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

Data Mining Practical Machine Learning Tools and Techniques

Neuromorphic computing has come to refer to a variety of brain-inspired computers, devices, and models that contrast the pervasive von Neumann computer architecture This biologically inspired approach has created highly connected synthetic neurons and synapses that can be used to model neuroscience theories as well as solve challenging machine learning problems The promise of the technology is to create a brain-like ability to learn and adapt, but the technical challenges are significant, starting with an accurate neuroscience model of how the brain works, to finding materials and engineering breakthroughs to build devices to support these models, to creating a programming framework so the systems can learn, to creating applications with brain-like capabilities In this work, we provide a comprehensive survey of the research and motivations for neuromorphic computing over its history We begin with a 35-year review of the motivations and drivers of neuromorphic computing, then look at the major research areas of the field, which we define as neuro-inspired models, algorithms and learning approaches, hardware and devices, supporting systems, and finally applications We conclude with a broad discussion on the major research topics that need to be addressed in the coming years to see the promise of neuromorphic computing fulfilled The goals of this work are to provide an exhaustive review of the research conducted in neuromorphic computing since the inception of the term, and to motivate further work by illuminating gaps in the field where new research is needed

A Survey of Neuromorphic Computing and Neural Networks in Hardware.

Activity detection and classification are very important for autonomous monitoring of humans for applications, including assistive living, rehabilitation, and surveillance. Wearable sensors have found wide-spread use in recent years due to their ever-decreasing cost, ease of deployment and use, and ability to provide continuous monitoring as opposed to sensors installed at fixed locations. Since many smart phones are now equipped with a variety of sensors, such as accelerometer, gyroscope, and camera, it has become more feasible to develop activity monitoring algorithms employing one or more of these sensors with increased accessibility. We provide a complete and comprehensive survey on activity classification with wearable sensors, covering a variety of sensing modalities, including accelerometer, gyroscope, pressure sensors, and camera- and depth-based systems. We discuss differences in activity types tackled by this breadth of sensing modalities. For example, accelerometer, gyroscope, and magnetometer systems have a history of addressing whole body motion or global type activities, whereas camera systems provide the context necessary to classify local interactions, or interactions of individuals with objects. We also found that these single sensing modalities laid the foundation for hybrid works that tackle a mix of global and local interaction-type activities. In addition to the type of sensors and type of activities classified, we provide details on each wearable system that include on-body sensor location, employed learning approach, and extent of experimental setup. We further discuss where the processing is performed, i.e., local versus remote processing, for different systems. This is one of the first surveys to provide such breadth of coverage across different wearable sensor systems for activity classification.

A Survey on Activity Detection and Classification Using Wearable Sensors

In a few years, the world will be populated by billions of connected devices that will be placed in our homes, cities, vehicles, and industries. Devices with limited resources will interact with the surrounding environment and users. Many of these devices will be based on machine learning models to decode meaning and behavior behind sensors’ data, to implement accurate predictions and make decisions. The bottleneck will be the high level of connected things that could congest the network. Hence, the need to incorporate intelligence on end devices using machine learning algorithms. Deploying machine learning on such edge devices improves the network congestion by allowing computations to be performed close to the data sources. The aim of this work is to provide a review of the main techniques that guarantee the execution of machine learning models on hardware with low performances in the Internet of Things paradigm, paving the way to the Internet of Conscious Things. In this work, a detailed review on models, architecture, and requirements on solutions that implement edge machine learning on Internet of Things devices is presented, with the main goal to define the state of the art and envisioning development requirements. Furthermore, an example of edge machine learning implementation on a microcontroller will be provided, commonly regarded as the machine learning “Hello World”.

https://www.mdpi.com/1424-8220/20/9/2533/pdf

Edge Machine Learning for AI-Enabled IoT Devices: A Review

One of the most extended algorithms for image scaling is bicubic interpolation. In this paper, a hardware architecture for bicubic interpolation (HABI) is proposed. The HABI proposed is integrated by three main blocks: the first one generates the interpolation coefficients, which implements the bicubic function to be used in HABI; the second one performs the interpolation process and the third one is a control unit that synchronizes the processing and the pipeline stages. The architecture work with monochromatic images, but it can be extended for working with RGB color images. Our design description is coded in Handel-C language and implemented on a Xilinx Virtex II Pro FPGA. The proposed system runs 10 times faster than an Intel Pentium 4-based PC at 2.4 GHz. Comparison with other related works are provided

