module hackathon.perf2018

Inheritance diagram of ensae_projects.hackathon.perf2018

Short summary

module ensae_projects.hackathon.perf2018

Compute the performance for the hackathon 2018.

source on GitHub

Classes

class truncated documentation
MLStoragePerf2018 Computes the performances the a hackathon.
MLStoragePerf2018Image Overloads compute_perf for images. Example of use:
MLStoragePerf2018TimeSeries Overloads compute_perf for timeseries. Example of use:

Methods

method truncated documentation
__init__  
__init__  
__init__  
_label_mapping Computes the label based on a subfolder name.
_label_mapping Computes the label based on a subfolder name.
_load_cached_performance Retrieves performances already computed.
_load_cached_performance Retrieves performances already computed.
_load_cached_performance Retrieves performances already computed.
_load_ml_storage Creates an instance of a MLStorage
_load_ml_storage Creates an instance of a MLStorage
_load_ml_storage Creates an instance of a MLStorage
_save_performance Saves cached performance.
_save_performance Saves cached performance.
_save_performance Saves cached performance.
compute_perf Computes the performances for every image and one particular model.
compute_perf Computes the performances for every image and one particular model.
compute_perf Computes the performances for every image and one particular model.
compute_performance Computes the performance for the not cached one if use_cache is True.
compute_performance Computes the performance for the not cached one if use_cache is True.
compute_performance Computes the performance for the not cached one if use_cache is True.

Documentation

Compute the performance for the hackathon 2018.

source on GitHub

class ensae_projects.hackathon.perf2018.MLStoragePerf2018(storage, examples, cache_file='cache_file.csv')[source]

Bases: object

Computes the performances the a hackathon.

source on GitHub

Parameters:
  • storage – storage location
  • examples – deep learning models

source on GitHub

__init__(storage, examples, cache_file='cache_file.csv')[source]
Parameters:
  • storage – storage location
  • examples – deep learning models

source on GitHub

_load_cached_performance(cache_file=None)[source]

Retrieves performances already computed.

Parameters:cached_file – file

source on GitHub

_load_ml_storage(root)[source]

Creates an instance of a MLStorage based on a folder.

Parameters:root – folder

source on GitHub

_save_performance(df, cache_file=None)[source]

Saves cached performance.

Parameters:
  • df – dataframe
  • cache_file – destination

source on GitHub

compute_perf(name)[source]

Computes the performances for every image and one particular model.

source on GitHub

compute_performance(use_cache=True, fLOG=None)[source]

Computes the performance for the not cached one if use_cache is True.

Parameters:
  • use_cache – use cache
  • fLOG – logging function
Returns:

dataframe

source on GitHub

class ensae_projects.hackathon.perf2018.MLStoragePerf2018Image(storage, examples, cache_file='cache_file.csv')[source]

Bases: ensae_projects.hackathon.perf2018.MLStoragePerf2018

Overloads compute_perf for images. Example of use:

from ensae_projects.hackathon.perf2018 import MLStoragePerf2018Image
mstorage = "storage_brgm"
mexample = "hackathon_test/sample_labelled_test"
mpref = MLStoragePerf2018Image(mstorage, mexample)
mres = mpref.compute_performance(fLOG=print, use_cache=True)
mres = mres.sort_values("precision", ascending=False)
print(mres)
mbody = "<html><body><h1>Hackathon EY-ENSAE 2018 - BRGM</h1>

mcontent = “{0}{1}</body></html>”.format(mbody, mres.to_html()) with open(“brgm.html”, “w”, encoding=”utf-8”) as f:

f.write(mcontent)

source on GitHub

Parameters:
  • storage – storage location
  • examples – deep learning models

source on GitHub

__init__(storage, examples, cache_file='cache_file.csv')[source]
Parameters:
  • storage – storage location
  • examples – deep learning models

source on GitHub

_label_mapping(subs)[source]

Computes the label based on a subfolder name.

source on GitHub

compute_perf(name)[source]

Computes the performances for every image and one particular model.

source on GitHub

class ensae_projects.hackathon.perf2018.MLStoragePerf2018TimeSeries(storage, examples, cache_file='cache_file.csv')[source]

Bases: ensae_projects.hackathon.perf2018.MLStoragePerf2018

Overloads compute_perf for timeseries.

Example of use:

from ensae_projects.hackathon.perf2018 import MLStoragePerf2018TimeSeries
mstorage = "storage_microdon"
mexample = "hackathon_test/sample_labelled_test"
mpref = MLStoragePerf2018TimeSeries(mstorage, mexample)
mres = mpref.compute_performance(fLOG=print, use_cache=True)
mres = mres.sort_values("cor", ascending=False)
print(mres)
mbody = "<html><body><h1>Hackathon EY-ENSAE 2018 - Microdon</h1>

mcontent = “{0}{1}</body></html>”.format(mbody, mres.to_html()) with open(“brgm.html”, “w”, encoding=”utf-8”) as f:

f.write(mcontent)

source on GitHub

Parameters:
  • storage – storage location
  • examples – deep learning models

source on GitHub

__init__(storage, examples, cache_file='cache_file.csv')[source]
Parameters:
  • storage – storage location
  • examples – deep learning models

source on GitHub

_label_mapping(subs)[source]

Computes the label based on a subfolder name.

source on GitHub

compute_perf(name)[source]

Computes the performances for every image and one particular model.

source on GitHub