Event Report: NTT Communication Science Laboratories Open House 2018

(1) “The real worth of quantum computers with elementary operations—Analysis of computational power of gate-based quantum computers,” by Dr. Yasuhiro Takahashi, Media Information Laboratory

Dr. Takahashi focused on a gate-based quantum computer, which has theoretical evidence of being more powerful than today’s computers. There are many problems in realizing gate-based quantum computers equipped with a sufficient amount of computational resources, but he nevertheless introduced two of his recent results demonstrating that gate-based quantum computers are still powerful even under the more realistic condition where computational resources are limited (Photo 2).

Photo 2. Dr. Yasuhiro Takahashi giving research talk.

(2) “Role of haptics in improving wellbeing—Science and design of touch can enhance human flourishing,” by Dr. Junji Watanabe, Human Information Science Laboratory

Dr. Watanabe introduced a recently launched study on wellbeing. Wellbeing is gathering attention because research on designing mind-enriching information technology has been pursued recently from a standpoint that is different from the one aiming to merely achieve greater efficiency. He introduced several topics including measuring and specifying factors of wellbeing and modifying cognitive and empathetic attitudes towards oneself and others with haptic experiences (Photo 3).

Photo 3. Dr. Junji Watanabe giving research talk.

(3) “Beyond combinatorial explosion—Enumeration and optimization with Binary Decision Diagrams,” by Dr. Masaaki Nishino, Innovative Communication Laboratory

Dr. Nishino focused on a phenomenon called a combinatorial explosion, in which the number of possible combinations increases exponentially with the number of elements, which we have often experienced in a grouping or travel routes search. He explained how the number of combinations created by the combinatorial explosion can be counted by algorithms based on the data structure called a binary decision diagram (Photo 4).

Photo 4. Dr. Masaaki Nishino giving research talk.

(4) “Speech production and perception share common brain pathways—Investigation of the mechanisms of speech communication by speech conversion and brain imaging,” by Dr. Sadao Hiroya, Human Information Science Laboratory

Dr. Hiroya introduced research on neural mechanisms underlying speech communication and showed the existence of the common pathways in the brain between speech production and perception, even though it has been considered that the pathways are different. He also discussed the difference between human brain mechanisms and speech recognition and synthesis techniques (Photo 5).

Photo 5. Dr. Sadao Hiroya giving research talk.

4.1 Science of Machine Learning

• Finding similar voice recordings in big data—Graph index-based audio similarity search
• Learning feature combinations from multiple tasks—MOFM: low-rank regression for learning common factors
• Where do they come from? Where are they going?—“Data assimilation and navigation learning for crowd”
• Datafying cities—Event analysis by environmental sensing and machine learning
• Memory efficient deep learning for mobile devices—Quantized neural networks for model compression
• Optics makes machine learning much faster—Photonic reservoir computing for high-speed machine learning
• Interpreting deep learning from network structure—Detecting communities in trained layered neural networks

4.2 Science of Communication and Computation

• Can I borrow your quantum memory?—High-speed quantum computations with uninitialized qubits
• Designing fault-tolerant networks—Maximizing network reliability via binary decision diagrams
• Can computer translate considering context?—Context understanding tests for neural machine translation
• Early vocabulary development in late talkers—Collecting and analyzing data from pediatric medical fields
• Chatting with robots broadens your knowledge—Integration of chat and QA based on two-robot coordination
• Anytime, anywhere, we can speak like a native!—Speech rhythm conversion by mobile application
• Sharing enthusiasm between remote sites—Applause coding for bi-lateral immersive sharing

4.3 Science of Media Information

• Illumination-based color saturation control—Spectral operation using color enhancement factors
• Pay attention to the speaker you want to listen to—Computational selective hearing based on deep learning
• Solving two-choice questions makes AI clever—Deep pairwise comparison model for ASR hypothesis selection
• Estimating objects’ visuals only from audio—Cross-media scene analysis
• Cast shadows add dimensions—Projection mapping giving depth illusion to real objects
• Converting impression and intelligibility of speech—Speech attribute conversion using deep generative models
• Creating favorite images with selective decisions—Hierarchical image analysis and synthesis with DTLC-GAN

4.4 Science of Humans

• Predicting attention to the ears by the eyes—Auditory spatial attention revealed as pupillary response
• Measuring, understanding, and empowering wellbeing—Cross-disciplinary research toward “eudaimonic wellbeing”
• Understanding human hearing with AI—Analyzing auditory neural mechanisms with machine learning
• How do excellent batters hit the ball?—Cognitive processes revealed by body movements in batting
• How do excellent batters look at the ball?—Cognitive processes revealed by eye movements in batting
• The sooner you decide, the better you can localize—Visual motion processing in the perception and action
• Let’s FEEL shape and action by a force—Can we receive environmental information by Buru-Navi4?
• Feeling bumps on a flat sheet—Magnetic haptic printing technology