Fig 1 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
Deep learning (DL) has been considered as a breakthrough technique in the field of artificial intelligence and machine learning. Conceptually, it relies on a many-layer network that exhibits a hierarchically non-linear processing capability. Some DL architectures such as deep neural networks, deep belief networks and recurrent neural networks have...
Context in source publication
Citations
... Deep Learning is a subset of machine learning that relies on a many-layer network which exhibits a hierarchically non-linear processing capability [53]. As well as the interaction behaviour at the surface level, it can learn the deeper relations between these behaviours. ...
This thesis covers the development of fabrication and calibration methods for soft stretch sensors. A variety of material approaches are discussed that are applicable to soft strain sensing. Printing methods are discussed in detail as a method of producing sensors via cavity channels within inert elastomer substrates that can then be manually filled with active compounds, and the direct syringe printing of nanocomposites. Next is presented a general method for calibrating highly hysteretic resistive stretch sensors and the results for this applied to 1D and 2D sensors. The method involves three-stages. The first stage requires a calibration step in which the strain of the sensor material is measured using a webcam while the electrical response is measured via a set of arduino-based electronics. During this data collection stage, the strain is applied manually by pulling the sensor over a range of strains and strain rates corresponding to the realistic in-use strain and strain rates. In the second stage the data is passed to a bespoke neural network architecture and trained on part of the dataset. The ability of the networks to predict the strain state given a stream of unseen electrical resistance data is then assessed. In the third stage, the sensor system is removed from the camera and used as desired. The final section discusses such use case - the fabrication of bespoke prosthetic socket liners with growth tracking for child prosthetics. Using 3D scanning technology to create the desired shape of the liner, I show my method for producing a castable mould, and the embedding of stretch sensors within it to achieve potential growth tracking of the residual limb.
... The reason behind this is the developments in ANN models and hardware systems that can handle and implement these models. (Sugiarto & Pasila, 2018) The ANNs can be separated into three generations based on their computational units and performance (Figure 1). ...
... The first generation of the ANNs has started in 1943 with the work of Mc-Culloch and Pitts (Sugiarto & Pasila, 2018). Their work was based on a computational model for neural networks where each neuron is called "perceptron". ...
... They do give an output (spike) if the neurons collect enough data to surpass the internal threshold. Moreover, neuron structures can work in parallel (Sugiarto & Pasila, 2018). In theory, thanks to these two features SNNs consume less energy and work faster than second-generation ANNs (Maass, 1997). ...
Technical report on Neuromorphic Computing within Human Brain Project
Classification algorithms are used mainly in design Brain-Computer Interface (BCI) systems for ElectroEncephaloGraphy (EEG) signals. This paper presents the commonly employed deep learning algorithms and describe their critical properties. This paper presents the recent research of applying deep learning in EEG signal classifications.
