Dataset for the challenge at the 2nd MODE workshop on differentiable programming 2022

<p>Data is in HDF5 format (with LZF compression). For specifics and details, please see&nbsp;<a href="http://github.com/GilesStrong/mode_diffprog_22_challenge">https://github.com/GilesStrong/mode_diffprog_22_challenge</a></p> <p>The training file contains two datasets:</p> <ul> <li>`&#39;x0&#39;`: a set of voxelwise X0 predictions (float32)</li> <li>`&#39;targs&#39;`: a set of voxelwise classes (int):</li> <li>0 = soil</li> <li>1 = wall</li> </ul> <p>&nbsp;</p> <p>The format of the datasets is a rank-4 array, with dimensions corresponding to (samples, z position, x position, y position).</p> <p>All passive volumes are of the same size: 10x10x10 m, with cubic voxels of size 1x1x1 m, i.e. every passive volume contains 1000 voxels.</p> <p>The arrays are ordered such that zeroth z layer is the bottom layer of the passive volume, and the ninth layer is the top layer.</p> <p>It can be read using e.g. the code below:</p> <p>&nbsp;</p> <p><em>with h5py.File(&#39;train.h5&#39;, &#39;r&#39;) as f:</em></p> <p><em>&nbsp; inputs = h5[&#39;x0&#39;][()]</em></p> <p><em>&nbsp; targets = h5[&#39;targs&#39;][()]</em></p> <p>The test file only contains the X0 inputs:</p> <p><em>with h5py.File(&#39;test.h5&#39;, &#39;r&#39;)&nbsp;as h5:</em></p> <p><em>&nbsp; inputs = h5[&#39;x0&#39;][()]</em></p> <p>The private testing sample also contains targets.&nbsp;The private and public splits can be recovered using:</p> <p><em>from sklearn.model_selection import train_test_split</em></p> <p><em>pub, pri = train_test_split(targets, test_size=25000, random_state=3452, shuffle=True)</em></p> <p>&nbsp;</p>