logo

Aiex.ai

3 Mins Read

Model-Driven Vs Data-Driven Approach

Data-Driven approach

Model-Driven Vs Data-Driven Approach

It is challenging to improve the accuracy and output of deep learning models. There are two main ways to improve a model’s output: by improving the deep learning model or by improving the dataset. Developing a deep learning model is a complex process and In order to improve its performance, many challenges must be overcome. The performance of the model can be improved by adjusting hyperparameters, trying multiple algorithms, ensemble models, etc. These methods can be very challenging to use and require a lot of hardware resources and trial and error. The model-driven approach refers to improving the output of the deep learning model by focusing on the model. An alternative method is the data-driven approach.

It should be noted that both of these two approaches are necessary to improve the output of a deep learning model, but will the effort spent improving the model be worth it if we spend the time improving data? We will examine this question throughout this article.

Model-Driven Approach

Generally, this category includes techniques that improve output by changing the model. This includes changing the algorithm, tuning hyperparameters, and using ensembles. These methods require some trial and error to get right which is time-consuming and limited by hardware availability. Moreover, a deep knowledge of deep learning models and their parameters is required to apply these techniques, which makes it more difficult.  Instead of manipulating the model to improve accuracy, it would be better to use the data-driven approach, which can sometimes greatly improve the output with a few changes.

Problems of Model-Driven Approach
Figure1. Problems of Model-Driven Approach

Data-Driven Approach

The Data-Driven approach consists of improving  the training data and problem definition to enhance  the results. The following techniques fall under this approach:

  • More data
  • Data Augmentation
  • Data Rescaling
  • Feature Selection
  • Dealing with Outliers
  • Dealing with Unbalanced Datasets

Clearly, data-driven approaches do not require deep and extensive knowledge of deep learning models, don’t require much training and are much less time-consuming. Moreover, there is great potential for improving accuracy and output with this approach.

The impact of dataset size on the performance of deep learning models
Figure2. The impact of dataset size on the performance of deep learning models. Slide by Andrew Ng

AIEX’s Data-Driven Approach

AIEX has been able to offer a data-driven approach for a variety of industries by providing an end-to-end framework. With the data-driven approach, no deep learning knowledge is required, therefore various industries that can benefit from using artificial intelligence and deep learning algorithms can easily create advanced models using the AIEX platform. The AIEX platform has made it easy to improve model performance by providing data-driven tools such as augmentation, image tiling, dataset health check, image engine, etc.

Various industries can use the data and problem-focused models offered on the AIEX platform to benefit from artificial intelligence in their fields.

Related articles
augmentation
To train a model or use transfer learning in machine vision, there must be enough data. Data Augmentation is...
Train, Test, and Validation Datasets
An artificial intelligence model output is affected by how we divide the input dataset. There are several factors to...
Tensorboard
In this article, we will introduce Tensorboard and explain how it can be used on AIEX....
Loss-Function
The majority of machine learning algorithms work by minimizing or maximizing an 'objective function'. Loss Functions are a group...
backbone
Backbone is a network that extracts a feature map of the input image , the map is then utilized...
evaluation metrics
This article examines the different metrics used to evaluate machine vision models, and the metrics implemented on the AIEX...
Subscribe to our newsletter and get the latest practical content.

You can enter your email address and subscribe to our newsletter and get the latest practical content. You can enter your email address and subscribe to our newsletter.