News : YOKOTA Laboratory : Science Tokyo, GSIC

2023.12.19

Okazaki Lab, Yokota Lab, and AIST have released a large language model “Swallow” with the highest Japanese capability among open models.
Okazaki and Yokota labs at the Tokyo Institute of Technology, and the National Institute of Advanced Industrial Science and Technology has improved the Japanese language capabilities of Llama 2 7B, 13B, and 70B, achieving the highest performance in Japanese among the 13B and 70B open LLMs.
https://www.titech.ac.jp/news/2023/068089

2023.08.11

Prof. Yokota appeared in the TV series "Gaia no Yoake"
Prof. Yokota introduced his recent efforts to develop foundational technology to train generative AI models on the supercomputer Fugaku in the TV series "Gaia no Yoake”
https://www.tv-tokyo.co.jp/gaia/

2023.04.25

We will be hosting a special seminar by members of Stability AI Japan - David Ha, Jerry Chi, and Kamil Rocki. We plan to stream the talks on Zoom.

Date/time：	April 25 (Tue) 11:00-12:00
Location：	Tokyo Institute of Technology, Ookayama campus, GSIC building 2F meeting room
Zoom：	https://zoom.us/j/96903641065?pwd=ZEdrZ3ZkbUFUMDNYL3FrOGtnSG5aUT09 No registration required, presentations will be in English.
Presenters：	David Ha, Jerry Chi, Kamil Rocki
Title：	Collective Intelligence of Generative AI
Abstract：	Generative machine learning models, such as generative adversarial networks for images, and language models for text have been around for many years. While impressive, their use cases had been largely confined to research or technical communities. However, from 2022-2023, we are witnessing an explosion of interest in text-to-image models, which are generative ML models that combine both text and image modalities. The ability to generate images simply from text input created interest far beyond the technical community, resulting in an explosion of casual creativity. What is it about the multi-modality of the models that make them so interesting to people and how will this technology fundamentally change our relationship with technology, and with each other? In this talk, we first give an overview of the state text-to-image technology, and explore deep questions concerning what makes this technology so compelling to large groups of users and continually drives iterative exploration. We explain why text-to-image models are a fundamental technology that enables our desire as social animals to share novel cultural artifacts and innovative uses of the technology itself with others, and, like previous culture-changing technology such as photography and mobile phones, will drive large shifts in our collective behavior that will fundamentally alter human culture.
Bio：	- David Ha is the Head of Strategy at Stability AI. He previously worked as a Research Scientist at Google, working in the Brain research team. His research interests include complex systems, self-organization, and creative applications of machine learning. Prior to joining Google, He worked at Goldman Sachs as a Managing Director, where he co-ran the fixed-income trading business in Japan. He obtained his undergraduate and masters degrees from the University of Toronto, and his PhD from the University of Tokyo. - Jerry is a quadrilingual Taiwanese American based in Tokyo, currently working on building the Japan team at Stability AI. Previously, he had a wide array of analytics and machine learning roles at Google, Supercell, SmartNews, and Indeed. Also, he is an ex-entrepreneur with experience advising startups. He is especially passionate about generative AI and creative applications of machine learning. Jerry holds degrees from Stanford's School of Engineering and the University of Pennsylvania's Wharton School. - Kamil Rocki graduated from the University of Tokyo with a PhD degree in Computer Science in 2011. Currently, he is a Senior Machine Learning Research Engineer at Stability AI, where he pushes the GPUs to their limits. Prior to that the led the dense linear solver team at NVIDIA. Before that he worked at Neuralink, Cerebras Systems and IBM Research. During that time he worked he decoded brain signals using a AI on an 64-MHz ARM chip in real-time, wrote implemented a neural network as well as faster-than-realtime Fluid Dynamics simulation on a 400.000-core wafer in assembly as well as prototyped a distributed Reinforcement Learning on an 1000-node FPGA cluster.

2023.03.04

The work by Satoki Ishikawa (4th year Bachelor) won the Student Encouragement Award at the 85th National Convention of IPSJ
The work by Satoki Ishikawa (4th year Bachelor) titled “Examination of Gradient Preconditioning in Deep Learning” was presented at the 85th National Convention of IPSJ and won the Student Encouragement Award.

2023.03.02

The work by Hiroyuki Ootomo (3nd year PhD) won the best paper award at HPC Asia 2023
The work by Hiroyuki Ootomo (3nd year PhD) “Reducing shared memory footprint to leverage high throughput on Tensor Cores and its flexible API extension library” won the best paper award at HPC Asia 2023.

