Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userMitigating print-through effects through an optimized method for CFRP mirror production in Chile
66
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
In the manufacturing process of Carbon Fiber Reinforced Polymer (CFRP) mirrors (replicated from a mandrel) the orientation of the unidirectional carbon fiber layers (layup) has a direct influence on different aspects of the final product, like its general (large scale) shape and local deformations. In particular, optical methods used to evaluate the surfaces quality, can reveal the presence of print-through, a very common issue in CFPR manufacture. In practical terms, the surfaces irregularities induced, among other artifacts, by print-through, produce unwanted scattering effects, which are usually mitigated applying extra layers of different materials to the surface. Since one of the main goals of CFPR mirrors is to decrease the final weight of the whole mirror system, adding more material goes in the opposite direction of that. For this reason a different layup method is being developed with the goal of decreasing print-through and improving sphericity while maintaining mechanical qualities and without the addition of extra material in the process.
rate research
Read More
Convolutional neural networks have demonstrated high accuracy on various tasks in recent years. However, they are extremely vulnerable to adversarial examples. For example, imperceptible perturbations added to clean images can cause convolutional neural networks to fail. In this paper, we propose to utilize randomization at inference time to mitigate adversarial effects. Specifically, we use two randomization operations: random resizing, which resizes the input images to a random size, and random padding, which pads zeros around the input images in a random manner. Extensive experiments demonstrate that the proposed randomization method is very effective at defending against both single-step and iterative attacks. Our method provides the following advantages: 1) no additional training or fine-tuning, 2) very few additional computations, 3) compatible with other adversarial defense methods. By combining the proposed randomization method with an adversarially trained model, it achieves a normalized score of 0.924 (ranked No.2 among 107 defense teams) in the NIPS 2017 adversarial examples defense challenge, which is far better than using adversarial training alone with a normalized score of 0.773 (ranked No.56). The code is public available at https://github.com/cihangxie/NIPS2017_adv_challenge_defense.
Planet Formation research is blooming in an era where we are moving from speaking about protoplanetary disks to planet forming disks (Andrews et al., 2018). However, this transition is still motivated by indirect (but convincing) hints. Up to date, the direct detection of planets in the making remains elusive with the remarkable exception of PDS70b and c (Haffert et al., 2019; Keppler et al., 2018; Muller et al., 2018). The scarcity of detections is attributable to technical challenges, and even for the rare jewels that we can detect, characterization (resolving their hill spheres) is unachievable. The next step in this direction demands from near to mid-infrared interferometry to jump from $sim$100 m baselines to $sim$1 km, and from very few telescopes (two to six) to 20 or more (PFI like concepts, Monnier et al. 2018). This transition needs for more affordable near to mid-infrared telescopes to be designed. Since the driving cost for such telescopes resides on the primary mirror, in particular scaling with its diameter and weight, our approach to tackle this problem relies on the production of low-cost light mirrors.
The surface quality of replicated CFRP mirrors is ideally expected to be as good as the mandrel from which they are manufactured. In practice, a number of factors produce surface imperfections in the final mirrors at different scales. To understand where this errors come from, and develop improvements to the manufacturing process accordingly, a wide range of metrology techniques and quality control methods must be adopted. Mechanical and optical instruments are employed to characterise glass mandrels and CFRP replicas at different spatial frequency ranges. Modal analysis is used to identify large scale aberrations, complemented with a spectral analysis at medium and small scales. It is seen that astigmatism is the dominant aberration in the CFRP replicas. On the medium and small scales, we have observed that fiber print-through and surface roughness can be improved significantly by an extra resin layer over the replicas surface, but still some residual irregularities are present.
Current large-scale language models can be politically biased as a result of the data they are trained on, potentially causing serious problems when they are deployed in real-world settings. In this paper, we describe metrics for measuring political bias in GPT-2 generation and propose a reinforcement learning (RL) framework for mitigating political biases in generated text. By using rewards from word embeddings or a classifier, our RL framework guides debiased generation without having access to the training data or requiring the model to be retrained. In empirical experiments on three attributes sensitive to political bias (gender, location, and topic), our methods reduced bias according to both our metrics and human evaluation, while maintaining readability and semantic coherence.
Automatic detection of toxic language plays an essential role in protecting social media users, especially minority groups, from verbal abuse. However, biases toward some attributes, including gender, race, and dialect, exist in most training datasets for toxicity detection. The biases make the learned models unfair and can even exacerbate the marginalization of people. Considering that current debiasing methods for general natural language understanding tasks cannot effectively mitigate the biases in the toxicity detectors, we propose to use invariant rationalization (InvRat), a game-theoretic framework consisting of a rationale generator and a predictor, to rule out the spurious correlation of certain syntactic patterns (e.g., identity mentions, dialect) to toxicity labels. We empirically show that our method yields lower false positive rate in both lexical and dialectal attributes than previous debiasing methods.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
