greentec's blog game designer, scripter, researcher

en kr

CoG 2019 attendance review

In the last week, my paper was accepted at CoG(Conference on Games) 2019 held at London QMUL University, so I presented my paper at CoG2019 and came back home. It was my first time to speak English when I attended the first conference of my life, but I did what...

Learn Reinforcement Learning (5) - Solving problems with a door and a key

In the previous article, we looked at the Actor-Critic, A2C, and A3C algorithms for solving the ball-find-3 problem in Grid World and did an action visualization to see how the agent interpreted the environment. A3C was a high-performance algorithm with a simple idea of using multiple agents instead of a...

Learn Reinforcement Learning (4) - Actor-Critic, A2C, A3C

In the previous article, we looked at several methods to improve the performance of the DQN algorithm and the Deep SARSA algorithm to solve the ball-find-3 problem in Grid World. Today we will look at Actor-Critic algorithms that outperform previous algorithms in ball-find-3, and visualize agent actions. We will also...

Learn Reinforcement Learning (3) - DQN improvement and Deep SARSA

In the previous article, we learned the DQN in Grid World that was developed by Deepmind and used to play Atari games. The average reward of the random action agent was obtained from -14 to -17, whereas the average reward of the DQN agent was obtained from -11 to -12....

Learn Reinforcement Learning (2) - DQN

  Value Function and Discount Rate In the previous article, we introduced concepts such as discount rate, value function, as well as time to learn reinforcement learning for the first time. The two concepts are summarized again as follows. Discount Rate: Since a future reward is less valuable than the...

Learn Reinforcement Learning (1) - Value Function

  What is reinforcement learning? In 2016, AlphaGo versus Lee Sedol became the topic of the event in which artificial intelligence won the world’s first professional supremacy in Baduk. In 2018, OpenAI’s researchers at DOTA2, a 5-to-5 team-fighting game, won a pro-amateur team in a pre-determined heroic confrontation, and in...

Shadertoy - Rain drops

“Heartfelt” by BigWIngs   Introduction Today, let’s analyze the effects of raindrops, one of the most visible effects on Shadertoy. The code to analyze is ‘Rain drops on screen’. The comments in the code are meticulous, which helped me to analyze it. So let’s get started. ‘Rain drops on screen’...

Shadertoy - 'Creation'

“Creation by Silexars” by Danguafer   Introduction Today I’m going to analyze one of Shadertoy’s popular shaders, ‘Creation’. If we subtract an annotation, it is a total of 19 lines of short code, but as we can see in the image above, it does not feel as simple. I unintentionally...

Shadertoy - Fire Shader

“Pyroclastic fireball” by Duke   Introduction Today I’m going to look at one of the most basic things in shaders, the fire shader. We’ll look at the basic components of the shader, including random, noise, fbm, and bumpMap, and see how they can be represented in the fragment shader used...

Shadertoy - Metaball

“Metaballs++ “ by TheBuoy   Introduction Following the last article, we will look at one of the most fundamental implementations of 3D graphics, the metaball. The idea of this posting is heavily dependent on Ryan Geiss’s this article. From his home page, he has been working on Google since 2010....

Shadertoy - Default Shader Analysis

“Canyon” by iq   Introduction Shader is a language used to render computer graphics, primarily light and shadows. With the development of hardware, especially the GPU, Shader is now able to express many things that previously could not be expressed as performance problems. Shadertoy is an online shader code sharing...

Create a hexagonal map

image link   Introduction It is now a technology that should not be used because of the risk of hacking because of security vulnerabilities, but about 10 to 15 years ago Flash created the best web contents. JavaScript has not yet set the right standards, so the code you need...

Create a random circular path

Introduction Two weeks ago, I had to find a non-overlapping circular path in a square grid. The requirements were as follows: circulation path (first == end) a clean 1 pixel line that does not overlap some randomness Depending on the problem definition, we can also use the search algorithm like...

Start blogging

I’m starting a new blog on Github Pages. The resources originally in the greentec.github.io path are located in the Inventory of the Sidebar.