Mercurial > pylearn

comparison doc/v2_planning/requirements.txt @ 1096:2bbc294fa5ac

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
requirements: Added a use case
author Olivier Delalleau <delallea@iro>
date Mon, 13 Sep 2010 09:38:26 -0400
parents a65598681620
children 4eda3f52ebef
comparison
equal deleted inserted replaced
1095:520fcaa45692 1096:2bbc294fa5ac
40 ----- 40 -----
41 41
42 R1. reproduce previous work (our own and others') 42 R1. reproduce previous work (our own and others')
43 43
44 R2. explore MLA variants by swapping components (e.g. optimization algo, dataset, 44 R2. explore MLA variants by swapping components (e.g. optimization algo, dataset,
45 hyper-parameters). 45 hyper-parameters)
46 46
47 R3. analyze experimental results (e.g. plotting training curves, finding best 47 R3. analyze experimental results (e.g. plotting training curves, finding best
48 models, marginalizing across hyper-parameter choices) 48 models, marginalizing across hyper-parameter choices)
49 49
50 R4. disseminate (or serve as platform for disseminating) our own published algorithms 50 R4. disseminate (or serve as platform for disseminating) our own published algorithms
56 random search) 56 random search)
57 57
58 R7. provide implementations of standard pre-processing algorithms (e.g. PCA, 58 R7. provide implementations of standard pre-processing algorithms (e.g. PCA,
59 stemming, Mel-scale spectrograms, GIST features, etc.) 59 stemming, Mel-scale spectrograms, GIST features, etc.)
60 60
61 R8. provide high performance suitable for large-scale experiments, 61 R8. provide high performance suitable for large-scale experiments
62 62
63 R9. be able to use the most efficient algorithms in special case combinations of 63 R9. be able to use the most efficient algorithms in special case combinations of
64 learning algorithm components (e.g. when there is a fast k-fold validation 64 learning algorithm components (e.g. when there is a fast k-fold validation
65 algorithm for a particular model family, the library should not require users 65 algorithm for a particular model family, the library should not require users
66 to rewrite their standard k-fold validation script to use it) 66 to rewrite their standard k-fold validation script to use it)
67 67
68 R10. support experiments on a variety of datasets (e.g. movies, images, text, 68 R10. support experiments on a variety of datasets (e.g. movies, images, text,
69 sound, reinforcement learning?) 69 sound, reinforcement learning?)
70 70
71 R11. support efficient computations on datasets larger than RAM and GPU memory 71 R11. support efficient computations on datasets larger than RAM and GPU memory
72 72
73 R12. support infinite datasets (i.e. generated on the fly) 73 R12. support infinite datasets (i.e. generated on the fly)
74 74
75 75 R13. from a given evaluation experimental setup, be able to save a model that
76 can be used "in production" (e.g. say you try many combinations of
77 preprocessing, models and associated hyper-parameters, and want to easily be
78 able to recover the full "processing pipeline" that performs best, to be
79 used on future "real" test data)
76 80
77 Basic Design Approach 81 Basic Design Approach
78 ===================== 82 =====================
79 83
80 An ability to drive parallel computations is essential in addressing [R6,R8]. 84 An ability to drive parallel computations is essential in addressing [R6,R8].