2023.02.28

The work by Sora Takashima (2nd year Master) in collaboration with AIST was accepted to CVPR2023
The work by Sora Takashima (2nd year Master) in collaboration with AIST “Visual Atoms: Pre-training Vision Transformers with Sinusoidal Waves” has been accepted to IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR2023) (acceptance rate 25.78%)

2023.02.01

Kazuki Osawa (graduated) has joined DeepMind
Kazuki Osawa (graduated) has joined the Optimization and Bayesian Deep Learning Team at DeepMind.

2022

2022.12.18	The work by Aoyu Li (alumni) received the Best paper award at ACM Multimedia Asia 2022
2022.12.18	The work by Aoyu Li (alumni) titled “Informative Sample-aware Proxy for Deep Metric Learning” received the Best paper award at ACM Multimedia Asia 2022Research Award
2022.07.28	The work by Hiroyuki Otomo (3^nd year PhD) received the Yamashita Memorial Research Award The work by Hiroyuki Otomo (3^nd year PhD) titled “Error Corrected Single Precision Matrix Multiplication on TensorCores” received the Yamashita Memorial Research Award https://www.ipsj.or.jp/award/yamashita2022.html
2022.06.13	The work by Qianxiang Ma (2^nd year PhD) and Sameer Deshmukh (3^rd year PhD) was accepted to SC22. The work by Qianxiang Ma (2^nd year PhD) and Sameer Deshmukh (3^rd year PhD) titled ”Scalable Linear Time Dense Direct Solver for 3-D Problems Without Trailing Sub-Matrix Dependencies” has been accepted to The International Conference for High Performance Computing, Networking, Storage, and Analysis (SC22)
2022.05.29	The work by Muhammad Ridwan Apriansyah (2^nd year PhD) was accepted to ACM TOMS. The work by Muhammad Ridwan Apriansyah (2^nd year PhD) titled ”Parallel QR Factorization of Block Low-Rank Matrices” has been accepted to ACM Transactions on Mathematical Software
2022.03.04	The work by Shukai Nakamura (4^th year Bachelor) won the Student Encouragement Award at the 84^th National Convention of IPSJ The work by Shukai Nakamura (4^th year Bachelor) titled “The effect of batch size on the generalization of vision transformers” was presented at the 84^th National Convention of IPSJ and won the Student Encouragement Award.
2022.03.03	The work by Edgar Martinez Noriega (Postdoc), Sora Takashima (1^st year Master), Xinyu Zhang (1^st year Master) in collaboration with AIST was accepted to CVPR2022. The work by Edgar Martinez Noriega (Postdoc), Sora Takashima (1^st year Master), Xinyu Zhang (1^st year Master) in collaboration with AIST titled ”Replacing Labeled Real-image Datasets with Auto-generated Contours” has been accepted to IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR2022). (Acceptance rate 25.33%)
2022.02.28	The work by Hiroyuki Ootomo (2^nd year PhD) was accepted to IJHPCA. The work by Hiroyuki Ootomo (2^nd year PhD) titled ”Recovering Single Precision Accuracy from Tensor Cores While Surpassing the FP32 Theoretical Peak Performance” has been accepted to The International Journal of High Performance Computing Application
2022.02.01	The work by Hana Hoshino (2^nd year Master) was accepted to ICRA2022. The work by Hana Hoshino (2^nd year Master) titled ”OPIRL: Sample Efficient Off-Policy Inverse Reinforcement Learning via Distribution Matching” has been accepted to IEEE International Conference on Robotics and Automation (ICRA2022). (Acceptance rate 43%)

2021

2021.09.17	CREST Project with Emtiyaz Khan at RIKEN AIP was awarded A joint proposal with Emtiyaz Khan at RIKEN AIP titled "A new Bayes-Duality principle for adaptive, robust, and life-long learning of AI” for the CREST area “Creating information utilization platform by integrating mathematical and information sciences, and development to society” has been awarded.
2021.07.23	The work by Shun Iwase (graduated) was accepted to ICCV2021. The work by Shun Iwase (graduated) titled ”RePOSE: Real-Time Iterative Rendering and Refinement for 6D Object Pose Estimation” has been accepted to The International Conference on Computer Vision (ICCV2021). (Acceptance rate 25.9%)
2021.04.13	The work by Hikaru Nakata (graduated) was accepted to CVPR2021 Workshop CLVISION The work by Hikaru Nakata (graduated) titled ”Self-supervised Continual Pretraining for Class Incremental Image Classification” has been accepted to Workshop on Continual Learning in Computer Vision at CVPR 2021.

