XCelBoost

Inspiration

Excel is everyone's favorite spreadsheets app. From running Doom to 3D physics engines, Excel can do it all. Thus, we figured that it should be possible to create a neural network on Excel too.

What it does

Our programme takes in preprocessed .csv files in the command line, and generates a neural network and predictions to machine learning problems (eg. binary classification) in the form of an Excel workbook.

How we built it

We implemented the neural network completely within Excel. We wrote a Python script that converts preprocessed .csv files into a neural network. All calculations will be performed within Excel using VBA.

Challenges we ran into

VBA was not the most intuitive language to use, and we didn't have much experience coding in it beforehand. We had to rely on MsgBox to print variables as pop-up alerts (cue infinite alerts and killing Excel in Task Manager).

The types of arrays and matrices were misleading to use as we had to declare nearly all matrices as "Variant" for the code to compile properly. Matrix multiplication was also very confusing. There were a lot of times where we got the dimensions mixed up, and we had to rely on our best friend, MsgBox, to help us debug the errors.

Accomplishments that we're proud of

It runs!!!! (Thanks to Leslie who did not sleep at all)

We now have a 2 layer neural network that is able to predictions with decent (?) accuracy, and we tested our data on binary classification datasets.

What we learned

VBA is not fun.

We also learnt more about neural networks, especially the mathematics behind it. (And we have newfound appreciation for the libraries that have done all this for us.)