stumpy

stumpy copied to clipboard

I think there is a ~minor~ bug in `stumpy/snippet.py`. one can see the bug by setting param `s` to a value less than or equal to `m / 2`. To...

NimaSarajpoor

[WIP] Unsigned Integer Indexing Optimizations (#658)

14

Added my first attempts at optimizing `stump` using unsigned integer indexing. I added a temporarity `stumpy.stump_uint` module, which (if successfully optimized) may replace `stump`. But I haven't had too much...

alvii147

This is in reference to the discussion around the differences in using `np.int64` and `np.uint64` to index numpy errors in [this PR](https://github.com/TDAmeritrade/stumpy/pull/463) and [this numba discussion thread](https://numba.discourse.group/t/uint64-vs-int64-indexing-performance-difference/1500?u=seanlaw). Based on the...

alvii147

Improving Multi-processor Performance

3

This [new paper](https://www.sciencedirect.com/science/article/pii/S0743731524000194) titled "Exploring Multiprocessor Approaches to Time Series Analysis" claims to significantly improve the performance of matrix profile calculations. We should consider looking into this. Additionally, we should...

seanlaw

Incorrect MIN Python Version in README

11

The MIN Python version mentioned In `README.rst` is outdated (see the following line) ``` STUMPY supports [Python 3.7+](https://python3statement.org/) and, due to ... ``` Is it okay to just change it...

NimaSarajpoor

Speeding up the computation of sliding dot product with fft

90

[Update] Here, I am providing some important links to external resources or the comments mentioned here: * So far: [6-8step-FFT-based MASS, with Numba (SingleThreaded)](https://gist.github.com/NimaSarajpoor/077c91a07858039b66cdfa0c98d2f4b4) * [A comment regarding: OFFT uses...

NimaSarajpoor

Computing matrix profile while considering a custom transformation for subsequences

10

Currently, STUMPY supports the parameter [normalize](https://github.com/TDAmeritrade/stumpy/blob/3559b38c9a45bf450359dab40eb081d0b87687a4/stumpy/stump.py#L547-L550) which allows users to compute matrix profile for the following cases: (1) `normalize == False`: Compute the distance between subsequences with no transformation (2)...

NimaSarajpoor

[This paper](https://www.merl.com/publications/docs/TR2022-040.pdf) by MERL discusses the use of double ended queues to speed up computation of matrix profiles but it isn't clear how this may/may not affect top-k nearest neighbors....

seanlaw

Tutorial notebook DAMP

27

This PR is a replacement for PR #872. This PR is to address #606. --- I will start with implementing the original algorithm, as proposed in the paper, as close...

NimaSarajpoor

Fixed#938: Speed up njit-decorated function for sliding dot product [WIP]

3

This PR addresses the issue raised in #938. We want to create a new function for `sliding_dot_product` that: (1) has a performance that is close to the performance of `core.sliding_dot_product`...

NimaSarajpoor