The decorrelating detector and the linear minimum mean-square error (MMSE) detector are known to be effective strategies to counter the presence of multiuser interference in code-division multiple-access channels; in particular, those multiuser detectors provide optimum near-far resistance. When training data sequences are available, the MMSE multiuser detector can be implemented adaptively without knowledge of signature waveforms or received amplitudes. This paper introduces an adaptive multiuser detector which converges (for any initialization) to the MMSE detector without requiring training sequences. This blind multiuser detector requires no more knowledge than does the conventional single-user receiver: the desired user’s signature waveform and its timing. The proposed blind multiuser detector is made robust with respect to imprecise knowledge of the received signature waveform of the user of interest.

HINE Parameters X0, X1, ....Xp Output E0, E1,....Ep Error Desired Output D0, D1,...Dp Y0, Y1,...Yp Input w w0 w1 AT&T Laboratories (c) COST FUNCTION Output E0, E1,....Ep Error Desired Output D0, D1,...Dp Y0, Y1,...Yp X0, X1, ....Xp Input Parameters w B R A COMPUTING THE GRADIENT WITH BACKPROPAGATION O = A(I1, I2) dI1 = dO ¶ A ¶ I1 dI2 = dO ¶ A ¶ I2 - The learning machine is composed of modules (e.g. layers) - Each module can do two things: 1- compute its outputs from its inputs (FPROP) 2- compute gradient vectors at its inputs from gradient vectors at its outputs (BPROP) A O, dO I1, dI1 I2, dI2 AT&T Laboratories (c) AN INTERESTING SPECIAL CASE: MULTILAYER NETWORKS X0, X1, ....Xp Output Desired Output D0, D1,...Dp Y0, Y1,...Yp Input || D - Y || 2 2 1 WX F() WX F() Mean Square Error Parameters (weights + biases) w Weight matrix E0, E1,....Ep Sigmoids + Biase

We propose a new learning approach for image retrieval, which we call adjustment learning, and demonstrate its use for face recognition and color matching. Our approach is motivated by a frequently encountered problem, namely, that variability in the original data representation which is not relevant to the task may interfere with retrieval and make it very difficult. Our key observation is that in real applications of image retrieval, data sometimes comes in small chunks - small subsets of images that come from the same (but unknown) class. This is the case, for example, when a query is presented via a short video clip. We call these groups chunklets, and we call the paradigm which uses chunklets for unsupervised learning adjustment learning. Within this paradigm we propose a linear scheme, which we call Relevant Component Analysis; this scheme uses the information in such chunklets to reduce irrelevant variability in the data while amplifying relevant variability. We provide results using our method on two problems: face recognition (using a database publicly available on the web), and visual surveillance (using our own data). In the latter application chunklets are obtained automatically from the data without the need of supervision.

Abstract—A common problem that arises in adaptive filtering, autoregressive modeling, or linear prediction is the selection of an appropriate order for the underlying linear parametric model. We address this problem for linear prediction, but instead of fixing a specific model order, we develop a sequential prediction algorithm whose sequentially accumulated average squared prediction error for any bounded individual sequence is as good as the performance attainable by the best sequential linear predictor of order less than some w. This predictor is found by transforming linear prediction into a problem analogous to the sequential probability assignment problem from universal coding theory. The resulting universal predictor uses essentially a performance-weighted average of all predictors for model orders less than w. Efficient lattice filters are used to generate the predictions of all the models recursively, resulting in a complexity of the universal algorithm that is no larger than that of the largest model order. Examples of prediction performance are provided for autoregressive and speech data as well as an example of adaptive data equalization. Index Terms—Adaptive filters, Bayes procedures, learning systems, least squares methods, model order, prediction methods,

Abstruct- We consider transmission of data over multiple coupled channels, such as bundles of twisted-pair copper wires in the local sub-scriber loop, and between central offices in the public switched tele-phone network. Transceiver designs for such channels typically treat the crosstalk between adjacent twisted pairs as random noise uncorrelated with the transmitted signal. We propose a transmitterheceiver pair that compensates for crosstalk by treating an entire bundle of twisted pairs as a single multiinput~ultioutput channel with a (slowly varying) matrix transfer function. The proposed transceiver uses multichannel adap-tive FIR filters to cancel near- and far-end crosstalk, and to pre- and postprocess the input/output of the channel. The linear pre- and post-processors that minimize mean squared error between the received and transmitted signal in the presence of both near- and far-end crosstalk are derived. The performance of an adaptive near-end crosstalk canceller using the stochastic gradient (LMS) transversal algorithm is illustrated via numerical simulation. Plots of mean squared error versus time and eye diagrams are presented assuming a standard transmission line model for the channel. A signal design algorithm that maps a vector input bit stream to a stream of channel symbol vectors is also presented. This algorithm is illustrated explicitly for a simple model of two coupled channels. It is shown that the achievable rate using the proposed signaling scheme is very close to the rate attainable in the absence of far-end crosstalk, and is significantly greater than the achievable rate assuming that far-end crosstalk is treated as additive noise with unknown statistics. 0 I.