2020

2020.11.06	A new post-doc Edgar Josafat Martinez Noriega has joined our group A new post-doc Edgar Josafat Martinez Noriega has joined our group as a Researcher for the NEDO project “Technology development business related to next-generation artificial intelligence that evolves with people”.
2020.10.02	The team “shallowverse” in which Yuichiro Ueno (M2) participates has passed the qualifying for the Web application acceleration contest “ISUCON10" in the second place overall and the first place for student teams.
2020.09.25	The work by Kazuki Osawa (3^rd year PhD) was accepted to NeurIPS 2020 as an oral presentation The joint work of Kazuki Osawa (3^rd year PhD) and Ryo Karakida at AIST titled "Understanding Approximate Fisher Information for Fast Convergence of Natural Gradient Descent in Wide Neural Networks"was accepted to the Thirty-forth Conference on Neural Information Processing Systems (NeurIPS 2020) as an oral presentation.（Acceptance rate 1.1%）
2020.08.25	The joint work by Yuichiro Ueno (2^nd year Master) and Kazuki Osawa (3^rd year PhD) was presented at KDD 2020 The joint work by Yuichiro Ueno (2^nd year Master) and Kazuki Osawa (3^rd year PhD) titled "Rich Information is Affordable: A Systematic Performance Analysis of Second-order Optimization Using K-FAC" was presented at KDD 2020. (Acceptance rate 16.9% 216/1279)
2020.08.05	Our HPCI project received the outstanding research award. Our HPCI project titled “Development of an Innovative Optimization Method for Massively Parallel Deep Learning” received the HPCI outstanding research award at the 7^th results briefing.
2020.07.03	The joint work of Mikiya Shibuya (1^st year Master) and Ken Sakurada at AIST was accepted to ECCV2020 The joint work of Mikiya Shibuya (1^st year Master) and Ken Sakurada at AIST titled "Privacy Preserving Visual SLAM" was accepted to ECCV2020 (Acceptance rate 27%)
2020.06.22	Hiroyuki Otomo (1^st year PhD) and Sameer Deshmukh (1^st year PhD) presented their posters at ISC2020
2020.06.20	The joint work by Kazuki Osawa (3^rd year PhD) and Yuichiro Ueno (2^nd year Master) was accepted to the IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). The joint work by Kazuki Osawa (3^rd year PhD) and Yuichiro Ueno (2^nd year Master) titled "Scalable and Practical Natural Gradient for Large-Scale Deep Learning" was accepted to one of the top journals in the AI/ML field -- the IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI).
2020.05.16	The joint work by Yuichiro Ueno (2^nd year Master) and Kazuki Osawa (3^rd year PhD) was accepted to KDD 2020 The joint work by Yuichiro Ueno (2^nd year Master) and Kazuki Osawa (3^rd year PhD) titled "Rich Information is Affordable: A Systematic Performance Analysis of Second-order Optimization Using K-FAC" was accepted to the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2020, Virtual Event, USA) research track. (Acceptance rate 16.9% 216/1279)
2020.04.23	Hiroyuki Otomo (1^st year PhD) and Sameer Deshmukh (1^st year PhD) had their posters accepted to ISC2020 The work by Hiroyuki Otomo (1^st year PhD) titled “Randomized SVD on TensorCores” and the work by Sameer Deshmukh (1^st year PhD) titled “Distributed Memory Task-Based Block Low Rank Direct Solver” was accepted to ISC2020 as a research poster.
2020.03.17	Linsho Kaku (2^nd year Master) won 1^st place in Kaggle Linsho Kaku (2^nd year Master) won 1^st place in the Kaggle competition for “Bengali.AI Handwritten Grapheme Classification”. https://www.kaggle.com/c/bengaliai-cv19/leaderboard
2020.01.22	Hiroyuki Otomo (2^nd year Master) won 2^nd place in the HAKKO experiment contest The work by Hiroyuki Otomo (2^nd year Master) for “An Experiment of Heat Released from Computer Parts” won 2^nd place in the HAKKO experiment contest.
2020.01.16	Sameer Deshmukh (1^st year PhD) and Muhammad Ridwan Apriansyah (2^nd year Master) presented their posters at HPC Asia 2020 The work by Sameer Deshmukh (1^st year PhD) titled “Distributed Memory Task-Based Block Low Rank Direct Solver” and the work by Muhammad Ridwan Apriansyah (2^nd year Master) titled “QR Decomposition of Block Low-Rank Matrices” was presented at the Poster Session at HPC Asia 2020. http://sighpc.ipsj.or.jp/HPCAsia2020/index.html

