deep-multipit/scripts/config_architecture/config_late_MSKCC.yaml at main · sysbio-curie/deep-multipit · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
name: MSKCC_latefus #model name (all results associated with this model will be saved in directories with this name)
MSKCC: True #whether the data to consider are similar to or the same as the one from Vanguri et al. (i.e., specific radiomic transform).

#Note: The function which builds the model from the configuration file is available in scripts/_utils.build_model

architecture:

  #whether to use intermediate or late fusion strategies
  intermediate_fusion: False

  #modalities to include in the model, in that specific order
  order: ['pdl1', 'pathomics', 'radiomics_PC', 'radiomics_LN', 'radiomics_PL', 'omics']

  #embedder for each modality (embedding models are available in dmultipit.model.embeddings)
  modality_embeddings:

    radiomics_PC:
      type: "ModalityEmbedding"
      args:
        dim_input:
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

    radiomics_PL:
      type: "ModalityEmbedding"
      args:
        dim_input:
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

    radiomics_LN:
      type: "ModalityEmbedding"
      args:
        dim_input:
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

    pathomics:
      type: "ModalityEmbedding"
      args:
        dim_input: 18
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

    omics:
      type: "ModalityEmbedding"
      args:
        dim_input: 11
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

    pdl1:
      type: "ModalityEmbedding"
      args:
        dim_input: 1
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

    clinicals:
      type: "ModalityEmbedding"
      args:
        dim_input: 13
        h_sizes: []
        p_dropout: 0
        dim_output: 1
        final_activation: "tanh"

  #attention model (attention models are available in dmultipit.model.attentions)
  attention:
    type: "MSKCCAttention"
    args:

  #predictor to obtain the final prediction from the multimodal embedding (not used for late fusion strategies !)
  predictor:
    type:
    args: