| Figure 19.1 - Various types of sequence-to-sequence models.png | 23.9 KB | |
| Figure 19.10 - Stacked LSTM classification—cross-validation performance.png | 135.5 KB | |
| Figure 19.11 - Stacked LSTM regression—out-of-sample performance.png | 32.4 KB | |
| Figure 19.12 - Original and transformed time series.png | 398.3 KB | |
| Figure 19.13 - Cross-validation and test results for RNNs with multiple macro series.png | 226.6 KB | |
| Figure 19.14 - Cross-validation for RNN using IMDB data with custom embeddings.png | 150.8 KB | |
| Figure 19.15 - Cross-validation and test results for RNNs with multiple macro series.png | 159.9 KB | |
| Figure 19.16 - Cross-validation and test results for RNNs with multiple macro series.png | 116.4 KB | |
| Figure 19.17 - Cross-validation test results for RNNs using SEC filings to predict weekly returns.png | 101.1 KB | |
| Figure 19.2 - Recurrent and unrolled view of the computational graph of an RNN with a single hidden unit.png | 33.4 KB | |
| Figure 19.3 - Information flow through an unrolled LSTM cell.png | 47 KB | |
| Figure 19.4 - The logic of, and math behind, an LSTM cell.png | 61.3 KB | |
| Figure 19.5 - The three dimensions of an RNN input tensor.png | 26.4 KB | |
| Figure 19.6 - Input-output pairs with a T=5 size window.png | 82.9 KB | |
| Figure 19.7 - Cross-validation performance.png | 155.3 KB | |
| Figure 19.8 - RNN performance on S&P 500 predictions.png | 301.3 KB | |
| Figure 19.9 - Stacked LSTM architecture with additional features.png | 28.8 KB | |
