China
Africa
Explore chapters and articles related to this topic
Spike Processing with SOA Dynamics
Published in Paul R. Prucnal, Bhavin J. Shastri, Malvin Carl Teich, Neuromorphic Photonics, 2017
Paul R. Prucnal, Bhavin J. Shastri, Malvin Carl Teich
Although this model successfully performs two qualitative features of biological computation—integration and thresholding—it lacks a reset condition, the ability to generate optical pulses, and truly asynchronous behavior. Several modifications to this bench-top model, including a delayed output pulse fed back to reset the SOA, make it suitable for preliminary experiments in learning (Chapter 12) and simple lightwave neuromorphic circuits (Section 4.2). The original fiber-based photonic neuron is hardly scalable to networks of many neurons, but it identifies a new domain of ultrafast cognitive computing, which has informed the development of more advanced photonic neuron devices based on excitable laser dynamics, a model that is described in Chapter 5. Furthermore, network-compatible versions of this circuit could be created, as discussed in Section 10.3.3.
Identification of cancer types from gene expressions using learning techniques
Published in Computer Methods in Biomechanics and Biomedical Engineering, 2022
Swati B. Bhonde, Sharmila K. Wagh, Jayashree R. Prasad
Cancer is also the most common cause of mortality worldwide, ranging from roughly one-sixth of fatalities. To reduce the effect of cancer on human health, significant research efforts have been dedicated to cancer detection and treatment techniques (Bahri et al. 2020). The goal of cancer detection is to classify tumor categories and establish indicators for each malignancy by a learning technique that automatically recognizes certain metastatic tumors or diagnose cancer in an early phase. Cancer prediction focuses on cancer susceptibility, recurrence, and prognosis by offering precise cancer treatment depending on unique genetic biomarkers (Li et al. 2020). The last decade has witnessed abundant use of DL algorithms, which has the exciting potential to uncover complicated interactions buried in large-scale information including bioinformatics (Luo et al. 2019). Although it is often considered synonymous with computational biology, bioinformatics is a discipline of science that is related but distinct from biological computation (Chiu et al. 2020). Bioinformatics uses computing to better understand biology, while biological computation uses bioengineering and biology to construct biological computers (Karim et al. 2020). The use of DL algorithms has grown rapidly in bioinformatics, demonstrating exciting abilities to show relationships concealed in extensive biological and biomedical evidence. DL is a class of multi-layer Neural Network models (NN) that progressively succeeds in the enormous amount of data learning (Mallik et al., 2020). It also comprises a training phase wherein the network characteristics are predicted from a training dataset and a testing phase in which the learned network is used to estimate subsequent outputs (Shanthi and Rajkumar 2021). The development of the DL model is used to improve accuracy and interpretability for cancer type prediction which is now made possible by the accumulation of whole transcriptomic profiling of tumor samples (Tabares-Soto, Orozco-Arias, et al.).