| <html><head><style>body { | |
| color: black; | |
| } | |
| </style></head><body><h1 id="aggregated-weight-matrices-used-in-transformers-enhance-the-predictive-power-of-network-medicine-">Aggregated Weight Matrices Used in <em>Transformers Enhance the Predictive Power of Network Medicine</em></h1> | |
| <p><em>Jonah Spector, Andrés Aldana Gonzalez, Michael Sebek, Joseph Ehlert, Christian DeFrondeville, Albert-László Barabási</em></p> | |
| <h2 id="data-structure">Data Structure</h2> | |
| <p>This directory contains all aggregated weight matrices used to make figures in the manuscript. The data structure is as follows</p> | |
| <pre><code>aggregated_matrices | |
| aggregated_attentions | |
| dcm_samples | |
| fine_tuned | |
| <span class="hljs-built_in">max</span> | |
| layer_0 | |
| layer_1 | |
| layer_2 | |
| layer_3 | |
| layer_4 | |
| layer_5 | |
| pretrained | |
| <span class="hljs-built_in">max</span> | |
| genecorpus | |
| fine_tuned | |
| <span class="hljs-built_in">max</span> | |
| pretrained | |
| <span class="hljs-built_in">max</span> | |
| aggregated_embeddings | |
| dcm_samples | |
| fine_tuned | |
| <span class="hljs-built_in">max</span> | |
| layer_input | |
| layer_0 | |
| layer_1 | |
| layer_2 | |
| layer_3 | |
| layer_4 | |
| layer_5 | |
| pretrained | |
| <span class="hljs-built_in">max</span> | |
| genecorpus | |
| fine_tuned | |
| <span class="hljs-built_in">max</span> | |
| pretrained | |
| <span class="hljs-built_in">max</span> | |
| </code></pre><p>In each <code>max</code> folder, there are 3 files, called <code>counts.pkl</code>, <code>vals.pkl</code>, <code>gene_dict.pkl</code>. The counts matrix is a square matrix of size NxN, where N is the number of unique genes in the sample set. It counts how many times each pair of genes co-occurs in a sample. The vals matrix is the same shape, and contains the aggregated weight between each pair of genes. <code>gene_dict.pkl</code> is a dictionary that specifies the index of each gene in the matrix.</p> | |
| </body></html> |