Huaqiu PCB
Highly reliable multilayer board manufacturer
Huaqiu SMT
Highly reliable one-stop PCBA intelligent manufacturer
Huaqiu Mall
Self-operated electronic components mall
PCB Layout
High multi-layer, high-density product design
Steel mesh manufacturing
Focus on high-quality steel mesh manufacturing
BOM ordering
Specialized Research a Ethiopians Escort one-stop purchasing solution
Huaqiu DFM
One-click analysis of hidden design risks
Huaqiu Certification
The certification test is beyond doubt
Recently, three young Chinese AI PhDs have the latest personal work plans: Chen Tianqi and Zhu Junyan both announced that they will attend CMU in the spring of 2020 , serves as an assistant professor; Jin Chi announced that he will join Princeton in September this year as an assistant professor.
Excellent, may be a habit.
Recently, the personal work plans of three Chinese AI doctoral students have made the latest progress:
XGBoost author Chen Tianqi announced that he will join CMU in the spring of 2020 as an assistant professor;
CycleGAN author Zhu Junyan announced that he will join CMU in the spring of 2020 as an assistant professor;
Jin Chi, the winner of the “RL Exploration” Best Paper Award of the ICML 2018 Research Society, announced that he will join Princeton in September 2019 as an assistant professor.
The doctoral lives of three AI mastersET Escorts is coming to an end, and what awaits them is more challenging and meaningful scientific research work. As Chen Tianqi and Zhu Junyan said on Twitter, they both Hope to make more contributions in their respective fields.
Congratulations!
XGBoost and TVM author Chen Tianqi joined CMU
PhD student at the University of Washington, famous machine learning algorithm XGBoost, in-depth study Chen Tianqi, the author of the compiler TVM and others, recently announced on Twitter that he will join the CMU Machine Learning Department as an assistant professor in the spring of 2020.
Chen Tianqi said:
I will join CMU in the spring of 2020. Serves as an Assistant Professor. I am very grateful to my mentors, co-authors, and the University of Washington for their assistance in my doctoral career. I look forward to working with my CMU colleagues to conduct more research on future intelligent systems.
Dr. Chen Tianqi studied at the Paul G. Allen School of Computer Science and Engineering at the University of Washington, and received his master’s and bachelor’s degrees from the ACM class at Shanghai Lukang University.
Chen Tianqi
Google AI leader Jeff Dean, GAN inventor Ian Goodfellow, CMU associate professor Ma Jian and others all expressed their congratulations
AI expert Li Mu also revealed a little about Chen Tianqi’s joining CMU on Zhihu:
Two sighs:
1. CMU started to do joint research on system and machine learning very early (around 2011). After a lot of exploration and detours in the later period (missed several supEthiopians Sugardaddyer star), finally recruited the most suitable young teacher. Tianqi was still hesitating about whether to go to CMU, but he finally decided. A great blessing for CMU.
2. It is relatively straightforward to understand Tianqi’s path to finding a faculty. Having a good job is a prerequisite, but visiting the other school is also important. At least you go to the host for the interview. I really appreciate your work. For example, I really like the work of TVM, and several members of our group are invested in TVM, so I often go out to promote it. Through the process, I found that my previous boss of CMU also liked it (Tianqi). CMU interview host), my old office friend at CMU liked it very much (the host of another offer that Tianqi struggled with), and the boss of our group TVM tech leader also liked it (another big boss from a top school who made the offer) ). In addition, it is a pity that there are several other veryStrong schools did not follow up. Personally, I really hope that Tianqi can go to Stanford or Berkeley, so that we can work together more conveniently in the future (CMU is actually too remote, and the transfer from the Bay Area to Pittsburgh seems to be gone). It’s no surprise that I didn’t get it. When I chatted with some of their senior teachers before, they didn’t seem to be interested in this area Ethiopians Escort.
(https://www.zhihu.com/question/329657835/answer/717971252)
“Having good tasks is a condition.” In his doctoral life, Chen Tianqi did many good things that were widely used. On his homepage, he writes: “I am very interested in cross-research on machine learning and systems. What’s really exciting about this field is that when we combine advanced machine learning techniques and systems, What can be achieved? In addition, I am also promoting the direction of in-depth learning, knowledge transfer and lifelong learning.”