2019

2019.11.22	The work of Hiroki Naganuma (1^st year PhD) was presented as a poster at IBIS2019 http://ibisml.org/ibis2019/posters/
2019.11.20	The work by Hiroyuki Otomo (2^nd year Master) was selected as a best poster candidate at SC19 The work by Hiroyuki Otomo (2^nd year Master) titled “TSQR on TensorCores” was selected as a best poster candidate at SC19.
2019.11.19	The posters by Hiroyuki Otomo (2^nd year Master) and Qianxiang Ma (1^st year Master) were presented at SC19 The posters by Hiroyuki Otomo (2^nd year Master) titled “TSQR on TensorCores” and Qianxiang Ma (1^st year Master) titled “Runtime System for GPU-based Hierarchical LU factorization” were presented at SC19. https://sc19.supercomputing.org/
2019.11.17	The poster by Hiroki Naganuma (1^st year PhD) was presented at ACML2019 The poster by Hiroki Naganuma (1^st year PhD) titled “On Empirical Analysis of Layer-wised Learning Rate Schedule” was presented at ACML2019 Workshop on Statistics & Machine Learning Researchers.
2019.11.16-18	Linsho Kaku (1^st year Master) participated in the ICPC 2019 Asia Yokohama Regional as “unlimited greedy” https://icpc.iisf.or.jp/2019-yokohama/regional-standings/
2019.09.07	The work by Hiroyuki Otomo (2^nd year Master) and Qianxiang Ma (1^st year Master) were accepted as Research Posters to SC19 The posters by Hiroyuki Otomo (2^nd year Master) titled “TSQR on TensorCores” and Qianxiang Ma (1^st year Master) titled “Runtime System for GPU-based Hierarchical LU factorization” were accepted to SC19. https://sc19.supercomputing.org/
2019.09.03	The work by Hiroyuki Otomo (2^nd year Master) was presented at the JSIAM Annual Conference. The work by Hiroyuki Otomo (2^nd year Master) titled “TSQR on TensorCores” was presented at the JSIAM Annual Conference. https://annual2019.jsiam.org/
2019.09.03	The work by Kazuki Osawa (2^nd year PhD) was accepted to NeurIPS 2019. The work by Kazuki Osawa (2^nd year PhD) titled “Practical Deep Learning with Bayesian Principles” was accepted to a top conference in the field of machine learning -- the Thirty-third Conference on Neural Information Processing Systems (NeurIPS 2019, Vancouver, CANADA).（Acceptance rate 21.1% 1428/6743） https://neurips.cc/Conferences/2019/
2019.08.26-28	The CREST Deep Joint Workshop was held at LecTore Hayama.
2019.07.29	The work by Shun Iwase (2^nd year Master) was presented at MIRU2019. http://cvim.ipsj.or.jp/MIRU2019/
2019.07.26	The work by Hiroyuki Otomo (2^nd year Master) and Peter Spalthoff (2^nd year Master) was presented at the 170^th HPC Seminar https://www.ipsj.or.jp/kenkyukai/event/hpc170.html
2019.07.12	The team of Linsho Kaku (1^st year Master) “unlimited greedy” won the domestic qualifying round of ICPC2019 The team of Linsho Kaku (1^st year Master) “unlimited greedy” won the domestic qualifying round of ICPC2019, allowing them to proceed to the Asian qualifying round in December. https://icpcsec.firebaseapp.com/standings/
2019.05.29	The work by Hiroki Naganuma (1^st year PhD) received the Outstanding Presentation Award at xSig2019. The work by Hiroki Naganuma (1^st year PhD) received the Outstanding Presentation Award at xSig2019, held at Keio University. https://xsig.ipsj.or.jp/2019/awards/
2019.05.28	The work by Hiroki Naganuma (1^st year PhD) was presented at xSig2019. The work by Hiroki Naganuma (1^st year PhD) titled “Effectiveness of Smoothing in Natural Gradient Learning Method for Large Batch Learning” was presented at xSig2019. https://xsig.ipsj.or.jp/2019/
2019.05.14	The work by Hiroki Naganuma (1^st year PhD) was presented at CCGrid2019 The work by Hiroki Naganuma (1^st year PhD) titled “A Performance Improvement Approach for Second-Order Optimization in Large Mini-batch Training” was presented at the 2nd High Performance Machine Learning Workshop Held in conjunction with CCGrid2019. https://hpml2019.github.io/
2019.03.20	The work by The work by Hiroki Naganuma (2^nd year Master) won the Excellent Poster Award at the IEICE General Conference The work by The work by Hiroki Naganuma (2^nd year Master) was presented at the IEICE General Conference and won the Excellent Poster Award https://www.ieice-taikai.jp/2019general/jpn/
2019.03.15	The work by Hiroyuki Otomo (1^st year Master) won the Student Encouragement Award at the 81^st National Convention of IPSJ The work by Hiroyuki Otomo (1^st year Master) was presented at the 81^st National Convention of IPSJ and won the Student Encouragement Award. http://www.ipsj.or.jp/award/taikaigakusei.html
2019.03.14	Kazuki Osawa (1^st year PhD), Hiroki Naganuma (2^nd year Master), Hiroyuki Otomo (1^st year Master), and Hikaru Nataka (4^th year Bachelor) presented at the 81^st National Convention of IPSJ. Kazuki Osawa (1^st year PhD), Hiroki Naganuma (2^nd year Master), Hiroyuki Otomo (1^st year Master), and Hikaru Nataka (4^th year Bachelor) presented at the 81^st National Convention of IPSJ. https://www.ipsj.or.jp/event/taikai/81/
2019.03.02	The joint work by Kazuki Osawa (1^st year PhD) and Yuichiro Ueno (4^th year Bachelor) was accepted to CVPR2019 The joint work by Kazuki Osawa (1^st year PhD) and Yuichiro Ueno (4^th year Bachelor) titled “Large-scale Distributed Second-order Optimization Using Kronecker-factored Approximate Curvature for Deep Convolutional Neural Networks” was accepted to the IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2019, Long Beach, CA) (Acceptance rate 25.2% 1300/5160）
2019.02.16	The work by Yuichiro Ueno (4^th year Bachelor) was accepted to CCGrid 2019. The work by Yuichiro Ueno (4^th year Bachelor) titled “Exhaustive Study of Hierarchical AllReduce Patterns for Large Messages Between GPUs” was accepted to the IEEE/ACM International Symposium in Cluster, Cloud, and Grid Computing (CCGrid 2019)（Acceptance rate 22.7% 47/207）

