More Related Content
Viewers also liked (19)
Similar to Self Introduction for people interested in me. (20)
Recently uploaded (20)
Self Introduction for people interested in me.
- 2. ? Vector Supercomputers ? Pseudo-Vector Supercomputers ? Parallel Supercomputers S-820 S-3800 CP-PACS SR2201 SR8000
- 3. ? Tsukuba University and Hitachi cooperated on development of Pseudo- vector mechanism (original technology), and I lead its compiler team. ? Based on evaluation of a customer's benchmark programs, I proposed several CPU instructions and it lead to orders from important customers.
- 4. ? For parallelization of big program, Hitachi developed interprocedural parallelizer, and I developed core of it i.e. inter-procedural analysis. ? Wrote several US-patent on compiler technology.
- 5. ? Hitachi joined Japanese National Project RWC (Real World Computing) and developed WPP (Whole Program Parallelizer). Interprocedural analysis played core role in WPP. This research result was exhibited at SC99. http://keima.la.coocan.jp/rwcp/10years/10-2/docs/chapter007/3para/16/36.html ? The project was hand over to next National Project Advanced Paralleling Compiler Project http://www.apc.waseda.ac.jp/index.html ? I was a committee member of HPC research group of IPSJ in 2000 - 2001.
- 6. ? In 2001, Hitachi made decision to concentrate on Storage research and many researcher including me moved to storage research. ? I lead cross laboratory NAS (Network Attached Storage) project and achieved 3 times speedup of the NAS. ? In this project, I proposed hardware-counter based profiling of Kernel, and discovered that memory access govern the performance. I proposed to change the structure of hash table and it was very effective. ? I left Laboratory after this research in 2003.
- 7. ? 2003 C 2009: In SAN Solution Division, I cooperated with Hitachi Data Systems (headquarters in San Jose) and contributed to the shift from hardware to software/services in storage business. It includes planning and promotion of new storage, its software and services, and alliance with storage related vendors: Symantec/Veritas, OpenText, AppIQ (acquired by HP 2005/9), Archivas (acquired by Hitachi 2007/2), etc. As for Archivas, I conducted evaluation and commercialization of their software. ? 2009 C 2015: In Platform Solution Division and Cloud Service Division, I contributed the shift from product based business to service based business (cloud). I was engaged in planning and development of cloud service 3 years. After that, I was charged in public relations of Hitachi Cloud. It had been called Harmonious Cloud. I contributed news release of alliance with Microsoft, alliance with Amazon, Hitachi Cloud, etc. AppIQ
- 8. ? 2015/10 C 2016/2 : Retired Hitachi 2015/9E with early retirement program with additional severance pay. Enjoyed travel to Switzerland with my wife and free days. But, it had been becoming boring. ? 2016/2 - : In the middle of February, I started learning of latest web technology, Java, deep learning, etc. It has been very interesting and enjoyable for me. I started blog, twitter and github at first. ? I wrote roughly 100 articles, and got 22,258 access in total. http://itsukara.hateblo.jp/ ? I uploaded 18 repositories in github (5 repositories are related to AI). https://github.com/Itsukara ? I uploaded 3 slide in slideshow.net (all are related to deep learning). Currently, total view is 3,321. http://www.slideshare.net/ItsukaraIitsuka/ ? I started Learning of deep learning in the end of May. I attended Do2dle research group meeting several times and read 4 books on deep learning (before that, I read 1 book). (Do2dle: come from DUDL i.e. Deep Understanding of Deep Learning. After entered research group, I named it as this) ? In the beginning of August, DQN group of Do2dle set target of group as Montezumas Revenge, because it was the most difficult in DeepMind paper.
- 9. ? Only I had a time to try Montezumas Revenge. I got stuck neear the end of August. Because I realized that there is little chance to get reward in Montezumas Revenge. I felt that some kind of mile stone or reward decreasing when the same state continue is needed. I wrote it in blog. ? I received twitter message from Mr. Miyoshi, who is the author of very fast A3c code which I was using, that DeepMind submitted a paper related to Montezumas Revenge, although he had not been read it yet. ? After that, I concentrated on Montezumas Revenge. For details, I will introduce it in another slide. ? I also has a slide on my works in github, etc. ? My learning history: ? 2016/2-5: Mastered web-front-technology (HTML5/CSS3, Javascript) and some of backend- technology NodeJS, sqlite3 and Java (Oracle OJCP Silver certificate in one month). ? 2016/5-8: Learned basics of AI and deep learning. Tried DQN and A3C with Breakout. ? 2016/8- : Concentrated on Deep Reinforcement Learning of Montezumas Revenge. ? 2016/9/14: Made presentation in WBAI meeting held at Dwango AI Lab. lead by Mr. Yamakawa WBAI: Whole Brain Architecture Initiative, NGO to promote research and development of general AI like human, lead by Mr. Yamakawa http://wba-initiative.org/en/about/ http://wba-initiative.org/wp-content/uploads/2015/05/wbai-leaflet.pdf
- 11. Thank you for listening