Recently, more and more people are acquiring qualifications called “G test” and “E-qualification”. These are qualifications that certify knowledge of deep learning provided by the Japan Deep Learning Association (JDLA). Since the 2018 exam, as of March 2022, the total number of people who have passed the G test and E qualification has exceeded 50,000, attracting a great deal of attention.
Among them, the E qualification is a qualification for AI engineers, and a wide range of advanced knowledge is required, centering on problems related to deep learning, such as technology related to implementation and the topic of the latest papers. Therefore, advance preparation is essential to pass”(E-qualification).
Also, in order to take the E qualification, it is necessary to take and complete the JDLA certification program . With multiple certification programs to choose from, it’s very important.
This time, we will ask those who actually passed the E qualification about the measures and study tips to pass the E qualification, and introduce them as a testimonial. We will also explain the points and countermeasures of the new syllabus that will be changed from August 2022 .
This is a must-see especially for those who have started studying AI(E-qualification), those who aim to become AI engineers, and those who want to expand their field of activity as an AI engineer!
If you want to know the outline of the E qualification exam, the difficulty level, and the pass rate, please check the article below.
Contents
- What is E qualification?
- E-Qualification Successful Interview
- Syllabus revision points and countermeasures
- summary
1. What is E qualification?
The E qualification is a qualification that proves deep learning knowledge and implementation skills sponsored by the Japan Deep Learning Association (JDLA) .
*For an overview of the E qualification, please refer to “What is the E qualification – Japan Deep Learning Association [Official] “.
Since the E qualification was held in 2018, there have been a total of 4,838 successful applicants E-qualification).
The pass rate is relatively high at about 60% to 70% each time, but the E qualification asks not only whether you understand the mechanism of AI, but also whether you know the theoretical content and the flow of implementation. will be split. It will be difficult to pass the exam if you do not master these wide-ranging knowledge correctly.
2.Interviw with E qualified candidates
From here, we will introduce the experiences of those who took the E-qualification exam held in August 2021 and passed it.
■Profile of Successful
Candidates Mr. Iwahori: Worked in the manufacturing industry. She is in charge of middle management in a department that develops product technology.
Okuda: I work in the manufacturing industry. He is in charge of operations in the department responsible for product evaluation and development.
-Why did you get the E qualification?
Mr. Iwahori: It all started when I was put in charge of the digital promotion team in my current department. Many of the members who gathered were good at AI and programming, so I felt a sense of crisis, so I started by studying for the G test E-qualification). I had never touched digital or programming before, and I had no knowledge, so I started from a negative point.
Mr. Okuda: I thought it was cool to see Mr. Iwahori, who is my direct boss, trying to get the E qualification, so I decided to take on the challenge myself.
As a prerequisite knowledge, I have been studying programming itself for about three years, mainly in MATLAB (*), and I also dealt with automation at work.
* Numerical analysis software developed by MathWorks and the programming language used in it
-What was your impression when you started studying for the E qualification?
Mr. Iwahori: When I first saw the workbook, I had a very difficult impression. I was worried about the future because I had almost no knowledge of mathematics such as singular value decomposition and matrices. For the first 1-2 months, I was copying Python in a situation where I didn’t know anything.
Mr. Okuda: When I saw the reference book, it was a shockingly difficult impression . I thought I was good at mathematics because I dealt with mathematics at work, but I was blown away E-qualification). From that point on, I started studying with my hands while watching YouTube and reference books every day.
2-1. About the E qualification “JDLA Certification Program”
-Which JDLA certification program and materials did you use?
Mr. Iwahori: After considering various certification programs, I took the Skill Up AI certification program. The reason I chose it is because the lecturer ‘s explanation was very easy to understand when I watched the free trial version . It was easy to hear the voice, such as saying difficult or important points twice.
Mr. Okuda: I also took the skill-up AI certification program. The reason is that the lecturer ‘s explanation was very easy to understand. I became a big fan when I watched the trial version.
◆ JDLA certification program for E-qualification measures for skill improvement AI ◆
-Have you ever been glad that you chose the Skill Up AI program?
Mr. Iwahori: Learning items are written in the “Individual progress management file” distributed to students, and I was able to see my own learning schedule and achievements . It was easy to look back and easy to understand the points to remember (E-qualification) .
Mr. Okuda: The ease of understanding the materials . In addition to the explanations given in the lectures, I have never seen materials that are so easy to understand , such as illustrations, so I was surprised that I could treasure them for the rest of my life.
In particular, I was able to understand the explanation of difficult algorithms such as neural networks, which I had the impression of being difficult in other materials, by looking at the materials of Skill Up AI.
2-2. Study methods and measures for passing the E qualification
-How long did you prepare for the study period?
Mr. Iwahori: The study period was about eight months. I studied about 3 hours on weekdays and 5-6 hours on weekends.
Mr. Okuda: The study period is about 8 months like Mr. Iwahori. As for the amount of study, I was studying about 4 hours on weekdays and 7 hours on weekends.
-Could you give me some tips on how to study and the secrets of preparing for the E qualification?
Mr. Iwahori: I created a study cycle by
dividing it into three areas: (1) memorization, (2) mathematics, and (3) programming . For the part that I had to memorize in ①, I used a vocabulary notebook so that I could study even in my spare time, and I made a “list to remember on the day” and studied until just before the exam.
However, regarding mathematical formulas, there are many cases where questions are asked at different points from the part that was memorized in the actual exam . Especially in fields such as deep reinforcement learning, it is difficult to answer the questions just by memorizing them, so I think this is a point to be aware of.
Okuda:
First , I recommend that you master the basics of mathematics and Python syntax . I felt like I was able to do it just by solving examples from reference books such as black books, but in the end I had to redo the basics.
For that reason, I think it is important to challenge applied fields little by little after solidifying the foundation . As a result, I became very good at fill-in-the-blank problems.
I also struggled with formulas in the field of deep reinforcement learning, so I think it is important not only to see and memorize formulas, but also to unravel them and understand what they mean .
When I stumbled in my studies, I sometimes consulted in the skill-up AI Slack community.
Participants who have already acquired qualifications will answer questions and kindly teach you, so I think that using such a community will also be a source of emotional support.
-Do you have a reference book for the E qualification?
Mr. Iwahori: I mainly studied based on the problem collection called “Kuromoto”, ” Thorough Strategy Deep Learning E Qualification Engineer Problem Collection 2nd Edition “. I also learned a little bit in ” Deep Learning from Scratch “.
Source: Shinya Kogata, Shota Saito, Satoshi Mizoguchi, Kazuyuki Wakasugi “Thorough Strategy Deep Learning E Qualified Engineer Problem Collection 2nd Edition Thorough Strategy Series” (E-qualification)
Mr. Okuda: It was also recommended by an acquaintance who had obtained the E qualification, so I focused on reading ” Deep Learning from scratch ” for about a month before the exam .
Source: Yasuki Saito “Deep Learning from Scratch -Theory and Implementation of Deep Learning Using Python” (O’Reilly Japan 2016)
2-3. Impressions of taking the E qualification
– How was it when you actually received the E qualification?
Mr. Iwahori: I felt that the exam time was very short. You can also experience how to operate the test terminal in advance, but during the test, I didn’t know how to put out the calculator, so I was rushing to solve the problems while being pressed for time.
Okuda: I took the G test at home, so I was really nervous at the test center. I had about 40 minutes left when I finished solving the problem, so I was in a hurry. I didn’t expect to have extra time, but I think I can proceed with peace of mind if I solve any problem calmly without panicking .
2-4. Glad I passed the E qualification
-Did the reaction of those around you change after you passed the exam?
Mr. Iwahori: Many of the people I work with usually use AI, so having an E qualification has allowed me to gain a certain level of trust from those around me.
Okuda: Actually, until now, I was said to be the number one “analog person” in the team I’m currently on. I didn’t know anything about AI, but by passing the E qualification, I was able to use my knowledge to develop products and systems .
People around me start to surprise me, saying, “Can you do something like this?”
-Are there any changes in your personal feelings?
Okuda: The change in my personal state of mind is huge. In my daily work, I often receive consulting and advice from customers, but the range of things I can provide has expanded, and I can now give advice when creating a system .
Until now, I often wondered which system or algorithm to offer, so I feel that the knowledge I gained from my E qualification is useful in my work.
2-5. Utilize the knowledge and skills acquired in the E qualification in the field
-Do you have any challenges to promote AI in your company?
Mr. Iwahori: In our daily work, we receive information about defects from the market on a daily basis. At that time, I would like to be able to use natural language processing to determine what is the cause of this problem, and I am working on it.
Now, veterans are reading the customer’s words and investigating the cause. If we can use AI technology to classify people based on their words , I think it will lead to work reduction and efficiency improvement in the future .
-Do you have an image of what you want to challenge in the future or what you aim for?
Mr. Iwahori: First of all, I would like to take on the responsibility of developing AI that is useful in my work .
Specifically, I would like to analyze customer feedback using natural language processing and develop AI that can embody human sensibilities. I also want to convey the necessity and enjoyment of AI to as many people as possible .
Mr. Okuda: I would like to disseminate new information to people around me under the theme of “enjoying AI and data science . ” Rather than making things, I think that development will naturally follow through understanding activities including DX.
* Details of the experience are available on YouTube “ AI Business Channel ”.
3. Key points for revision of the syllabus and countermeasures
It has been announced that the syllabus will be revised from the August 2022 exam. Many people who are planning to take the exam are wondering how to deal with the changes. From here on, for those who are planning to take the E qualification in 2022 #2 or later, Yamada
, a skill-up AI lecturer who is familiar with the revised syllabus, will teach you about the contents and points of the new syllabus, and how to take measures .
3-1.So that the implementation using the framework is questioned
Implementations with deep learning frameworks are added to the exam scope. Candidates must choose PyTorch or TensorFlow before starting the exam.
In the previous E qualification, he was asked about scratch implementation using NumPy, but some of the problems are replaced with implementation using frameworks.
Skill Up AI offers an introductory course to PyTorch for deep learning . In addition, in the deep learning basic course that can be used in the E qualification course , the PyTorch course is provided as a set along with this curriculum revision, so students can take measures for implementation using PyTorch .
Other recommended books
| field | book name |
|---|---|
| TensorFlow | Takuya Niimura, “Introduction to DeepLearning implementation starting with TensorFlow” Impress, 2018. |
| PyTorch | Keiichiro Miyamoto et al., PyTorch Neural Network Implementation Handbook, Shuwa System, 2018. |
3-2. Items added to each field of deep learning
New items have also been added to the deep learning section from the old syllabus. Additional points include:
- A3C added to deep reinforcement learning field
- ResNet and EfficientNet newly added to the field of image recognition
- BERT and GPT-n added to natural language processing field
All of these can be addressed with supplementary teaching materials that are compatible with the new syllabus of the deep learning basic course that can be used in the field .
Other recommended books and articles
| field | Division | Book/article name |
|---|---|---|
| deep reinforcement learning | books | Yutaro Ogawa “Learning while making! Deep Reinforcement Learning Practical Programming with PyTorch, Mynavi Publishing, 2018 |
| Qiita | [Reinforcement learning] Learning A3C while implementing [Stick stand with CartPole: Complete with 1 file] →Explanation about A3C |
|
| image recognition | blog post | Pytorch – Explain the mechanism and implementation of ResNet → Explain the mechanism of ResNet with an implementation example using PyTorch |
| natural language processing | media articles | Mechanism and usage of language understanding AI – How to use “BERT” and “GPT-3”? → Explain the basic mechanism of BERT and GPT-3 |
| Qiita | I tried to implement sentence classification & attention visualization by Japanese BERT with PyTorch → Explain how to implement BERT using PyTorch |
If you want to fully understand the mechanism, it is a little difficult, but I recommend reading the original paper or its Japanese commentary article.
3-3. Added new sections such as distance learning, meta-learning, and deep learning explainability
Distance learning, meta-learning, deep learning explainability, etc. have also been added. In this field, too, it is certain to learn from the original papers and their derivative contents, but it may be difficult to understand without sufficient background knowledge.
Knowledge in these relatively new areas is difficult to learn on your own, as there is little information available in books and the like.
For newly added fields, it is possible to take countermeasures with supplementary teaching materials for deep learning basic courses that can be used in the field .
In addition, the free study session ” Skill Up AI Camp ” will deal with the new fields of XAI (explainable AI) and meta-learning as themes, so you can learn more effectively by participating!
Other recommended books and articles
| field | Division | Book/article name |
|---|---|---|
| distance learning | Qiita | Standard of Deep Metric Learning⁈ Thorough explanation of Triplet Loss → Explanation of basic knowledge of distance learning |
| meta-learning | blog post | I experienced implementing MAML (Model-Agnostic Meta-Learning) with Pytorch. → Explain the basics of meta-learning and how to implement it with PyTorch |
4. Summary
AI is now so important that it can be said to be the leading role of DX. In order to become a person responsible for DX, it is necessary to understand the theory of deep learning and have the skills to implement it(E-qualification).
If you are aiming to pass the E qualification to acquire such skills, please refer to this blog and take measures.
Currently, you can watch a part of the JDLA certification program (equivalent to about 6 hours) for E qualification provided by Skill Up AI for free as a trial version(E-qualification).
You can experience the atmosphere and clarity of the course, so please try it.
In addition, at the “‘JDLA Certification Program’ Online Information Session for E-Qualifications”, we will explain the features, outline, and schedule of the certification program. If you are interested in the details , please join us for a free briefing session .
