To start using PyGGi, first register using a valid email address and create an account. Login using the valid credentials to reach the personal dashboard. To use PyGGi online, click on the respective toolkits such as PyGGi Bank to explore the glass database, PyGGi Seer to predict the properties, or PyGGi Zen to optimize glass compositions.
To explore the glass database for selected glasses, first, choose the components of interest. The components can be chosen by simply clicking on it. You can also search for the components in the search box on the right top. The selected components are displayed on the top. Once the components are selected, choose a property of interest. If you also want to see the source of the property, select “With Reference”. Once all the selections are done, click on “Get Preview”. You will be able to see all the results associated with the selected components.
The experimental data in PyGGi Bank has been obtained from various literature and the open-source glass database SciGlass. The GitHub repository of the database can be found here.
To predict the properties of your glass compositions of interest, first, choose the components of interest. The components can be chosen by simply clicking on it. You can also search for the components in the search box on the right top. The selected components are displayed on the top. Once the components are chosen, click on the “Add and proceed” button.
The next page will display the chosen components. Fill the percentage of each of the components in mol%. Ensure that the sum adds up to 100. Select the methodology of interest, that is, “Neural Network” or “Gaussian Process”. The results displayed will be based on the model chosen. Details of the models can be found here. The philosophy of the models are briefly explained below.
- Neural Network:
A neural network is a network composed of artificial neurons or nodes, which learns the input-output relationship from databases. Here, large database of glass properties from literature is used to develop neural networks for predicting the properties
- Gaussian Process:
Gaussian process regression assumes that the experimental data points used for training come from a multivariate normal distribution, i.e., every finite linear combination of them is normally distributed. Thus, by training on the large datasets of glasses Gaussian process allows predicting the properties of glasses along with the standard deviation in the prediction.
Once the methods are selected, click on the “Get Results” button. The results of all the properties are displayed in the next page. These results can be saved by clicking on “Save the Result”.
PyGGi Seer and its commercial version PyGGi Seer Pro can be downloaded for offline use. To use the downloadable version offline, click on the Download button on the top panel of the website to directly download the offline version of PyGGi Seer for your operating system.
Windows: Unzip the downloaded package and run the setup file- PyGGiSetup.exe to install the software to your preferred location. After installation, click on the PyGGi icon to launch the software.
Open the .dmg file and move the software to the Applications folder. Click on the PyGGi icon to launch the software.
PyGGi Seer & PyGGi Seer Pro:
|PyGGi Seer||PyGGi Seer Pro|
|Number of oxides available||9||34|
|All components available for a glass formation||Limited to 3||✔|
|Input as range for individual||✔||✔|
|Screens at the same time||No Limit||1|
Please write to us to get "PyGGi Seer Pro" or a custom-tailored solution for your organisation.
To optimize glass compositions toward a targeted value, first, choose the components of interest. Then, choose the target property along with the target value. Note that the target value can be an actual number or one among the keywords “min” or “max” to minimize or maximize, respectively, the chosen property. Additional constraints can be added to either the compositions or other properties. For example, it can be specified that among the chosen input components, SiO2 should be between 50% and 60%, and Na2O should be between 5% and 15%. Similarly, constraints can be added to other properties as well. Once the components and constraints are chosen, select the optimization method of interest from “Genetic Algorithm”, “Ant Colony Optimization”, “Particle Swarm Optimization”, or “Gradient Descent”. Once the method is chosen, click on the “Get Results” button. The displayed composition represents the optimized glass composition with targeted properties and constraints.