![]() When we make PEEK, we use a process which controls the length of the chains, or molecular weight. Designing to the needs of the application Compared to metals, PEEK-based materials are very light weight, easily shaped, resistant to corrosion and can have considerably higher specific strength (strength per unit weight). The resulting polymer is widely regarded as one of the highest performing thermoplastics in the world. The regular structure of the repeat unit means that PEEK molecules can partially crystallise, and crystallinity provides a combination of wear, creep, fatigue and chemical resistance – more on this later. The ether groups provide some degree of flexibility, for toughness, and like the aryl and ketone groups are unreactive, so providing resistance to chemical attack. The aryl and ketone groups are fairly rigid and provide stiffness which means good mechanical performance combined with a high melting point. The P comes from the Greek “poly” meaning many, so many EEKs make PEEK. This “repeat unit”, shown in the square brackets above, is replicated many times – on average somewhere between 200-300 times – to make a single PEEK polymer chain. Here’s how the building blocks fit together and thus we get Ether Ether Ketone or EEK: From a chemical point of view, PEEK is a largely linear, semi-crystalline polymer. PEEK itself however – like most thermoplastics – is odourless under normal conditions. Small molecules of this type, like toluene & naphthalene, have distinct odours, hence the name. "Aromatic", usually meaning distinctive or sweet-smelling, may seem a strange word here, but scientists use it to describe some molecules containing or made from ring-like structures (like the aryl building block above). ![]() We also show that these insights can potentially provide guidance on improving NN's performance.R&D into PAEKs has its origins in the 1960’s but it wasn’t until 1978 that ICI (Imperial Chemical Industries) filed their patent on PEEK which was first commercialised as Victrex PEEK polymer in 1981. With extensive experiments, we empirically show VRX can meaningfully answer "why" and "why not" questions about the prediction, providing easy-to-understand insights about the reasoning process. By means of knowledge distillation, we show VRX can take a step towards mimicking the reasoning process of NNs and provide logical, concept-level explanations for final model decisions. Given a trained classification model, the proposed VRX extracts relevant class-specific visual concepts and organizes them using structural concept graphs (SCG) based on pairwise concept relationships. In this work, we propose a framework (VRX) to interpret classification NNs with intuitive structural visual concepts. While recent developments in explainable artificial intelligence attempt to bridge this gap (e.g., by visualizing the correlation between input pixels and final outputs), these approaches are limited to explaining low-level relationships, and crucially, do not provide insights on error correction. Download a PDF of the paper titled A Peek Into the Reasoning of Neural Networks: Interpreting with Structural Visual Concepts, by Yunhao Ge and 7 other authors Download PDF Abstract:Despite substantial progress in applying neural networks (NN) to a wide variety of areas, they still largely suffer from a lack of transparency and interpretability.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |