Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

The Internet of Things (IoT) is expected to require more effective and efficient wireless communications than ever before. For this reason, techniques such as spectrum sharing, dynamic spectrum access, extraction of signal intelligence and optimized routing will soon become essential components of the IoT wireless communication paradigm. In this vision, IoT devices must be able to not only learn to autonomously extract spectrum knowledge on-the-fly from the network but also leverage such knowledge to dynamically change appropriate wireless parameters ( e.g. , frequency band, symbol modulation, coding rate, route selection, etc.) to reach the network’s optimal operating point. Given that the majority of the IoT will be composed of tiny, mobile, and energy-constrained devices, traditional techniques based on a priori network optimization may not be suitable, since (i) an accurate model of the environment may not be readily available in practical scenarios; (ii) the computational requirements of traditional optimization techniques may prove unbearable for IoT devices. To address the above challenges, much research has been devoted to exploring the use of machine learning to address problems in the IoT wireless communications domain. The reason behind machine learning’s popularity is that it provides a general framework to solve very complex problems where a model of the phenomenon being learned is too complex to derive or too dynamic to be summarized in mathematical terms. This work provides a comprehensive survey of the state of the art in the application of machine learning techniques to address key problems in IoT wireless communications with an emphasis on its ad hoc networking aspect. First, we present extensive background notions of machine learning techniques. Then, by adopting a bottom-up approach, we examine existing work on machine learning for the IoT at the physical, data-link and network layer of the protocol stack. Thereafter, we discuss directions taken by the community towards hardware implementation to ensure the feasibility of these techniques. Additionally, before concluding, we also provide a brief discussion of the application of machine learning in IoT beyond wireless communication. Finally, each of these discussions is accompanied by a detailed analysis of the related open problems and challenges.

Machine learning for wireless communications in the Internet of Things: A comprehensive survey

We have developed a computational method based on polyploid phasing of long sequence reads to resolve collapsed regions of segmental duplications within genome assemblies. Segmental Duplication Assembler (SDA; https://github.com/mvollger/SDA ) constructs graphs in which paralogous sequence variants define the nodes and long-read sequences provide attraction and repulsion edges, enabling the partition and assembly of long reads corresponding to distinct paralogs. We apply it to single-molecule, real-time sequence data from three human genomes and recover 33-79 megabase pairs (Mb) of duplications in which approximately half of the loci are diverged ( 99.9%) and that the diverged sequence corresponds to copy-number-variable paralogs that are absent from the human reference genome. Our method can be applied to other complex genomes to resolve the last gene-rich gaps, improve duplicate gene annotation, and better understand copy-number-variant genetic diversity at the base-pair level.

/pdf/long-read-sequence-and-assembly-of-segmental-duplications-3w7fdvmcf4.pdf

Long-read sequence and assembly of segmental duplications.

Performing link adaptation in a multiantenna and multiuser system is challenging because of the coupling between precoding, user selection, spatial mode selection and use of limited feedback about the channel. The problem is exacerbated by the difficulty of selecting the proper modulation and coding scheme when using orthogonal frequency division multiplexing (OFDM). This paper presents a data-driven approach to link adaptation for multiuser multiple input mulitple output (MIMO) OFDM systems. A machine learning classifier is used to select the modulation and coding scheme, taking as input the SNR values in the different subcarriers and spatial streams. A new approximation is developed to estimate the unknown interuser interference due to the use of limited feedback. This approximation allows to obtain SNR information at the transmitter with a minimum communication overhead. A greedy algorithm is used to perform spatial mode and user selection with affordable complexity, without resorting to an exhaustive search. The proposed adaptation is studied in the context of the IEEE 802.11ac standard, and is shown to schedule users and adjust the transmission parameters to the channel conditions as well as to the rate of the feedback channel.

Learning-Based Adaptive Transmission for Limited Feedback Multiuser MIMO-OFDM

In matrix factorization problems, one seeks to decompose a data matrix into a product of two matrices—frequently, one captures meaningful information contained in the data, and the other specifies how this information is combined to generate the data matrix. In this paper, matrix factorization that arises in haplotype assembly, an important NP-hard problem in genomics, is studied. Haplotypes are sequences of chromosomal variations in an individual’s genome, which are of critical importance for understudying the individual’s susceptibility to various diseases. A novel formulation of haplotype assembly as the partially observed low-rank matrix factorization problem is proposed and efficiently solved via a modified gradient descent method that exploits salient structural properties of sequencing data. In particular, the observed matrix in the problem at hand contains noisy samples of the product of an informative matrix with rows having entries from a finite alphabet and a matrix with rows that are standard unit basis. Convergence of the proposed algorithm is analyzed and its performance tested on both synthetic and experimental data. The results demonstrate superior accuracy and speed of the proposed method as compared to state-of-the-art haplotype assembly techniques.

