Louis Tiao

Layer shape compatibility check

1

Add utility for checking compatibility of layer input / output shapes in stack operation

For building deeper and more complex models like Inception and ResNet, ops like [repeat](https://www.tensorflow.org/api_docs/python/tf/contrib/layers/repeat) makes it easy to replicate layers or even blocks of layers (e.g. ResNet blocks). See https://github.com/tensorflow/models/blob/master/slim/nets/inception_resnet_v2.py...

Look into: http://arxiv.org/pdf/1312.6055v3.pdf

Refactor Dockerfile dependency building steps to take advantage of Docker caching and eliminate the need for having a separate base image.

Implementation of Cholesky rank-1 update is orders of magnitude slower than the naive approach?

1

A common problem is to compute the Cholesky factor of `A + u @ u.T`, given a PD matrix `A` (shape `n x n`) and a rank-1 update vector `u`...

The Todo List

1

Very interested in getting involved. I have a few questions from looking at the todo list: - Consider adding capitalised phrase identification - Consider adding a repeated-char remover (e.g. grrrrrrr->grr)...

Refactor data-loading for LCBench surrogate benchmark problem

10

This diff improves the LCBench data-loading and benchmark problem setup. Key changes include: - Introducing an alternative constructor option for `LearningCurveBenchmarkTestFunction` in `get_lcbench_early_stopping_benchmark_problem` - Making data retrieval in `LearningCurveBenchmarkTestFunction` more...

Allow `map_keys_as_parameters` to be specified independently from `fit_only_completed_map_metrics` in Adaptor data-loading logic

1

Differential Revision: D71394207

Adds FCNet/HPOLib Tabular Benchmarks (AutoML Problem with Discrete Search Space)

2

Summary: TODO: Add tests, docstrings, instructions for mounting data to devservers from Manifold Differential Revision: D68461287

Temporary Commit at 2/7/2025, 6:32:26 PM

1

Differential Revision: D69335324