2018

2018.12.08	The team of Peter Spalthoff (1^st year Master) won 1^st place at the Tokyo Tech division of the HULT PRIZE 2019 The team of Peter Spalthoff (1^st year Master) won 1^st place at the Tokyo Tech division of the HULT PRIZE 2019, allowing them to proceed to the regional qualifiers to be held in April 2019 at Melbourne, Australia. https://www.titech.ac.jp/news/2019/043498.html
2018.12.07	Hiroki Naganuma (2^nd year Master) won the JAMES DYSON AWARD2018. Hiroki Naganuma (2^nd year Master) won the JAMES DYSON AWARD2018 https://www.titech.ac.jp/english/news/2018/042324.html
2018.09.19	The work by Hiroki Naganuma (2^nd year Master) was presented at the 17^th Forum on Information Technology The work by Hiroki Naganuma (2^nd year Master) titled “Hyperparameter Optimization in Massively Parallel Deep Learning Using Natural Gradient Approximation” was presented at the 17^th Forum on Information Technology. https://www.ipsj.or.jp/event/fit/fit2018/
2018.08.20	Hiroyuki Otomo (1^st year Master) participated in the 24^th Supercomputing Contest as a tutor Hiroyuki Otomo (1^st year Master) participated as a tutor in the 24^th Supercomputing Contest, which was jointly held by the Global Scientific Information and Computing Center at the Tokyo Institute of Technology and the Cybermedia Center at Osaka University. http://www.gsic.titech.ac.jp/supercon/main/attwiki/index.php?SupercomputingContest2018
2018.03.14	Hiroki Naganuma (1^st year Master) participated in the KOBE HPC Spring School 2018 http://www.eccse.kobe-u.ac.jp/simulation_school/kobe-hpc-spring-school-2018/
2018.03.13	The work by Hiroyuki Otomo (4^th year Bachelor) and Yuji Kuwamura (4th year Bachelor) was presented at the 80^th National Convention of IPSJ The work by Hiroyuki Otomo (4^th year Bachelor) and Yuji Kuwamura (4^th year Bachelor) was presented at the 80^th National Convention of IPSJ. https://www.ipsj.or.jp/event/taikai/80/
2018.03.01	The work by Hiroyuki Otomo (4^th year Bachelor) was presented at the 163^rd HPC Seminar The work by Hiroyuki Otomo (4^th year Bachelor) titled “Distributed Learning of Deep Neural Networks Using Kronecker Factorization of Fisher Information Matrix” was presented at the 163^rd HPC Seminar. https://www.ipsj.or.jp/kenkyukai/event/hpc163.html
2018.01.29	The work by Hiroki Naganuma (1^st year Master) was presented at HPC Asia 2018. The work by Hiroki Naganuma (1^st year Master) titled “Accelerating Convolutional Neural Networks Using Low Precision Arithmetic” was presented at HPC Asia 2018. http://sighpc.ipsj.or.jp/HPCAsia2018/

