Visualize numeric pixel data as an image to recognize digit patterns. Watch this tutorial to learn the key concepts and techniques.
Understand how neural networks interpret numerical arrays to recognize visual patterns, similar to the human brain's perception. Learn how data visualization aids in comprehending machine learning processes.
Key Insights
- The article illustrates how numerical arrays representing images, like a handwritten number 5, are interpreted by neural networks based on the relative position and values of numbers.
- Visualization techniques, such as displaying arrays as images without line wrapping, clarify how numerical data corresponds to visual shapes, aiding understanding of machine learning concepts.
- The neural network identifies patterns in data similarly to how the human brain perceives visual patterns, allowing it to recognize numbers based solely on numerical relationships.
This lesson is a preview from our Data Science & AI Certificate Online (includes software) and Python Certification Online (includes software & exam). Enroll in a course for detailed lessons, live instructor support, and project-based training.
Common ML Algorithms
-
Linear/Logistic Regression
Interpretable baselines — start here.
-
Random Forest
Robust, handles mixed data, minimal tuning.
-
Gradient Boosting
XGBoost/LightGBM dominate tabular ML competitions.
-
Neural Networks
Best for images, audio, text, and high-dimensional data.
Let's take one more look at this data—this array—as something that might visibly resemble a 5. We saved an image of the 5 array so that you could see how this works. The 28-item rows in the version above aren’t clearly visible as a 0 or a 5. Let me revert that to the version displaying the 5. It's kind of visible, but Jupyter Notebook is wrapping each of these lists—each of these rows—onto two lines, which makes it tough to visualize.
But we made an image to help you visualize this, where it’s not wrapped. Let’s take a look at it. Here’s the same list that we saw before, but we’ve stretched it out so that each list appears as a full row.
And now you can see—this looks like a 5. However, of course, the computer isn't going to look at this and say, "That's a 5, I see it." It’s going to have to learn based solely on the numbers, their positions relative to one another, and the intensity of each number. But this is a good visualization showing that—even just looking at the numbers—we can recognize it as a 5, even though we aren’t machine learning models ourselves.
But this is what your brain is doing. It’s similar to what the neural network is doing. It’s looking for patterns, and it’s detecting a pattern in the non-zero values here.
That looks kind of like a 5. And our hope is that our neural network can follow a similar process.
