Machine Learning » MLOps
A Simple Introduction
Machine Learning Model
A Machine Learning Model can be defined as a trained program make predictions, recognize patterns or behaviors based on data/previous experience with a given dataset, but before everything, we need to train the model over a specific set of data training, giving it an specific algorithm (Supervised Learning) that uses computational methods to learn information from the data training without depending on a specific equation, and than, the result (the model) will be used to reason and learn from those specific type of data/patterns/behaviors.
Developing ML Models involves a lot of iterations, it is an experimental process where you need to change different model parameters, data preprocessing steps, and many others which could result in new data and model versions to achieve the optimal model. Every failure in these processes leads to rework, which means - repetition of a lot of steps, which means - wasted time, wasted computational resources, which can mean - incremental cost.
Types of Machine Learning Models and Algorithms
Machine Learning Experiment Tracking
Well, if we need to test different model parameters/data preprocessing and others in each step, so would be imperative to tracking our experiments (experiment management):
evaluation models, hyperparamentes, data, metrics, charts, codes
scripts used to running the experiment
tracking parameter configurations
use different training or evaluation data
run the same code in different environments
organize and compare those experiments
model weigths
performance visualizations
sample predictions
prediction distributions
and others
some images (graphs) were collected on google just to illustrate the post, they do not necessarily have useful information.
will be continue soon……