2017

2017.12.12	The work by Hiroki Naganuma (1^st year Master) was present at GTC Japan 2018 The work by Hiroki Naganuma (1^st year Master) titled “Verification of speedup using low-precision operations in convolutional neural networks” was present at GTC Japan 2018, held at Hilton Odaiba, Tokyo. https://www.gputechconf.jp/sessions.html
2017.11.23	Hiroki Naganuma (1^st year Master) won the Digital Creative Department Grand Prize at the 2025 Japan World Expo Committee Creative Competition. Hiroki Naganuma (1^st year Master) won the Digital Creative Department Grand Prize at the 2025 Japan World Expo Committee Creative Competition. https://pulse-beat.com/articles/a7vAC
2017.11.12	Kazuki Osawa (2^nd year Master) was selected as a student volunteer for SC17 http://sc17.supercomputing.org/studentssc/student-volunteers/
2017.10.21	Hiroki Naganuma (1^st year Master) won 3^rd place in the Stanford's Health Hackathon Health++ 2017 Hiroki Naganuma (1^st year Master) participated in the Stanford's Health Hackathon Health++ 2017 and won 3^rd place and received the Persistent-Neodesign award. https://www.titech.ac.jp/news/2017/039929.html
2017.10.12	Kazuki Osawa (2^nd year Master) and Hiroki Naganuma (1^st year Master) presented their work at PRMU 2017 http://www.ieice.org/iss/prmu/jpn/index.html
2017.09.13	The work by Kazuki Osawa (2^nd year Master) was presented at ICANN 2017 The work by Kazuki Osawa (2^nd year Master) titled “Evaluating the Compression Efficiency of the Filters in Convolutional Neural Networks” was presented at the 26^thInternational Conference on Artificial Neural Networks. http://www.icann2017.org/
2017.09.06	The work by Hiroki Naganuma (1^st year Master) was presented at the JSIAM Annual Conference The work by Hiroki Naganuma (1^st year Master) titled “Speeding Up Compressed Models Using Half-Precision Operations in Deep Learning” was presented at the JSIAM Annual Conference. http://annual2017.jsiam.org/
2017.09.04	Hiroyuki Otomo (4^th year Bachelor) participated in the KOBE HPC Summer School 2017 Hiroyuki Otomo (4^th year Bachelor) participated in the KOBE HPC Summer School 2017. http://www.eccse.kobe-u.ac.jp/simulation_school/kobe-hpc-summer-school-2017/
2017.08.25	Hiroyuki Otomo (4^th year Bachelor) participated in the 23^rd Supercomputing Contest as a tutor Hiroyuki Otomo (4^th year Bachelor) participated in the 23^rd Supercomputing Contest, which was jointly held by the Global Scientific Information and Computing Center at the Tokyo Institute of Technology and the Cybermedia Center at Osaka University. http://www.gsic.titech.ac.jp/supercon/main/attwiki/index.php?SupercomputingContest2017
2017.07.26	Naganuma Hiroki (1^st year Master) received the Son Masayoshi Foundation Scholarship Naganuma Hiroki (1^st year Master) received the Son Masayoshi Foundation Scholarship, selected as one of the most talented 96 young scholars under 25 years old among the more than 1,100 applicants. http://masason-foundation.org/
2017.07.17	The work by Kazuki Osawa (2^nd year Master) was presented at HPCS2017. The work by Kazuki Osawa (2^nd year Master) titled “Accelerating Matrix Multiplication in Deep Learning by Using Low-Rank Approximation” was presented at the 2017 International Conference on High Performance Computing & Simulation. http://hpcs2017.cisedu.info/
2017.06.01	The work by Kazuki Osawa (2^nd year Master) was presented at the 22^nd JSCES Conference. The work by Kazuki Osawa (2^nd year Master) titled “Accelerating Convolutional Neural Networks Using Low-Rank Approximation” was presented at the 22^nd JSCES Conference. http://www.jsces.org/koenkai/22/
2017.03.27	The team of Hiroki Naganuma (4^th year Bachelor) won 1^st Prize at the Imagine Cup 2017 Japan qualifiers The team of Hiroki Naganuma (4^th year Bachelor) won 1^st Prize at the Imagine Cup 2017 Japan qualifiers. They also received the Dean's Award, School of Computing at Tokyo Institute of Technology for this achievement. https://www.microsoft.com/ja-jp/education/imagine-cup.aspx
2017.03.08	The work by Akira Sekiya (1^st year Master) was presented at the 158^th HPC Seminar The work by Akira Sekiya (1^st year Master) titled “Accelerating Matrix Products in Deep Learning Using Low-Rank Approximation” was presented at the 158^th HPC Seminar. http://www.ipsj.or.jp/kenkyukai/event/hpc158.html
2017.02.21	The Tokyo Tech. – AIST Real World Big Data Computation Open Innovation Laboratory (RWBC-OIL) was founded The Tokyo Tech. – AIST Real World Big Data Computation Open Innovation Laboratory (RWBC-OIL) was founded. Our lab is involved in the Real World HPC Systems for Big Data/AI Research Team. http://www.titech.ac.jp/news/2017/037519.html http://www.aist.go.jp/aist_j/news/announce/pr20170220.html