Structured Low-Rank Matrix Factorization for Haplotype Assembly

High-throughput DNA sequencing technologies allow fast and affordable sequencing of individual genomes and thus enable unprecedented studies of genetic variations. Information about variations in the genome of an individual is provided by haplotypes, ordered collections of single nucleotide polymorphisms. Knowledge of haplotypes is instrumental in finding genes associated with diseases, drug development, and evolutionary studies. Haplotype assembly from high-throughput sequencing data is challenging due to errors and limited lengths of sequencing reads. The key observation made in this paper is that the minimum error-correction formulation of the haplotype assembly problem is identical to the task of deciphering a coded message received over a noisy channel---a classical problem in the mature field of communication theory. Exploiting this connection, we develop novel haplotype assembly schemes that rely on the bit-flipping and belief propagation algorithms often used in communication systems. The latter algorithm is then adapted to the haplotype assembly of polyploids. We demonstrate on both simulated and experimental data that the proposed algorithms compare favorably with state-of-the-art haplotype assembly methods in terms of accuracy, while being scalable and computationally efficient.

Decoding genetic variations: communications-inspired haplotype assembly

Single carrier frequency domain equalization (SC-FDE) uses cyclically prefixed quadrature amplitude modulation to permit simple frequency domain equalization at the receiver. Link adaptation for SC-FDE systems, where the modulation and coding rate are adapted based on the current channel state, is straightforward with perfect channel state information due to the simple analytical form of the post-processing signal-to-noise ratio (SNR). Imperfect channel state information, however, introduces adaptation errors. This paper proposes a machine learning-based approach for link adaptation in bit interleaved convolutionally encoded SC-FDE systems. To improve performance in the presence of channel uncertainty, principal component analysis is used to reduce the feature space dimensionality consisting of the channel coefficients, noise variance, and post-processing SNR. The reduced dimension feature set improves performance of the link adaptation classifier and leads to higher performance versus just the post-processing SNR estimate. Simulation results indicate that the proposed algorithm increases the goodput while maintaining the target packet error rate, achieving optimal adaptation in 95% of the tested cases.

A Machine Learning Approach to Link Adaptation for SC-FDE System

Homologous chromosome pairs have similar DNA sequences and encode the same type of information but are not identical. The most common form of variations between them are the so-called single nucleotide polymorphisms (SNPs), where two chromosomes in a homologous pair differ at individual corresponding positions. The problem of haplotype assembly is concerned with finding an ordered sequence of SNPs associated with each one of the chromosomes in a pair. The information about genetic variations (i.e., SNPs) is inferred using short reads provided by the high-throughput DNA sequencing systems. The reads are potentially erroneous and haplotype assembly can thus be interpreted as the process of separating fragments into two potentially inconsistent (due to noise) classes, each corresponding to one of the haplotypes in a pair. Solving this problem is known to be NP-hard, and hence practical haplotype assembly typically involves heuristics and low complexity approximations. In our work, we propose novel graphical models of the hap-lotype assembly problem and design efficient message passing algorithms for solving it. The developed algorithms perform close to the existing state-of-the-art techniques while being more computationally efficient.

A message passing algorithm for haplotype assembly

In recent years, advancements in high-throughput DNA sequencing technologies enabled heretofore impractical studies of genetic variations. Cells of diploid organisms, including humans, have a number of chromosome pairs that are homologous — they encode essentially the same genetic information and are almost identical but vary in certain location. These variations are referred to as single nucleotide polymorphisms. The complete information about genetic variations in an individual genome is given by haplotypes, ordered sequences of single nucleotide polymorphisms for each homologous pair of chromosomes. In this paper, we derive a graphical formulation of the haplotype assembly problem, propose an iterative scheme for single individual haplotyping, and demonstrate the performance of the algorithm on experimental data. The results demonstrate that the proposed method has better accuracy than state-of-the-art haplotype assembly techniques.

Zrinka Puljiz

Papers

Decoding genetic variations: communications-inspired haplotype assembly

A Machine Learning Approach to Link Adaptation for SC-FDE System

A message passing algorithm for haplotype assembly

Iterative learning of single individual haplotypes from high-throughput DNA sequencing data