Add Readme

Author: THinnerichs

src/data/DeepCoder_2016/README.md

@@ -0,0 +1,21 @@
+# DeepCoder Benchmark
+
+The DeepCoder specializes in functional programs that manipulate lists. 
+Each problem is written as a set of input-output examples.
+
+The DeepCoder benchmark is derived from Balog et al. (2016) using the setup from Neo (Feng et al., 2018), as the evaluation benchmarks are not publicly available.
+Neo thus generated 100 benchmarks following this workflow:
+
+> We enumerate DSL programs with
+> at least 5 components and randomly generate inputs and the
+> corresponding output. This procedure is repeated for a fixed
+> number of times until we either obtain 5 valid input-output
+> examples or no examples have been found within the iter-
+> ation limit. In the latter case, we restart this process and
+> randomly search for a different program.
+
+See
+> Balog, M., Gaunt, A. L., Brockschmidt, M., Nowozin, S., & Tarlow, D. (2016). Deepcoder: Learning to write programs. arXiv preprint arXiv:1611.01989.
+and
+> Feng, Y., Martins, R., Bastani, O., & Dillig, I. (2018). Program synthesis using conflict-driven learning. ACM SIGPLAN Notices, 53(4), 420-435.
+