TabNAS

A Neural Network model generalized for any tabular dataset (Classification/Regression Task)

TabNAS is a generalized neural network model that can be fed any kind of tabular dataset. It has been generalized for both classification as well as regression tasks.

We have benchmarked the model for 4 datasets - two of which are classification tasks while the other two are regression. The datasets tested on are - the Adult Income dataset, Car dataset, Housing Prices dataset

Objective

We have generalized the function in order to automatically understand:-

The number of Features
Each feature type - (Categorical or Numerical)
The number of layers and neurons required for each task

We also provide a size - performance tradeoff which ensures that the model takes up less space on the disk as well as has a good performance overall. The tradeoff between size and performance is observed to ensure the best possible values for hyperparameters can be applied.

Input

We require the user to input the following parameters:-

data: Dataset
target_var: Target Variable - The variable which needs to be predicted
classification = True: Classification(True/False) - Whether its a Classification task or Regression task
lr: Learning rate - The learning rate for the optimizer to train on

Neural Architecture Search(NAS)

Neural architecture search (NAS) is a technique for automating the design of artificial neural networks (ANN), a widely used model in the field of machine learning. NAS has been used to design networks that are on par or outperform hand-designed architectures. We employ evolutionary algorithms for optimizing,

params = {
  "number-of-layers",
  "neurons each layers",
  "activation function"
}

Output

The project saves the model's weights which can then be loaded and worked on.

To Execute