Real-time FPGA-based architecture for bicubic interpolation: an application for digital image scaling

Mobile artifacts such as smartphones have made possible the development of wearable systems for user activity monitoring and recognition due to the synergy of communication, computation and sensing capabilities in battery-powered systems-on-chip. Due to user acceptability, smartphones are able to measure nonintrusively proprioceptive motion outside of a controlled environment for rather long periods of time using embedded inertial sensors. Though work has been done for accelerometer-based activity recognition, the portability of the smartphone to a single fixed tight position has been a major constraint to easy the interpretation of the collected data. In this paper, a human activity hierarchical recognition system based on time-domain features and neural networks without the need of the smartphone to be constrained to a single fixed body position is presented. Experimental results on Android-capable smartphones on four on-body locations show that the recognition system achieves high classification rates, above 92%, for five activities including static, walking, running, and up-down stairs walking, running continuously in near real-time with reduced power consumption.

Robust smartphone-based human activity recognition using a tri-axial accelerometer

Detect, classify and keep track, in real-time, on different kinds of objects or vehicles that are moving on a road is crucial for traffic managements systems, among other research areas. In this paper, a vision based system to detect, track, count and classify moving vehicles, on any kind of road, is shown. The data acquisition system consists of a HD-RGB camera placed on the road, while the information processing is performed by clustering and classification algorithms. The system obtained an efficiency score over the 95 percent in test cases, as well, the correct classification of 85 percent of the test objects. Also, the system achieves 30 fps in image processing with a resolution of 1280×720.

http://ieeexplore.ieee.org/iel7/6900043/6908326/06908506.pdf

Computer vision based real-time vehicle tracking and classification system.

Indoor positioning is a recent technology that has gained interest in industry and academia thanks to the promising results of locating objects, people or robots accurately in indoor environments. One of the utilized technologies is based on algorithms that process the Received Signal Strength Indicator (RSSI) in order to infer location information without previous knowledge of the distribution of the Access Points (APs) in the area of interest. This paper presents the design and implementation of an indoor positioning mobile application, which allows users to capture and build their own RSSI maps by off-line training of a set of selected classifiers and using the models generated to obtain the current indoor location of the target device. In an early experimental and design stage, 59 classifiers were evaluated, using data from proposed indoor scenarios. Then, from the tested classifiers in the early stage, only the top-five classifiers were integrated with the proposed mobile indoor positioning, based on the accuracy obtained for the test scenarios. The proposed indoor application achieves high classification rates, above 89%, for at least 10 different locations in indoor environments, where each location has a minimum separation of 0.5 m.

On-Device Learning of Indoor Location for WiFi Fingerprint Approach.

Biometric identification (BI) is one of the most explored topics in recent years. One of the most important techniques for BI is face recognition. Face recognition systems (FRSs) are an important field in computer vision, because it represents a non-invasive BI technique. In this paper, a FRS is proposed. In the first step, a face detection algorithm is used for extracting faces from video frames (training videos) and generating a face database. In a second step, filtering and preprocessing are applied to face images obtained in the previous step. In a third step, a collection of machine learning algorithms are trained using as input data the faces obtained in the previous step. Finally, the classifiers are used for classify faces obtained from video frames (test videos). The obtained results shows the suitability of this approach for analyzing large collections of videos where previous face labels are not available.

Marco Aurelio Nuño-Maganda

Papers

Real-time FPGA-based architecture for bicubic interpolation: an application for digital image scaling

Robust smartphone-based human activity recognition using a tri-axial accelerometer

Computer vision based real-time vehicle tracking and classification system.

On-Device Learning of Indoor Location for WiFi Fingerprint Approach.

Evaluation of machine learning techniques for face detection and recognition