X-ray Imaging through Machine LearningI 本周物理學講座

2021-02-21 中國物理學會期刊網

報告人:Ge Wang,Rensselaer Polytechnic Institute

時間:8月9日(周四)15:00

單位:中科院高能所

地點:化學樓C305

Computer vision and image analysis are great examples of machine learning, especially deep learning. While computer vision and image analysis deal with existing images and produce features of these images (images to features), tomographic image reconstruction produces images of internal structures from measurement data, which are various features (attenuated/non-attenuated line integrals, Fourier/harmonic components, echoed/scattered/transmitted ultrasound signatures, diffused/excited/interfered light signals, and so on) of the underlying images (features to images). Recently, machine learning, especially deep learning, techniques are being actively developed worldwide for tomographic image reconstruction, which is a new area of research as shown by the perspective https://ieeexplore.ieee.org/document/7733110, the 20 high-quality papers included in the latest TMI special issue https://ieee-tmi.org/fast-facts/featured-article.asp?id=30&title=Special-Issue-on-Machine-Learning-for-Image-Reconstruction,-June-2018, as well as similar publications in other journals and conferences. In addition to well-known analytic and iterative methods for tomographic image reconstruction, machine learning is an emerging approach for image reconstruction, and likewise image reconstruction is a new frontier of machine learning. There are exciting research and application opportunities ahead for smart imaging and precision medicine. 

報告人:Gang LI

時間:8月7日(周二)12:00

單位:北京大學

地點:1st floor meeting room, KIAA

We present our analysis of ∼450 Kepler γ Doradus stars. These stars are pulsating mainly in gravity modes, but the periods are no longer equally-spaced due to core rotation and molecular weight gradients. So far we have discovered ∼600 period spacing patterns from ∼450 stars. These results constitute the largest γ Dor sample thus it is possible to reveal some new statistical relations. For dozens of slowly-rotating stars, we show how rotation distorts the asymptotic relation. In rapid rotators, we find that g-mode period spacings and their gradient (slope) are both positively correlated with period. Additionally, Rossby modes are seen in tens of stars. They have a negative correlation between the slope and mean period, which is twice the mean period of prograde gravity modes.  These stars are very useful for understanding the physics of rotation and angular momentum transport in main sequence stars, and can place much needed constraints on diffusive mixing and chemical gradients, whilst also providing stellar ages. As this is first sample with significant Rossby modes, we have the unprecedented opportunity to study these unfamiliar modes, like their instability strip, excitation mechanism and surface rotation.

3Quantum Dynamics ?Space-time Crystal and Bethe String States

報告人:CongjunWu

時間:8月8日(周三)10:00

單位:北京大學

地點:Room W563, Physics Building

We present recent works on two different aspects of quantum dynamics ? symmetry and strong correlations. For the symmetry aspect, we propose a new concept of ?space-time? crystal as a general framework for studying intertwined space-time periodic structures, which include both the static crystal and the Floquet lattice as special cases. A new mathematic structure of ?space-time? group is constructed to describe the symmetries of a space-time crystal, which augments the space group with non-symmorphic operations involving fractional translations along the time domain: ?time-screw rotation? and ?time-glide reflection?. Classifications for the 1+1 D and 2+1D space-time crystals (groups) are completed, and their consequences on dynamic band structures will be discussed. For the strong correlation aspect, we have studied the real frequency response at high energy which is a hard core problem of condensed matter physics. We studied the role of Bethe-string states in the quantum spin dynamics in antiferromagnetic spin chains in high magnetic fields based on algebraic Bethe ansatz via the form-factor formulae. Close to quantum criticality, the string excitations govern the quantum spin dynamics, whereas the fractional excitations, which are dominant at low energies, reflect the antiferromagnetic quantum fluctuations. These states have been recently observed in the electron-spin-resonance spectroscopy measurement on SrCo2V2O8. This work is helpful for experimental studies on spin dynamics in both condensed matter and cold atom systems beyond the low energy effective Luttinger liquid theory.

報告人:Boaz Katz

時間:8月9日(周四)16:00

單位:北京大學

地點:KIAA-PKU Auditorium

How do supernovae explode? The simple truth is that we don't know.  While decades of observational and theoretical research have lead to many important breakthroughs in our understanding of these powerful explosions, for all we know, they need not occur. In this talk I will focus on type Ia supernovae and describe our recent ongoing efforts to find an answer to this long standing question.

1. 90歲的物理所,依然年輕!

2. 二維超導材料

3. 拓撲相和拓撲相變的量子模擬

4. 物理學咬文嚼字之九十八:Phase:a phenomenon

5. 對大學物理實驗教學的思考、改革與實踐

6. 物理像什麼?

