Key Takeaways in Solving an Open-ended Problem


Hopfield Network mimicks how memory works. It is an autoassociative memory which can learn (remember) one or more patterns and given the partial (incomplete) learned pattern (some information of pattern are missing), the network can recall and recover a whole learned pattern.
There are two steps in this application, train and predict.

Train
- In this step we will train 3 patterns on the screen
- You can draw the 3 patterns by yourself by clicking the grid
- Or you can use the preset image provided
- Click the Train 3 Patterns button to start training and updating network weights

Predict
- In this step we the network will predict the learned pattern once you click the Predict button
- You can draw incomplete pattern similar with the learned pattern in the training phase
- Or you can use the preset image + noise provided


Visit the site here <u>https://hopfield-network.mabdh.now.sh/</u>


Hopfield Network


A snowman lives in a dangerous place. Its home is a tent. However, there are many locations with high intensity radiation. The snowman is going home but it does not know yet which lane are the safest. The algorithm will make snowman decide the most optimal (safest) trajectory to the home (a tent).  

<img src="/images/qlearning.jpg" width="400" style={{display: "block", margin: "0 auto"}}/>

There are several parameters that can be modified in this simulation.

- Grid Size (Integer)
- Start Point (Integer,Integer) - starting point of the snowman
- Finish Point (Integer,Integer) - location of the tent
- Number of Obstacles per Row (Integer) - At most n obstacles per row. Will be randomized.
- Number of episodes (Integer) - For training purpose. The more episode, the more chance to learn.

Visit the site here https://reinforcement-learning-q-learning.mabdh.now.sh/



Reinforcement Learning - Q-Learning


Location-based insurance premium prediction implemented as android application. The app uses geofencing to locate the user. The insurance premium data for every location in a city will be various depends on its condition (e.g. the locations with higher risk will have higher premium).Users can register themselves in on-demand manner when they are approaching the location with high risk.
All locations are updated in near real-time.

An MQTT back-end is used as a message broker to publish locations information and their premium that can be varied over time.

<img src="/images/asuran/2.png" width="200" style={{display: "block", margin: "0 auto"}}></img>
<img src="/images/asuran/3.png" width="200" style={{display: "block", margin: "0 auto"}}></img>
<img src="/images/asuran/6.png" width="200" style={{display: "block", margin: "0 auto"}}></img>
Android application Screenshots: 1. Login screen. 2. Initialization screen shows three important locations of the user (home, last location where he activates the insurance for car, and last location where he activates the life insurance). 3. Predictive locations where the risk might occur. The red color shows the highest risk.

- Date: October 2016
- Team: Muhammad Abduh, Rian Josua Masikome, Galih Gilang Wicaksono
- A Project of Zurich Insurhack 2016


Asuran


Generating spotify playlist based on an image's mood. Implemented as Web application using Node.js and iOS application and using Spotify API.

<img src="/images/mudjik/mudjik_presentation.png" width="500" style={{display: "block", margin: "0 auto"}}></img>

- Date: October 2016
- Team: Muhammad Abduh, Rian Josua Masikome
- A Project of Music Hack Day Berlin 2016


Mudjik



"The theremin (/ˈθɛrəmɪn/[1] THERR-ə-min; originally known as the ætherphone/etherphone, thereminophone[2] or termenvox/thereminvox), is an early electronic musical instrument controlled without physical contact by the thereminist (performer).  It is named after the Westernized name of its Russian inventor, Léon Theremin, who patented the device in 1928." - Wikipedia

<img src="/images/theremin/index.jpg" width="400" style={{display: "block", margin: "0 auto"}}/>

Creating an interactive web-based theremin with javascript, HTML 5, and web-audio API.

Check this on http://mabdh.github.io/interactive_theremin/


- Date: October 2016
- Team: Muhammad Abduh, Rian Josua Masikome
- A Project of Music Hack Day Berlin 2016
