Roadmap

[MatteoRobbiati]

Next steps with priorities ( High priority $\rightarrow$ 🔴 🟠 🟢 $\leftarrow$ low priority ):

Core developing

• merge model branch into main (🔴);
• clean up the main from old/unused files (🔴);
• merge PSR checking gradients (🔴);
• speed up tests (🔴):
  • reduce the tested combinations (🔴);
  • setup a wise test procedure (🟠);
• setup the differentiation procedure for data $x$ in the differentiation rules;
• enable a single BackPropagation(DifferentiationRule) node in our differentiation pipeline (🟠) ?? - open discussion here;
• Keras interface (🟠);
• Jax backend improvements (🟠);
• Implement reuploading encoding and make the model more flexible to support trainable encoding (🟠);
  • implement a first working example (🟠);
  • define how the QuantumModel structure should get as argument a trainable layer/a quantum circuit, as pointed out in comment Enabling PSR, analytical diff with pytorch and moving interfaces to dedicated module #42 (comment) (🟠).
• Jax (Flax) interface (🟢);
• test GPUs work with Pytorch, Tensorflow and Jax (🟢);
• support batch execution (🟢);
• real time error mitigation (🟢);
• support a set of hardware-compatible optimizers (🟢). Have a look to this repo of gradient-free optimizers;
• Add a function to draw the QuantumModel (just like Circuit.draw()) (🟢);

Support tasks

• migrate the pytorch backend from Qibo to Qiboml (🔴);
• merge docs draft (🔴);
• docs (🔴):
  • webpage (🟠);
  • tutorial (🔴);
• move variational examples from Qibo to Qiboml with tests (using Qiboml API) (🟠);