Structured data organizes information in rows and columns. Each row represents an independent sample or record, and row order carries no inherent meaning. section 4.3
Columns can contain different kinds of information, such as numerical measurements and labels, so the table is heterogeneous. Neural networks instead consume homogeneous numeric tensors, requiring deliberate encoding before training. section 4.3 section 4.3.1
Wine quality table
A wine dataset stores eleven chemical variables and a quality score in twelve CSV columns. A possible task is predicting quality from the chemical measurements. section 4.3.1
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality
Inspect the wine table’s column names, from acidity and sulfur measurements to alcohol and quality. section 4.3.1
Structured tabular data differs from time series: adding meaningful date or time information makes row order relevant and changes the representation problem. section 4.4
