def _predict_single(net, driver_id, options): """ Make predictions for given driver. Assumes output-2 of the net is the positive class """ x_var = T.ftensor3('x_var') # symbolic var net_output = get_output(net, x_var, deterministic=True) # dropout inactive output_fn = theano_function(inputs=[x_var], outputs=net_output) driver_data = load_driver(join(options.datapath, str(driver_id)), options, make_train_val=False) probs = output_fn(driver_data['x']) output = sorted(np.vstack((np.asarray(driver_data['trip_ids']), probs[:, 1])).transpose().tolist(), key=lambda x: x[0]) output = [[str(driver_id) + '_' + str(int(trip_id)), str(p)] for trip_id, p in output] return output
def logp_forw(out_vars, vars, shared): """Compile Theano function of the model and the input and output variables. Parameters ---------- out_vars: List containing :class:`pymc3.Distribution` for the output variables vars: List containing :class:`pymc3.Distribution` for the input variables shared: List containing :class:`theano.tensor.Tensor` for depended shared data """ out_list, inarray0 = join_nonshared_inputs(out_vars, vars, shared) f = theano_function([inarray0], out_list[0]) f.trust_input = True return f