7. 納維-斯託克斯方程:行到水窮處,坐看雲起時

8. 一個物理學家的萬米長跑和玻色-愛因斯坦凝聚

9. 相對論天體物理

10. 周末大家談——從選上學部委員到耄耋之年

相關焦點

  • 【預告】施汝為系列講座 | 武漢大學劉惠軍教授:Machine learning in the study of ...
    直播二維碼此次講座由合肥微尺度物質科學國家研究中心、國際功能材料量子設計中心
  • 量子幾何相位理論及其應用簡介 | 本周物理學講座
    薛丁格方程的提出,虛數i首次出現在物理學運動方程中,表明複數的引進不僅是數學工具的方便,更是量子力學描述世界的必須,相位開始成為物理學的核心概念。稍後,狄拉克引入波函數相位因子,並用它證明,如果世界上有一個磁單極,那麼電荷必須是量子化的,這一優美的理論把相位和拓撲聯繫在一起。1954年,楊振寧-米爾斯非阿貝爾規範場理論的提出,把U(1)相位推廣到SU(2)相位,構築了粒子之間相互作用的基石。
  • 加速器新原理 I 本周物理學講座
    地點:Conference Room 322We investigate the two- and three-body universal physics associated with the s-wave three-body system consisting of a cluster of two neutrons and one flavored meson, i.e
  • Machine Learning Cheat Sheet
    Gradient descent methodBatch gradient descent:Stochastic gradient descent:Choice of learning rateLoss function's changes
  • SlidePlayer課件精選:Basics of Chest X-Ray
    AP Films Obtaining Images Systematic method to reading CXR Common Signs Examples 3 Chest X-ray (CXR) Basics A standard chest X-ray consists of a – PA Image – Lateral Image – Images read together
  • 半導體低維結構中的量子相 | 本周物理學講座
    In this talk, firstly, I will describe a CMOS-compatible synthesis approach, i.e. gold-free indium-particle-catalyzed (or self-catalyzed) vapor-liquid-solid growth mode.
  • CFA二級思維導圖分享:機器學習(Machine Learning)
    CFA二級思維導圖分享:機器學習(machine learning)Reading7主要了解機器學習的一些常見概念,主要分類、了解常用算法的原理及其用途。常用的分類算法有支持向量機(Support vector machine (SVM))、近鄰算法(K-nearest neighbor (KNN) 、分類回歸樹(Classification and Regression Tree (CART)),以及集成算法,集成算法為將多種不同的算法或模型集成到一起,將各個不同模型的結果放到一起,按模型結果的最大值作為整個算法的結果,如Bootstrap Aggregating
  • 北語講座 | 李平教授系列講座
    How can we identify, predict, and promote rapid L2 learning-induced brain changes?Our studies help to identify (a) how neurocognitive changes occur as a function of learning contexts (traditional classroom vs. 3D virtual environment based learning), (b) how brain changes may capture
  • 聲子晶體材料:從聲石墨烯到聲拓撲絕緣體 I 本周物理學講座
    為了能夠讓大家更詳細的了解WITec 300R拉曼光譜儀的功能,更好的掌握並充分利用這一設備開展相關科研工作,解決用戶使用過程中的問題,拉曼實驗室邀請WITec中國子公司的應用工程師胡海龍博士為用戶提供講座。
  • Osmosis: 胸片判讀方法 Reading a Chest X-ray(附視頻連結)
    Let's go through this chest x-ray using an easy-to-remember checklist associated with the first seven letters of the alphabet: ABCDEFG.我們來用一個速記列表來幫助判讀這張胸片:以字母表的前七個字母作為口訣:ABCDEFG ——
  • 基於半導體單電子器件的量子計算 | 本周物理學講座
    Based on the intensive redshift survey, we explore the structure of local massive cluster Abell 2029 and its surroundings along with X-ray and weak lensing observations.
  • 【新開放書】醫學影像原理與應用,Medical Imaging Principles and Applications
    https://www.intechopen.com/books/medical-imaging-principles-and-applications/the-application-of-functional-magnetic-resonance-imaging-in-neuropathic-pain
  • 淺析射線檢測 raycast 的使用!Cocos Creator 3D!
    _point.x, touch._point.y, this._ray);接下來就是檢測這條線穿過了哪些物體啦。creator 3d 提供了三種檢測方案,可以一起看看是如何使用的。基於物理碰撞器的射線檢測:我們先給需要檢測的物體添加碰撞器。在代碼中調用物理碰撞器檢測方法就可以啦。
  • 【徵稿】Evolutionary Transfer Learning and Transfer Optimisation
    In machine learning, transfer learning aims to transfer knowledge acquired in one problem domain, i.e. the source domain, onto another domain, i.e. the target domain.
  • Strong Field QED | 本周物理學講座
    I will present the measurements of 32 X-ray selected AGNs (1.2<z<1.7), whose host mass are inferred using HST imaging data and BH masses are estimated using Balmer lines.
  • 《Machine Learning in Action》——白話貝葉斯,「恰瓜群眾」應該...
    基於屬性條件獨立性假設,式子3-1,我們可以重寫為:\[P(c|x)=\frac{P(c)P(x|c)}{P(x)}=\frac{P(c)}{P(x)}\Pi_{i=1}^8P(x_i|c) \tag{3-2} \]而我們知道,對於一個檢測西瓜樣本來講,該樣本每一個屬性特徵值在每個樣本類別裡計算得到的結果都是一樣的
  • npj:機器學習——用於匹配X射線吸收譜的Web應用程式
    該文近期發表於npj Computational Materials  4: 12 (2018); doi:10.1038/s41524-018-0067-x。英文標題與摘要如下,點擊閱讀原文可以自由獲取論文PDF。
  • 長知識‖X-RAY在分選設備上的原理及應用
    一、x-ray分選設備原理1、首先X-RAY設備這個裝備主要是利用X光射線的穿透作用,X光射線波長很短,能量特別大,照在物質上時,物質只能吸收一小部分,而大部分X光射線的能量會從物質原子的間隙中穿過去,表現出極強的穿透能力。