Abstract—A class of adaptive reduced-rank interference suppression algorithms is presented based on the multi-stage Wiener filter (MSWF). The performance is examined in the context of direct-sequence (DS) code division multiple access (CDMA). Unlike the Principal Components method for reduced-rank filtering, the algorithms presented can achieve near full-rank performance with a filter rank much less than the dimension of the signal subspace. We present batch and recursive algorithms for estimating the filter parameters, which do not require an eigen-decomposition. Algorithm performance in a heavily loaded DS-CDMA system is characterized via computer simulation. Results show that the reduced-rank algorithms require significantly fewer training samples than other reduced- and full-rank algorithms. Index Terms—Adaptive filters, code-division multiple access (CDMA), interference suppression. I.

Methods for automatically managing the performance of computing services must estimate a performance model of that service. This paper explores properties that are necessary for performance model estimation of black-box computer systems when used together with adaptive feedback loops. It shows that the standard method of least-squares estimation often gives rise to models that make the control loop perform the opposite action of what is desired. This produces large oscillations and bad tracking performance. The paper evaluates what combination of input and output data provides models with the best properties for the control loop. Plus, it proposes three extensions to the controller that makes it perform well, even when the model estimated would have degraded performance. Our proposed techniques are evaluated with an adaptive controller that provides latency targets for workloads on black-box computer services under a variety of conditions. The techniques are evaluated on two systems: a three-tier e-commerce site and a web server. Experimental results show that our best estimation approach improves the ability of the controller to meet the latency goals significantly. Previously oscillating workload latencies are with our techniques smooth around the latency targets. 1

During recent years much interest has been given to the application of Singular Value Decomposition in association with extended--order and overdetermined evaluation in the finite parametric spectral estimation domain. Such approaches have been shown to perform superior to other methods for discontinuous frequency signal, e.g. the harmonic retrieval problem. In this report a similar approach is applied to wide--banded AR processes. It is found that the so--called extraneous poles of the lower rank solution spoil the spectral estimate. A new approach, the direction weighted total least squares solution, which enforces the extraneous poles to be located at the origin while maintaining the good properties of the aforementioned approaches, is therefore introduced. Computer simulation experiments clearly indicate that this approach is superior to existing overdetermined and extended-- or parsimonic order methods. The author is currently being supported by a grant from the Danish Technical...

Abstract—The performance of adaptive least squares (LS) filtering is analyzed for the suppression of multiple-access interference. Both full-rank LS filters and reduced-rank LS filters, which reside in a lower dimensional Krylov space, are considered with training, and without training but with known signature for the desired user. We compute the large system limit of output signal-to-interference-plus-noise ratio (SINR) as a function of normalized observations, load, and noise level. Specifically, the number of users, the degrees of freedom, and the number of training symbols or observations all tend to infinity with fixed ratios and. Our results account for an arbitrary power distribution over the users, data windowing (e.g., recursive LS (RLS) with exponential windowing), and initial diagonal loading of the covariance matrix to prevent ill-conditioning. Numerical results show that the large system analysis accurately predicts the simulated convergence performance of the algorithms considered with moderate degrees of freedom (typically aQP). Given a fixed, short training length, the relative performance of full- and reduced-rank filters depends on the selected rank and diagonal loading. With an optimized diagonal loading factor, the performance of full- and reduced-rank filters are similar. However, full-rank performance is generally much more sensitive to the choice of diagonal loading factor than reduced-rank performance. Index Terms—Adaptive filter, large system analysis, least squares (LS), reduced-rank filters. I.

The increasing interest for using the OSVD in the real--time DSP domain necessitates an efficient computation of the OSVD. Special interest has been given to Jacobi--like algorithms which also is the case in this paper. After a description of the basic orthogonal transformations, algorithms for computing the OSVD are classified and shortly described. Various rotation schemes for Jacobi--like algorithms enabling concurrent computation are described and compared. It is found that the well--known cyclic--by--row scheme is the most suited for real--time DSP applications and it is shown that this scheme allows for concurrent implementations. Finally, some 6 Jacobi--like algorithms, including a new one presented here, are described and compared in detail. The differences of the various algorithms can be summarized in four. (i) The assumed structure of the matrix. (ii) How the rotation formula is expressed. (iii) The applied rotation scheme. (iv) How the result is delivered. All four items ar...