This document is a presentation by Ted Chang about creating new opportunities for Taiwan's intelligent transformation. It discusses paradigm shifts in technology such as mobile phones and cloud computing. It introduces concepts like the Internet of Things, artificial intelligence, and how they can be combined. It argues that key driving forces for the future will be machine learning, big data, cloud computing and AI. The presentation envisions applications of these technologies in areas like future medicine and smart manufacturing. It ends by emphasizing the importance of wisdom and intelligence in shaping the future.
- The document discusses how artificial intelligence can enable earlier and safer medicine.
- It provides background on the author and their expertise in biomedical informatics and roles as editor-in-chief of several academic journals.
- Key applications of AI in healthcare discussed include using machine learning on large medical datasets to detect suspicious moles earlier, reduce medication errors, and more accurately predict cancer occurrence up to 12 months in advance.
- The author argues that AI has the potential to transform medicine by enabling more preventive and earlier detection approaches compared to traditional reactive healthcare models.
Jane may be able to help. Let me check with her personal assistant Jane-ML.
NextPrevIndex
Meera checks with Jane-ML
User-Agent Interaction (V)
48
PA_Meera: Mina, do you
have trouble in
debugging?
Mina: Yes, is there
anyone who has done
this?
Personal Agent
[Meera]
Jane-ML: Jane has done a similar debugging problem before. She is available now and willing to help.
compiletheme
Compiling output
1) Kaggle is the largest platform for AI and data science competitions, acquired by Google in 2017. It has been used by companies like Bosch, Mercedes, and Asus for challenges like improving production lines, accelerating testing processes, and component failure prediction.
2) The document discusses the author's experiences winning silver medals in Kaggle competitions involving camera model identification, passenger screening algorithms, and pneumonia detection. For camera model identification, the author used transfer learning with InceptionResNetV2 and high-pass filters to identify camera models from images.
3) For passenger screening, the author modified a 2D CNN to 3D and used 3D data augmentation to rank in the top 7% of the $1
[台灣人工智慧學校] Bridging AI to Precision Agriculture through IoT台湾资料科学年会
?
The document describes a system for precision agriculture using IoT. It involves sensors collecting environmental data from fields and feeding it to a control board connected to actuators like irrigation systems. The data is also sent to an IoTtalk engine and AgriTalk server in the cloud for analysis and remote access/control through an AgriGUI interface. Equations were developed to estimate nutrient levels like nitrogen from sensor readings to help optimize crop growth.
The document discusses Open Robot Club and includes several links to its website and YouTube videos. It provides information on the club's computing resources like NVIDIA V100 GPUs. Tables with metrics like underkill and overkill percentages are included for different types of tasks like AI AOI and PCB inspection. The club's website and demos are referenced throughout.
[2018 台灣人工智慧學校校友年會] Practical experience in mining and evaluating information...台湾资料科学年会
?
1) The document discusses four common pitfalls in training and evaluating recommender systems.
2) The first pitfall is that training models on clickstream data can result in models that learn the layout and structure of pages rather than true user interests.
3) The second pitfall is that using live recommendation data to evaluate new models favors algorithms similar to the online one.
4) The third pitfall is that click-through rates alone do not accurately capture business goals like revenue generation.
5) The fourth pitfall is that accurately measuring increased purchases from recommendations, rather than redirected purchases, is challenging.
[2018 台灣人工智慧學校校友年會] Textual Data Analytics in Finance / 王釧茹台湾资料科学年会
?
Dr. Chuan-Ju Wang gave a talk on textual data analytics in finance. He discussed how natural language processing and text analytics can be used to analyze unstructured text data, such as financial reports, to gain insights for financial applications like risk prediction. Specifically, he described how sentiment analysis of financial reports using finance-specific lexicons can predict stock return volatility and relative risk levels of companies. He also discussed using continuous word embeddings to automatically expand financial lexicons with related keywords.
1. The document discusses explainable medical AI and techniques like Grad-CAM and SHAP that can help explain AI models in medical contexts like diabetic retinopathy detection and gene expression prediction.
2. It describes using Grad-CAM to help explain a model for detecting diabetic retinopathy and using SHAP to explain a gene expression prediction model by determining the contribution of different histone modification signals.
3. The author advocates that explainable AI is very important for medical AI to help ensure models are not seen as "black boxes" and to build trust in the use of AI in healthcare.
Yen-Yu Lin presents research on video synthesis through frame interpolation. His lab uses deep learning models like DVF to predict intermediate frames between two consecutive frames. However, existing methods produce artifacts or over-smoothed results. The proposed approach uses a two-stage training procedure with cycle consistency loss to address this. It first pre-trains DVF, then fine-tunes with cycle loss to make the model robust to lack of data and produce higher quality frames. Experimental results show the approach outperforms state-of-the-art methods on standard datasets.
The document discusses various techniques for optimizing deep learning models from software to hardware to address increasing computational demands. It covers approaches like pruning, quantization, task partitioning, liveness analysis, layer scheduling, memory allocation, software decomposition, and microarchitecture optimizations to improve efficiency. Precision management and quantization are also examined as ways to balance performance and accuracy.
H. T. Kung from Harvard University congratulates the graduates of the Taiwan AI Academy program for completing their four months of study in core AI techniques like deep learning and real-world applications. Kung encourages the graduates to apply what they've learned to benefit industry, business, and civic life, and to continue learning as the field progresses rapidly. Kung thanks those involved in the Taiwan AI Academy program and looks forward to seeing the graduates' contributions to society.
Major roles in an AI team include product managers, data scientists, machine learning engineers, data engineers, and IT/data infrastructure. Successful AI projects require cross-functional collaboration between these roles. Many AI projects fail due to data silos that prevent effective collaboration and evaluation of ROI. Change is constant, and effective AI deployment requires adapting to new challenges through experimentation and iteration.
2. 陳昇瑋 / 人工智慧民主化在台灣 2
Mobile computing, inexpensive sensors collecting terabytes of data, and the
rise of machine learning that can use that data will fundamentally change the
way the global economy is organized.
- Fortune, “CEOs:The Revolution is Coming,” March 2016
50. 陳昇瑋 / 人工智慧民主化在台灣
State of AI In The Enterprise, 2018
50
Deloitte interviewed 1,100 IT and line-of-business executives
from US-based companies in the 3rd quarter of 2018.
82% of enterprise AI early adopters are seeing a positive ROI from
their production-level projects this year.
69% of enterprises are facing a “moderate, major or extreme”
skills gap in finding skilled associates to staff their new AI-driven
business models and projects.
63% of enterprises have adopted machine learning, making this
category the most popular of all AI technologies in 2018.