Since we set up TOSHI STATS AI LAB, I research many theories and algorithms. Today, I want to present the first result of our machine intelligence from AI LAB. It uses attention mechanism which can be applied to many application. If you want to know more, could you see our blog here? Stay tuned!
Inspired by "AlphaGo by DeepMind", TOSHISTATS researches "deep reinforcement learning (DRL)" since 2016. Today, I announce that TOSHISTATS starts applying DRL to TEXT GENERATION to create Machine intelligence. I am very excited as DRL could work in the field of "Natural Language Processing" very well. I will update the progress of that. Stay tuned!
Since AI-Go player "AlphaGo" beat human professional Go player in 2016, Deep Reinforcement Learning becomes one of the key technologies to develop machine intelligence. Today I am very glad to announce that "OpenAI Gym (1)" is introduced as the framework to develop this technology. Although there are several frameworks for the environments of Deep Reinforcement Learning, OpenAI Gym is widely used in the communities of researchers and there are many open source projects with it. OpenAI Gym provides many game-based environments for Deep Reinforcement Learning. Here are several examples of the environments and the code to run it.
Hope I can update the progress of our development of Deep Reinforcement Learning soon.
(1) OpenAI Gym, Greg Brockman, Vicki Cheung, Ludwig Pettersson, Jonas Schneider, John Schulman, Jie Tang, Wojciech Zaremba, 2016
Happy new year everyone! As a new year starts now, I would like to announce our new project in TOSHI STATS.
This is about "Machine intelligence". With inputs of images/videos and texts, I would like to do experiments of "what can computers say in our languages?". I have been developing our base model to do that. With introducing both images and text at once, our model must be more intelligent than before. I will update the progress the project on our blog. Hope you can enjoy the results soon!
Last week, Google released an awesome tool for anyone who is interested in computer vision and deep learning. It is called "Tensorflow Object Detection API". TOSHI STATS introduces this awesome API as our core computer vision framework. The image above is one of the results of detection task from the API.
I would like to explain why Object Detection API is awesome
- It has major classification models.
- It enables us to combine classification models with object detection frameworks
- It can be run locally and on Google Cloud Platform
We will make the most of this API to get insights from images and videos. See you soon!
"Tensorflow Object Detection API" was released by Google on June 15 2017
"Speed/accuracy trade-offs for modern convolutional object detectors."
Huang J, Rathod V, Sun C, Zhu M, Korattikara A, Fathi A, Fischer I, Wojna Z,
Song Y, Guadarrama S, Murphy K, CVPR 2017
TOSHI STATS is going to "Self-driving car engineering" in 2017. This technology is consist of deep learning, computer vision, sensor fusion and robotics. It must be exiting as this technology is brand new. We will update the progress of our technology in 2017. See you there again.