2016

2016.12.19	Hiroki Naganuma (4^th year Bachelor) won the JPHacks Best Idea Award, Mashup Award, and best student award At the awards ceremony held at the University of Tokyo on November 19, 2016, members of TITAMAS received the Best Idea Award for developing "Walky," a handheld device to assist the visually impaired when walking. The team also took home the AbemaTV Award, Softbank Award, Mitsubishi UFJ Morgan Stanley Securities Award, Microsoft Imagine Cup Award, and Mashup Awards Prize. https://www.titech.ac.jp/english/news/2017/036942
2016.11.23	Dr. Miquel Pericas visited our group. Dr. Miquel Pericas from Chalmers University of Technology visited our group for collaboration on Task Assembly Objects (TAO) for Fast Multipole Methods (FMM).
2016.11.13	Hiroki Naganuma (4^th year Bachelor) was selected for the SC16 HPC for Undergraduates Program Hiroki Naganuma (4^th year Bachelor) was the first Japanese student to be selected for the HPC for Undergraduates Program at SC. http://sc16.supercomputing.org/
2016.08.28	Kazuki Osawa (1^st year Master) participated in the Silicon Valley Workshop 2016 Summer Kazuki Osawa (1^st year Master) participated in the Silicon Valley Workshop 2016 Summer organized by Recruit Holdings’ TECH LAB PAAK, where he spent 10 days touring the IT companies in Silicon Valley. Places he visited are Stanford University, UC Berkeley, NVIDIA, Google, Twitter, Facebook, Recruit Institute of Technology, Stripe, Niantic, Torre, Oracle. http://techlabpaak.com/blog/1548
2016.08.22	Kazuki Osawa (1^st year Master) and Hiroki Naganuma (4^th year Bachelor) participated in the 22^nd Supercomputing Contest as tutors Kazuki Osawa (1^st year Master) and Hiroki Naganuma (4^th year Bachelor) participated as tutors in the 22^nd Supercomputing Contest, which was jointly held by the Global Scientific Information and Computing Center at the Tokyo Institute of Technology and the Cybermedia Center at Osaka University. http://www.gsic.titech.ac.jp/supercon/main/attwiki/index.php?SupercomputingContest2016
2016.07.02	Dr. Miquel Pericas visited our group. Dr. Miquel Pericas from Chalmers University of Technology visited our group for collaboration on Task Assembly Objects (TAO) for Fast Multipole Methods (FMM).

