Explainable Deep Learning for the Analysis of MHD Spectrograms in Nuclear Fusion

In the nuclear fusion community, there are many specialized techniques to analyze the data coming from a variety of diagnostics. One of such techniques is the use of spectrograms to analyze the magnetohydrodynamic (MHD) behavior of fusion plasmas. Physicists look at the spectrogram to identify the oscillation modes of the plasma, and to study instabilities that may lead to plasma disruptions. One of the major causes of disruptions occurs when an oscillation mode interacts with the wall, stops rotating, and becomes a locked mode. In this work, we use deep learning to predict the occurrence of locked modes from MHD spectrograms. In particular, we use a Convolutional Neural Network (CNN) with Class Activation Mapping (CAM) to pinpoint the exact behavior that the model thinks is responsible for the locked mode. Surprisingly, we find that, in general, the model explanation agrees quite well with the physical interpretation of the behavior observed in the spectrogram.

The Cyclical Sine Model Explanation for Climate Change

This paper will show that the global warming/climate change underway on Earth today is a totally natural occurrence with solid scientific and historical support. The Earth is currently in the upswing part of its normal temperature cycle. Very warm (Medieval Warming) and very cold (Little Ice Age) cycles have been historically documented on Earth for at least the last 3,000 years. This cyclicity has a repeated period of approximately every 1,500 years [1]. The explanation for the Earth’s temperature increases since 1850 is captured in a mathematical model called the Cyclical Sine Model. This model fits past climate cycles, measured temperatures since 1850, and correlates closely with the thousand year cyclicity of solar activity from 14C/12C ratio studies [2], and Bond [3] Atlantic drift ice cycles. This model also agrees with sunspot history, the Atlantic Multidecadal Oscillation, and the Pacific Decadal Oscillation. In addition, this model quantitively explains the time span 1945-1975 when an impending ice age was feared [4]. Earth temperatures are controlled by three solar cycles of approximately 1,000, 70, and 11 years. The Cyclical Sine Model is the best explanation for the Earth’s recent temperature increases.

Exploring in the Latent Space of Design: A Method of Plausible Building Facades Images Generation, Properties Control and Model Explanation Base on StyleGAN2

GAN has been widely applied in the research of architectural image generation. However, the quality and controllability of generated images, and the interpretability of model are still potential to be improved. In this paper, by implementing StyleGAN2 model, plausible building façade images could be generated without conditional input. In addition, by applying GANSpace to analysis the latent space, high-level properties could be controlled for both generated images and novel images outside of training set. At last, the generating and controlling process could be visualized with image embedding and PCA projection method, which could achieve unsupervised classification of generated images, and help to understand the correlation between the images and their latent vectors.

Towards Fair and Transparent Algorithmic Systems

My research in the past few years has focused on fostering trust in algorithmic systems. I often analyze scenarios where a variety of desirable trust-oriented goals must be simultaneously satisfied; for example, ensuring that an allocation mechanism is both fair and efficient, or that a model explanation framework is both effective and differentially private. This interdisciplinary approach requires tools from a variety of computer science disciplines, such as game theory, economics, ML and differential privacy.

CausaLM: Causal Model Explanation Through Counterfactual Language Models

Understanding predictions made by deep neural networks is notoriously difficult, but also crucial to their dissemination. As all ML-based methods, they are as good as their training data, and can also capture unwanted biases. While there are tools that can help understand whether such biases exist, they do not distinguish between correlation and causation, and might be ill-suited for text-based models and for reasoning about high level language concepts. A key problem of estimating the causal effect of a concept of interest on a given model is that this estimation requires the generation of counterfactual examples, which is challenging with existing generation technology. To bridge that gap, we propose CausaLM, a framework for producing causal model explanations using counterfactual language representation models. Our approach is based on fine-tuning of deep contextualized embedding models with auxiliary adversarial tasks derived from the causal graph of the problem. Concretely, we show that by carefully choosing auxiliary adversarial pre-training tasks, language representation models such as BERT can effectively learn a counterfactual representation for a given concept of interest, and be used to estimate its true causal effect on model performance. A byproduct of our method is a language representation model that is unaffected by the tested concept, which can be useful in mitigating unwanted bias ingrained in the data.

Universe at Zero Second and The Creation Process of The First Particle from the Absolute Void

In this study, we discuss the properties of absolute void space or the universe at zero seconds, and how these properties play a vital role in creating a mechanism in which the very first particle gets created and we find the limit in which when the absolute void volume reaches will lead to the collapse that leads to the creation of the first particle. Later we discuss the standard model explanation through the elementary dimensions theory, as according to the elementary dimensions theory study that was peer-reviewed at the end of 2020, everything in the universe is made from four elementary dimensions, these dimensions are the three spatial dimensions (X, Y, and Z) and the force equivalent.