Chen Tianqi won the Google Ph.D. Fellowship, the 2012 KDDCup Champion, and the 2011 KDDCup. Second place and other awards, and multiple papers awarded at top artificial intelligence conferences such as ICML, ICLR, NeurIPS, and KDD.
Chen Tianqi Google scholar citation data
Chen Tianqi and The team developed three widely adopted machine learning systems:
TVM stack: an automatic end-to-end optimizing compiler for deep learning.
XGBoost, a scalable, end-to-end tree boosting system. This tool has become one of the tools that data scientists use every dayEthiopia Sugar Daddy.
Apache MXNet (co-author): One of the important deep learning frameworks currently used by AWS.
XGEthiopia Sugar DaddyBoost
Let’s look at XGBoost first. At that time in Distributed (Deep)Chen Tianqi from the Machine Learning Community (DMLC) group is in charge.
In 2014, Chen Tianqi, who was a Ph.D. in Washington at the time ET Escorts, open sourced the XgBoost algorithm and was sought after by the public. Agility has also become a frequent visitor in Kaggle competitions.
To this day, XgBoost is still widely used in competitions and performs very well. It is featured in many winning plans and is known as a weapon for winning competitions.
TVM Stack
Another major task of Chen Tianqi is TVM.
TVM is a neural network compiler that can directly compile models trained by other frameworks into executable code on the target platform, with significant optimization in speed and memory. TVM enables automatically or semi-automatically generated code to achieve the effect of handwritten code.
The framework of TVM
Chen Tianqi puts TVM+ NNVM is described as “a complete optimized tool chain from deep learning to various hardware.” He explained TVM clearly on Zhihu:
TVM attempts to summarize the manual optimization experience of deep learning OP from a higher abstract level, so that users can quickly explore in an automatic or semi-automatic way. Efficient op completion space.
TVM is different from existing solutions. Taking XLA as an example, TVM has taken a more traditional technical path than the current XLA. tvm can be used to make it easier to achieve the performance required by XLA: Existing solutions Designing your own rule transformation based on advanced graph representation can produce some graph-level combinatorial op optimizations, such as conv-bn fusion, but you still have to rely on handwritten rules to achieve the step from graph representation to code. The OP of the graph shows that there are too many things that the code itself can choose from, such as how to do threads and how to use shared memory, but most of them are not described in the graph language, making it difficult to automate. In this way, the bottleneck of the deep learning system will inevitably change from the complexity of op implementation to the complexity of implementing the model generation requirements in the graph compiler. Taking this direction requires the support of a very large engineering team, and we hope to achieve the same or even better results with less manpower.
We have taken a long-term technical path that is more risky but also has greater rewards. Simply put, TVM works by converting images toThe step of op generation rules is further abstracted step by step, and the generation rules themselves are divided into various operation primitives, which can be combined when needed. Based on tvm, we can quickly combine different schedule plans.
Link: https://www.zhihu.com/quET Escortsestion/64091792/answer/217722459
In addition, Chen Tianqi, together with Li Mu, is one of the important initiators and important contributors of DMLC (Distributed (Deep) Machine Learning Common) and its important project MXNet.
After joining CMU, Ethiopians Sugardaddy what tasks Chen Tianqi will continue to deliver is exciting. Some people say that CMU is likely to usher in the first AI systems course in history.
“Pioneer”: Zhu Junyan got a teaching position at CMU at only 30 years old
At the same time, another AI master Zhu Junyan also said that he will return to CMU in the spring of 2020 as an assistant professor One position.
On June 14, Zhu Junyan announced on his Twitter:
After spending a wonderful time at the University of California, Berkeley and MIT CSAIL, I will be studying at CSAIL in 2020 Returned to CMU as an assistant professor in the spring of next year. I look forward to doing more work at the intersection of graphics, vision, and machine learning.
In 2012, Zhu Junyan obtained a bachelor’s degree in engineering from the Computer Science Department of Tsinghua University. Then he went to CMU and UC Berkeley to study. After 5 years of study, Zhu JunEthiopians Sugardaddyyan obtained a bachelor’s degree in electrical engineering from UC Berkeley in 2017 with the Department of Computer Science (his supervisor is Alexei Efros, and doctoral research was supported by a Facebook scholarship).
His doctoral thesis, Learning to Generate Images, won the ACM SIGGRAPH 2018 “Outstanding Doctoral Thesis Award” at the top computer graphics conference.
