Things are starting to get weird with AI, that alternate reality in sci-fi movies where machines become superior and take over may not actually be so impossible after all. According to researchers from Loughborough University, Western General Hospital, the University of Edinburgh, and the Edinburgh Cancer Centre in the United Kingdom, they have recently developed a deep learning-based method that can analyze and detect compounds in human breath.
<img src="https://media.8ch.net/file_store/89d5fd6f8518c45b15a3066f08d621ca900a6b46468d92e18ae4269f4841820f.png" style="max-height:640px;max-width:360px;">
<span style="margin-top:15px;rgba(42,51,6,0.7);font-size:12px;">James Gathany/Smithsonian.com</span>
Related coverage: <a href="https://thegoldwater.com/news/27689-New-Robot-Hybrid-Grows-It-s-Own-Living-Muscles">New Robot Hybrid Grows It's Own Living Muscles</a>
The deep-learning AI is so successful it can even detect illnesses such as cancer with greater accuracy than average human performance. In an article referring to human smell in the Smithsonian, one researcher said, "The sense of smell is used by animals and even plants to identify hundreds of different substances that float in the air. But compared to that of other animals, the human sense of smell is far less developed and certainly not used to carry out daily activities. For this reason, humans aren’t particularly aware of the richness of information that can be transmitted through the air, and can be perceived by a highly sensitive olfactory system."
The researchers in Loughborough University and beyond used NVIDIA Tesla GPUs and the cuDNN-accelerated Keras, and TensorFlow deep learning frameworks to train their neural network on data from participants with varying forms of cancer who were receiving radiotherapy. A Ph.D. research student at Loughborough University named Angelika Skarysz said the team increased the original training data with data augmentation to increase the neural network's efficiency. The team reported augmenting the neural network 100 times.
Related coverage: <a href="https://thegoldwater.com/news/21708-Arizona-Governor-Bans-Driverless-Cars-Nvidia-Uber-And-Tesla-All-Suffer">Arizona Governor Bans Driverless Cars - Nvidia, Uber And Tesla All Suffer</a>
The researchers said, "This is the first successful machine learning attempt at learning ion patterns and detecting compounds from raw GC-MS data," the team said. "The convolutional neural network achieved the best performance when implemented with two particular features: one-dimensional filters to adapt to the particular structure of GC-MS data, and a three-channel input to read high, medium, and low-intensity signals from the highly variable GC-MS spectrum. The novel approach was shown to discover labeling errors from human experts, suggesting better-than-human average performance."
They also used the NVIDIA Tesla GPUs for an inference which was scanning the breath samples. Soltoggio said, "Computers equipped with this technology only take minutes to autonomously analyze a breath sample that previously took hours by a human expert." She also added that the whole process is cheaper and is making it more reliable. The team's full report can be found on <a href="https://www.researchgate.net/publication/324921031_Convolutional_neural_networks_for_automated_targeted_analysis_of_raw_gas_chromatography-mass_spectrometry_data">Research Gate</a> and it is due to be presented at the International Joint Conference on Neural Networks.
Tips? Info? Send me a message!