as a unifying framefork for several classical problems including Max Cut, Max k-Cover ...

From: VideoLectures on Fri, Jan 14 2011 3:04 AM

In a network orientation problem one is given a mixed graph, consisting of directed and undirected edges, and a set of sourcetarget vertex pairs. The goal is to orient the undirected edges so that a maximum number of pairs admit a directed path from the source to the target. This problem arises ...

From: VideoLectures on Fri, Jan 14 2011 3:04 AM

Most existing MKL approaches employ the large margin principle to learning kernels. However, we point out that the margin itself can not well describe the goodness of a kernel due to the negligence of the scaling. We use the ratio between the margin and the radius of the minimal enclosing ball o...

From: VideoLectures on Fri, Jan 14 2011 3:03 AM

Sparse methods for supervised learning aim at finding good linear predictors from as few variables as possible, i.e., with small cardinality of their supports. This combinatorial selection problem is often turned into a convex optimization problem by replacing the cardinality function by its con...

From: VideoLectures on Fri, Jan 14 2011 3:03 AM

Most organisms facing a choice between multiple stimuli will look repeatedly at them, presumably implementing a comparison process between the items’ values. Little is known about the exact nature of the comparison process in value-based decision-making, or about the role that the visual fixatio...

From: VideoLectures on Fri, Jan 14 2011 3:03 AM

Ordinary and partial differential equations are extensively used in different branches of science and engineering to model wide variety of phenomena, such as diffusion, stability, wave propagation, population growth and chemical reactions, to mention just a few. For most practical problems these...

From: VideoLectures on Fri, Jan 14 2011 3:02 AM

We develop a theory of online learning by defining several complexity measures. Among them are analogues of Rademacher complexity, covering numbers and fat-shattering dimension from statistical learning theory. Relationship among these complexity measures, their connection to online learning, an...

From: VideoLectures on Thu, Jan 13 2011 2:56 AM

From: VideoLectures on Thu, Jan 13 2011 2:56 AM

Structured prediction (SP) problems are characterized by strong interdependence among the output variables, usually with sequential, graphical, or combinatorial structure [17, 7]. Obtaining good predictors often requires a large effort in feature/kernel design and tuning (usually done via crossv...

From: VideoLectures on Thu, Jan 13 2011 2:56 AM

The Gaussian process (GP) is a popular way to specify dependencies between random variables in a probabilistic model. In the Bayesian framework the covariance structure can be specified using unknown hyperparameters. Integrating over these hyperparameters considers different possible explanations...

From: VideoLectures on Thu, Jan 13 2011 2:56 AM

This tutorial surveys methodology and theory for high-dimensional statistical inference when the number of variables or features greatly exceeds sample size. Particular emphasis will be placed on problems of model and feature selection. This includes variable selection in regression models or est...

From: VideoLectures on Thu, Jan 13 2011 2:56 AM

This tutorial surveys methodology and theory for high-dimensional statistical inference when the number of variables or features greatly exceeds sample size. Particular emphasis will be placed on problems of model and feature selection. This includes variable selection in regression models or est...

From: VideoLectures on Thu, Jan 13 2011 2:55 AM

The purpose of this tutorial is to explore the interplay between sensing and control, to highlight the "information knot" that ties them, and to design inference and learning algorithms to compute "representations" from data that are optimal, by design, for decision and control tasks. We will fo...

From: VideoLectures on Thu, Jan 13 2011 2:55 AM

The human visual system, and almost every digital camera, achieves trichromacy by using sensors of different spectral sensitivity arranged in some spatial pattern. What pattern of spatial sampling is best? I’ll describe work on learning an optimal (ok, locally optimal) color filter array. This l...

From: VideoLectures on Thu, Jan 13 2011 2:55 AM

It was realized soon after the introduction of Multiple Kernel Learning that l1 - regularization and its groupwise extension are related to MKL by dualization. MKL can therefore be viewed as providing an extension of sparsity to function spaces. However, this extension is not limited l1 - norm. ...

From: VideoLectures on Thu, Jan 13 2011 2:55 AM

The study of the structure and function of proteins serves up many problems that offer challenges and opportunities for computational and statistical research. I will overview my experiences in several such problem domains, ranging from domains where off-the-shelf ideas can be fruitfully applied...

From: VideoLectures on Wed, Jan 12 2011 2:49 AM

The world’s market on piezoceramics is about 10 billion dollar per year. Main applications are actuators, e.g. for fuel injection and for piezomotors, sensors e.g. for parking pilot and airbag, transducers for ultrasound investigations and transformers for example for mobile phones. Legislation ...

From: VideoLectures on Fri, Jan 07 2011 9:22 AM

Professor Saltzman continues his presentation on DNA technology by discussing control of gene expression using two methods of RNA silencing: anti-sense therapy and RNA interference. Molecular cloning techniques to mass-produce proteins using plasmid, restriction enzymes, ligase, and antibiotic s...

From: VideoLectures on Fri, Jan 07 2011 9:22 AM

In this final lecture, Professor Saltzman talks about artificial organs, with a stress on synthetic biomaterials. First, the body’s responses (immunological and scar healing responses) to foreign materials are introduced. This leads to discussion of different types of polymer/plastic materials (...