Dr. Zhu Junyan can be said to be the pioneer of modern machine learning applications in the field of computer graphics. His paper can be said to be the first paper to use deep neural network systems to deal with natural image decomposition problems.
Therefore, his research has great significance in this field Ethiopia Sugar has had a serious impact. Some of his scientific research results are not only used by researchers in fields such as computer graphics, but also become widely used tools by visual artists.
Among them, the most well-known is CycleGAN.
CycleGAN uses pixel2pixel technology to automatically convert a certain type of image into another type of image. It is so truly natural that it can be said to be one of the most followed models in 2017.
Not only CycleGAN, Zhu Junyan also has many well-known scientific research results.
NVIDIA released an amazing image generator-GauGAN at GTC 2019. It can be said that with a few lines, the sketch turns into a landscape in seconds Ethiopians Sugardaddy photo. Zhu Junyan is one of the authors.
Paper address Ethiopians Sugardaddy: https://arxiv.org/pdf/1903.07291. pdf
This software compiles human drawing methods and processes to create sketches in seconds and convert them into realistic photos. From early demos of the software, it appears it can do just that.
In addition, the recently popular MIT ten-dollar “Thanos” gloves are also Zhu Junyan’s masterpiece.
This magical glove, called the “scalable tactile glove” (STAG), uses flexible materials and is equipped with 550 micro-sensors on almost the entire hand. Using only tactile data, the AIET Escorts system can identify objects with an accuracy of up to 76%.
The experiment also proved that a large amount of pressure and its spatial resolution are the keys to successfully identifying targets. And similar deep learning models can estimate the weight of unknown objects. The results show that most objects weighing less than 60 grams can be accurately estimated.
This task was also published in the international scientific and technological journal “Nature”.
Paper address:
https://www.nature.com/articles/s41586-019-1234-z
In addition to the above-mentioned very well-known achievements, Zhu Junyan His other work is also very outstanding. For more scientific research results, please visit belowPersonal homepage:
Zhu Junyan’s personal homepage:
http://people.csail.mit.edu/junyanz/#sect-awards
In 2018, Zhu Junyan was awarded the UC Berkeley award Sakrison Memorial Prize, and Nvidia’s Pioneer Research Award.
He has also received the following awards and scholarships:
ACM SIGGRAPH Outstanding Doctoral Dissertation Award (2018)
Berkeley EECS (Ethiopians Sugardaddy2018) David J. Sakrison Outstanding Doctoral Research Memorial Award
NVIDIA Pioneer Research Award (2018)
Facebook Scholarship (2015)
Tsinghua University Outstanding Undergraduate Thesis (2012)
Outstanding Undergraduate Student at Tsinghua University (2012)
National Scholarship, issued by the Ministry of Education of China (2009 and 2010)
Singapore Science and Technology Engineering China Scholarship (2010, 2011 and 2012)
Jin Chi: Peking University student went to teach at Princeton
Jin Chi is currently a researcher in the Department of Electrical Engineering and Computer Science (EECS) at the University of California, Berkeley Sixth-year doctoral student, supervised by Michael I. Jordan. He is also a member of RISELab and the Berkeley Artificial Intelligence Research Institute (BAIR). Prior to this, Jin Chi obtained a bachelor’s degree in physics from Peking University and completed his undergraduate thesis under the leadership of Professor Wang Liwei.
Jin Chi’s research interests lie in machine learning, statistics and optimization. An important goal of his doctoral research is to design better learning algorithms that are physically sound and efficient in terms of sample complexity, run time, and space. To achieve this goal, his research focuses on providing a deeper understanding of fundamental problems in non-convex optimization, and more recently on reinforcement learning.
Jin Chi will join the Department of Electrical Engineering at Princeton University in September 2019 as an assistant professor. He will also attend IAS’s “Special Year in Optimization, Statistics and Practical Machine Learning” in the spring of 2019.
