Building a Machine Learning model involves the following steps.

  • Define what your model will do
  • Collect data
  • Choose model (Should read more about this)
  • Train model on your dataset
  • Evaluate model
  • Tune and deploy model

Tensors

Pytorch is a python library which helps in building an ML model. FIrst, lets talk about tensors. Imagine a spreadsheet which contains rows and columns of data, tensors are more complex which can hold data in several dimensions. The common term used is “containers” that holds numbers in a structured way.

Why is it important to convert data into tensors? it helps in easier computation of multi-dimentional data, it has a standardized format as they can represent image, numerical, text data in a consistent format, helpful in applying mathematical operations such as linear algebra.

So, as a first step, you would have to convert your data into tensors. This can be done with the help of the function torch.tensor() which can take 2 arguments- the numerical data (np array, list) and data type (int or float)

Layers

Now, you would have to choose an algorithm to run a model (I only learned about linear regression so, ill explain building a model with that). The function nn.Sequential() is an easier way to stack layers in a neural network.

model = nn.Sequential(
    nn.Linear(2,3),
    nn.ReLU(),
    nn.Linear(3,1)
)

Here, nn.Linear() applies the linear regression. The numbers enclosed in parenthesis denotes the no.of nodes, Begins with 2 input layers connects to 3 nodes in the second layer. nn.ReLU() (an enitre topic of its own) is an activation fucntion. Breifly, it takes number as input and gives a number as the output. This helps the model to learn faster. The above mentioned code is a simple model which has an input layer, a second layer in between (or a hidden layer) and an output layer. A model can contain n number or layers.

— more on my next blog.