2025.01.23	Two papers by Qi Sun (1st year PhD) in collaboration with Sakana AI were accepted to ICLR2025 Two papers by Qi Sun (1st year PhD) in collaboration with Sakana AI “Transformer^2 Self-adaptive LLMs” and “An Evolved Universal Transformer Memory” were accepted to the International Conference on Learning Representations (ICLR).
2025.01.23	Two papers by Satoki Ishikawa (2nd year Master) were accepted to ICLR2025 A paper by Satoki Ishikawa (2nd year Master) in collaboration with AIST “Local Loss Optimization in the Infinite Width: Stable Parameterization of Predictive Coding Networks and Target Propagation” and another in collaboration with OIST “PhiNets: Brain-inspired Non-contrastive Learning Based on Temporal Prediction Hypothesis” were accepted to the International Conference on Learning Representations (ICLR).
2025.01.23	Two papers by Taishi Nakamura (1st year Master) were accepted to ICLR2025 A paper by Taishi Nakamura (1st year Master) in collaboration with NII and Sakana AI “Drop Upcycling: Training Sparse Mixture of Experts with Partial Re-initialization” and another in collaboration with Sakana AI “Agent Skill Acquisition for Large Language Models via CycleQD” were accepted to the International Conference on Learning Representations (ICLR).

2025.01.23	Two papers by Qi Sun (1st year PhD) in collaboration with Sakana AI were accepted to ICLR2025 Two papers by Qi Sun (1st year PhD) in collaboration with Sakana AI “Transformer^2 Self-adaptive LLMs” and “An Evolved Universal Transformer Memory” were accepted to the International Conference on Learning Representations (ICLR).
2025.01.23	Two papers by Satoki Ishikawa (2nd year Master) were accepted to ICLR2025 A paper by Satoki Ishikawa (2nd year Master) in collaboration with AIST “Local Loss Optimization in the Infinite Width: Stable Parameterization of Predictive Coding Networks and Target Propagation” and another in collaboration with OIST “PhiNets: Brain-inspired Non-contrastive Learning Based on Temporal Prediction Hypothesis” were accepted to the International Conference on Learning Representations (ICLR).
2025.01.23	Two papers by Taishi Nakamura (1st year Master) were accepted to ICLR2025 A paper by Taishi Nakamura (1st year Master) in collaboration with NII and Sakana AI “Drop Upcycling: Training Sparse Mixture of Experts with Partial Re-initialization” and another in collaboration with Sakana AI “Agent Skill Acquisition for Large Language Models via CycleQD” were accepted to the International Conference on Learning Representations (ICLR).
2024.10.09	Okazaki Lab, Yokota Lab, and AIST have released large language models Llama3.1-Swallow-8B and Llama3.1-Swallow-70B. Okazaki and Yokota labs at the Tokyo Institute of Technology, and the National Institute of Advanced Industrial Science and Technology have released Llama3.1-Swallow-8B and Llama3.1-Swallow-70B, which are continually pretrained on Japanese from Llama3.1-8B and Llama3.1-70B.
2024.07.01	Okazaki Lab, Yokota Lab, and AIST have released large language models Llama3-Swallow-8B and Llama3-Swallow-70B. Okazaki and Yokota labs at the Tokyo Institute of Technology, and the National Institute of Advanced Industrial Science and Technology have released Llama3-Swallow-8B and Llama3-Swallow-70B, which are continually pretrained on Japanese from Llama3-8B and Llama3-70B.
2024.03.15	The joint work by Kazuki Fujii (4th year Bachelor) and Taishi Nakamura (3rd year Bachelor) with Okazaki Lab won the Outstanding Paper Award at NLP2024. The joint work by Kazuki Fujii (4th year Bachelor) and Taishi Nakamura (3rd year Bachelor) with Okazaki Lab “Swallow Corpus: Japanese large-scale web corpus” and “Building a large-scale language model that is excellent in Japanese using continuous pre-learning” won the Outstanding Paper Award at NLP2024.
2024.03.11	Okazaki Lab, Yokota Lab, and AIST have released large language models Swallow-MS and Swallow-MX. Okazaki and Yokota labs at the Tokyo Institute of Technology, and the National Institute of Advanced Industrial Science and Technology have released Swallow-MS and Swallow-MX, which are continually pretrained on Japanese from Mistral-7B and Mixtral-8x7B.

Institute of Science Tokyo, Institute of Integrated Research, Supercomputing Research Center

News

2025

2024