Apply Machine Learning to Design thinking

Tram Ho

Too many people think that AI is the business solution for many of the problems we face today.

Perhaps having a neural network for a data set large enough can help us make decisions about a problem and impact users. But in order to use this data effectively to create a product, we need to understand if the problem is even real. We need to understand people because at the end of the day, companies are consumer products that address the problems that people face.

To begin with, we need to ask ourselves whether the problem is a real problem or just a reason to apply AI innovation. On a larger scale, we need to ask if adopting a person-centered approach to data science is a way to solve problems in AI.

Many industry leaders are becoming strong supporters of this idea,

“I’m worried that the enthusiasm for AI is preventing us from calculating with its looming effects on society,” Li argues in the work. “Although its name is nothing, there is nothing artificial about this technology – it was created by humans, intended to behave like humans and affect people. So if we for it to play an active role in tomorrow’s world, it must be guided by people’s concerns. ” Fei Fei Li, Stanford University said

Jim Guszcza, data scientist at Deloitte Consulting LLP said, “An AI revolution is happening right now, but I believe it needs to be complemented by a design revolution,”

After empathizing and performing user research, we can know whether a problem is feasible to be solved with artificial intelligence. After understanding the problem and studying its use cases, we need to ask more questions to understand whether deep learning algorithms are an approach compared to conventional prediction algorithms. Some ways to know that are:

  • If it’s a problem in that the Neural Network gives you more advanced than traditional algorithms
  • If there is obviously enough data to train.
  • If the problem is complicated enough.
  • If you have the right kind of neural network to solve the problem.

Once we have determined that AI is a viable solution, the steps to create a Neural Network Pipeline are:

  • Find and determine the most appropriate neural network architecture for the problem
  • Divide data into batches
  • Pre-processing data with image processing
  • Augment data to increase size
  • Put courses into the Neural Network for training
  • Test your model and save it for future use

This Neural Network Pipeline can be thought of using the innovative 5-step design thinking method proposed by David M. Kelley. Design thinking process, a design approach that provides a solution-based approach to problem solving. These are steps that can be taken to approach Deep learning by using design thinking.

Deep learning uses Design Thinking

Step 1: Empathy and analysis

The first stage of the Design Thinking process is gaining empathetic understanding of the problem you’re trying to solve and your users are good.

Design Thinking : Get a better understanding of the problem you want to conquer by talking to people / users and advising the expert about the problem or getting deeper into the problem to better understand the problem.

Deep learning : Understanding a consumer problem in the real world, where the Neural Network application will significantly impact the problem. Neural Network is the solution for everything. We need to understand real-world problems from our users to create something meaningful using AI. We can start by understanding the user ‘s primary decision and capturing variables and metrics will be a better predictor of those decisions.

Step 2: Determining and Synthesis

Defining means gathering all the information gathered in the Empathy and data analysis steps to create a meaningful statement.

Design Thinking : Identify the problem as a problem statement in a person-centered manner.

Deep learning:

  1. Find or aggregate a data set based on the problem being solved.
  2. Download data in an appropriate place.
  3. Data preparation – randomization, visualization to see imbalances or relationships, preprocessing, splitting and increasing data will be sent for training.
  4. Divide data into training, evaluation, and testing.

Step 3: Ideas

Ideate is the stage of your design process in which you aim to create radical design alternatives. Give as many solutions to the problem that were thought of in the previous step.

Design Thinking: Use information from previous stages to generate ideas and think as many potential solutions as possible.

Deep Learning: In this step choose a model for your specific problem. Many models have been created for images, sequences such as text or music, numerical data, or text based data. If not, you can even define your own model architecture by adding layers one by one until you are happy with your network.

Step 4: Prototype and Adjustment

Prototyping creates some inexpensive, scaled-down versions of a product or specific features found in a product. The team will work to create some inexpensive products with specific features. This allows Design Thinkers to investigate possible solutions to problems that have been identified in the earlier stages of the Design Thinking process.

Design Thinking: Simulation design is created based on all ideas created in the previous period. These are low or high fidelity wireframes that implement the concept of ideas.

Deep learning: In this step, we train the model using the training data and then we start manipulating the data and adjust the hyperparameter based on the training results.

Step 5: Check and validate

Testing is an opportunity to get feedback on your solutions, fine-tune the solutions to make them better and keep repeating them conceptually. Design Thinking: We test our prototypes using user test techniques to see how they solve the problems that we originally analyzed in previous stages.

Deep Learning: We test the model using the test data set that provides the gold standard used to evaluate the model. Based on test results and validation, we repeat the hyperparameter adjustment process to improve the model’s accuracy.

  1. Research and prototyping
  2. Produce models for actual end users
  3. Screening system in the real world

Using design thinking for deep learning provides a framework and a process for something that has many steps and is a complex process with many stages. The prospect of design thinking helps integrate human perspectives into solving problems in AI and emphasizes repeatedly to build neural networks in the same way that designers create designs.

Share the news now

Source : Viblo