The paper written by Jin Chi and his supervisor Michael I. Jordan:
On Gradient Descent Ascent for Nonconvex-Concave Minimax Problems (about the gradient descent problem of non-convex-concave minimax problems)
StocEthiopia Sugarhastic Gradient Descent Escapes Saddle Points Efficiently (random gradient descent effectively escapes saddle points)
A short note on concentration inequalities for random vectors with subgaussian norm (Long exposition of concentration inequalities for random vectors with subgaussian norm)
Minmax Optimization: Stable Limit Points of Gradient Descent Ascent are Locally Optimal (Minmax Optimization: Stable Limit Points of Gradient Descent Ascent are Locally Optimal)
Sampling can be faster than optimization (sampling can be faster than optimization)
On the local minima of the empirical risk (the part that experiences risk is extremely small)
Is q-learning provably efficient ? (Can q-learning be proven effective?) – The best paper of the 2018 ICML Research Conference “RL Exploration”
Stochastic cubic regularization for fast nonconvex optimization (Stochastic cubic regularization for fast nonconvex optimization )
Accelerated gradient descent escapes saddle points faster than gradient descent (Accelerated gradient descent escapes saddle points faster than gradient descent)
How to escape saddle points eEthiopia Sugar Daddyfficiently (how to escape saddle points effectively)
Gradient descent can take exponential time to escape saddle points (Gradient descent can take exponential time to escape saddle points) For more papers, visit Jinchi’s Google Scholar homepage:
https://scholar.google.com/citations?user=GINhGvwAAAAJ&hl=en
News:
In October 2018, he gave an invited speech at the 56th Allerton Conference.
In July 2018, the paper “Can Q-Learning be proven useful?” won the best paper award in the “RL Exploration” of the ICML 2018 Research Conference.
In July 2018, the ICML 2018 seminar “Modern Trends in Non-convex Optimization of Machine Learning” was jointly held.
July 2017, blog post on how to effectively escape from saddle points.
Educational experience:
2013 to present
University of California, Berkeley
Doctoral student in computer science
2012-2013
ET Escorts University of Toronto
Visiting Professor of Statistics
2008-2012
Peking University
Bachelor of Physics
What happened during the practice Ethiopians Escort Situation:
Summer 2016
Microsoft Research Institute, Redmond
Dong Yu Research Intern
Summer 2015
Microsoft Research Institute, New England
Sham Kakade Research Intern
Jin Chi’s personal homepage:
https://sites.google.com/view/cjin/home
Original title: [AI Rising Star Shining in Famous Schools] Chen Tianqi, Zhu Junyan, and Jin Chi join the alliance CMU, Princeton
Article source: [Microelectronic signal: AI_era, WeChat public account: Xinzhiyuan] Welcome to add tracking and follow! Please indicate the source when transcribing and publishing the article.
The difference between three-and-a-half-digit and four-and-a-half-digit multimeters. Three-and-a-half-digit multimeters and four-and-a-half-digit multimeters are two types of digital multimeters commonly used in the field of electronic measurement. They are different in terms of measurement accuracy, display method, and application range. There are certain differences. This article will analyze the differences between these two multimeters in detail from many aspects. 1. Basic concepts of three-and-a-half-digit multimeter three-digit Issued on 07-14 14:23 •2416 views
JacEthiopia Sugarob: What kind of talent will be at Character.aiHave a long chat with the robot inside? “We use it when we get confused, like at 2 o’clock in the morning, especially when we can’t talk to any friends or a real therapist,” a true quote from a young user of Character.ai message. When we need to talk, there is no one to talk to. This was published on 07-14 08:05 •185 views
Kai-fu Lee looks forward to the future of AI: 2025 AI or super doctorate degree will be held on June 29 At the 10th Zhihu SaltEthiopians SugardaddyClub New Knowledge Youth Conference, technology giant Kai-fu Lee, who is also the CEO and founder of Zero One Thousand Things The dual composition of the chairman of Yifang Factory expressed his in-depth views on the future of artificial intelligence (AI). He recalled the 40 years he spent with AI Published on 06-29 15:56 •531 views
What are the characteristics of the three-position polarization relay? Polarizing relay is a special type of relay that has three positions: normal position, polarization position and reverse polarization position. This kind of relay is widely used in power systems, automation control, communication equipment and other fields. This article will introduce in detail the characteristics of the three-position polarization relay, including its working principle Published on 06-24 09:34 •234 views
