Arash Arbabi

Recently, I have been experimenting with a Graph Neural Network (GNN) and have attempted to train it on the FS-MOL dataset. However, I have used some additional preprocessing steps that were not included in the original FSMOL dataset. It is crucial for neural network researchers to visualize the exact input that goes in the neural network. This ensures that the input data is correct and there are no errors in the preprocessing steps that may result in faulty data....

The whole world is made up of molecules, you are made of molecules! To use molecules inside computer programs and especially neural networks, we need to be able to digitize them. Here I am going to show how molecules are represented inside the computer. SMILES Strings SMILE strings are very straightforward. They are always available on Wikipedia. For example, if you wanted to depict Ibuprofen as a SMILE string you can simply go to the Wikipedia page and look for SMILES inside the right table....

In this series we are going to talk about few-shot learning. Few shot learning is useful when there are many different related Tasks but data for each one of them is lacking. To give an example: There is branch of science called Math. But there are different subject areas of Math. In order to learn Math you have to learn the different subject areas. For example one might imagine that a person who knows how to do multiplication will have an easier time learning derivatives than a person who does not know how to do multiplication....

I remember the first days that I entered high school, I was registered at a high school that specialized in sending students to national scientific olympiads and since I was very fond of my love of computers I wanted to participate in the computer national olympiad. However, this dream soon went into dust when I was first introduced to combinatorics. I had never seen a subject as vague as combinatorics. So many symbols and new operations....