Ali Alibaba announced the distribution of approximately US$4 billion in dividends. The financial report showed that AI-related revenue achieved triple-digit year-on-year growth. Alibaba announced the distribution of approximately US$4 billion in dividends. The financial report showed that AI-related revenue achieved triple-digit year-on-year growth. A few days ago, Alibaba Group released Q4 and full-year results for fiscal year 2024. Annual performance data; the financial report shows that Alibaba’s performance exceeded market expectations, and Alibaba Group’s Q4 revenue was 221.874 billion yuan Published on 05-15 12:13 •689 views
Two Chinese experts were selected as academicians of the Canadian Academy of Engineering, and the results of enabling sensor manufacturing were officially announced. This year, 53 academicians were elected. According to the observation of the laser industry, two Chinese experts in the laser field were selected this time, namely: Dr. James Chen is a member of the Canadian Ministry of Natural Resources. Published on 05 -14 17:02 •287 views
Basic structure and tasks of three-position three-way solenoid valvePrinciple: The three-position three-way solenoid valve is a commonly used control execution component and is widely used in various automation fields. They are usually used to control the flow of liquids and bodies in hydraulic and pneumatic systems and achieve various control functions. This article will introduce in detail the basic structure of the three-position three-way solenoid valve and Published on 01-24 11:00 •3920 views
How to select three positions in matlab Each digit of a number In MATLAB, you can apply some basic arithmetic and logical operations to obtain each digit of a three-digit number. The above is a basic method: Step 1: Enter a three-digit number. First, you need to enter a three-digit number in MATLAB. You can use ‘inp Published on 12-28 14:44 •1660 views
What are the ACT_Z ACT_Y ACT_X bits in ADXL355 0X24 REGISTER used for? When measuring the three-axis acceleration value, do you need to set all three parameters to 1? May I ask what the ACT_ZACT_YACT_X bit in the ADXL3550X24 REGISTER is used for? Do you need to set these three bits to 1 when measuring the three-axis acceleration value? Ethiopians SugardaddyThanks! Published on 12-28 07:21
Holtek HT45F5Q controls a three-digit seven-segment display. Please Ethiopians Escort and use Holtek HT45F5Q I can control the three-digit seven-segment display through three PINs. Currently, there is no way to display three different numbers. I would like to ask if you have any ideas or if there is a problem with my program. Thanks
Baidu Smart The first Demo Day event of the Cloud AI Accelerator was successfully held in Beijing. Automated and intelligent enterprise AI intelligent legal affairs, private health twins, AI personal work planning assistant that breaks the job information gap, accurate and efficientEthiopians Escort, reliable AI wealth consultant… Issued on 12-13 16:00 •716 Views
Should we give up the embedded path? Should we give up the embedded path? Embedded is a stable and promising field, but we have to decide whether we should stick to this path.The road needs to comprehensively consider multiple reasons such as personal preferences, personal work plans and market demand. The following is a discussion on this issue to help everyone make better decisions ~ Embedded Published on 12-07 10:43
Participate in [2023 Electronic Engineers Conference and the 3rd Annual Community Awards 】The feeling of the conference will be unforgettable: 3. During the roundtable dialogue, I was fortunate to listen to some of the suggestions, personal work plans, and personal work plans of the six top bosses on the engineers. ET Escorts‘s words and deeds on starting a business: 4. Zhang Fei’s practical experience in electronics, Baiwen Technology, Xiaoyan Technology, Wandian Atom, and Opening Ethiopia Sugar Published by Hongzhigu, Qiliguangguo, Runkaihong, and Youbo on 11-24 08:40
#Build a technical community for FPGA developers to FPGA Ecosystem Like#+2023.1Ethiopians Escort1.7+IC engineers talk about the workplace: from entrance interview to promotion and salary increase…. ..to other industries. The demand itself is accumulated. As the saying goes, it takes three months to change jobs and three years to change careers. To increase efforts in learning, the first thing to solve is to keep up and not fall behind. Anxiety is useless, study more. 4. IC engineer’s personal work plan: How to find the path to unfettered wealth? Unfettered wealth is a wonderful thing Published on 11-07 10:06
Live broadcast review | Training manual for IC people, how to plan for personal work On September 25, the first lesson of the “Core Star Project” integrated circuit started The online course started successfully. Industry elites from the “Core StarEthiopia Sugar Daddy Project” project team, Acceleration Technology and other integrated circuit fields gathered in the live broadcast room to focus on the integrated circuit industry Discussions on topics such as development trends, job markets, unemployment trends and personal work plans, based on Published on 09-27 13:22 •1780 views