LICHENSTEIN HENRI (US)
PARRY BRADLEY (US)
ROTHBERG JONATHAN (US)
XU TIAN (US)
US20160328253A1 | 2016-11-10 | |||
US20140046683A1 | 2014-02-13 | |||
US20040076984A1 | 2004-04-22 | |||
US20160217395A1 | 2016-07-28 |
CLAIMS 1. A method of predicting an association among input data, comprising: mapping, by at least one processor, the input data to at least one space; calculating, by the at least one processor, an energy metric based on a distance, in the at least one space, separating members of a data pair of the input data; and predicting, by the at least one processor based on the energy metric, that the members of the data pair are associated with one another. 2. The method of claim 1, wherein calculating the energy metric includes calculating an exponential term. 3. The method of any of claims 1-2, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 4. The method of any of claims 1-3, wherein mapping the input data to the at least one space includes grouping vector representations of the input data into a plurality of clusters in the at least one space. 5. The method of any of claims 1-4, wherein the at least one space includes a hyper- geometric space. 6. The method of claim 5, wherein the hyper-geometric space includes a surface of a hypersphere. 7. The method of any of claims 5-6, wherein calculating the energy metric includes using a hyper-parameter of the hyper-geometric space. 8. The method of any of claims 1-7, wherein predicting that the members of the data pair are associated with one another includes determining, based on the energy metric, a likelihood that the members of the data pair are associated with one another. 9. The method of any of claims 1-8, further comprising filtering out, by the at least one processor, portions of the input data prior to mapping the input data to the at least one space. 10. The method of claim 9, wherein filtering out the portions of the input data includes removing portions of the input data having less than a threshold level of correlation. 11. The method of any of claims 1-10, wherein the members of the data pair are of a same data domain. 12. The method of claim 11, wherein a first member of the data pair is of a first data modality within the same data domain, and a second member of the data pair is of a second data modality within the same data domain. 13. The method of claim 12, wherein mapping the input data to the at least one space includes: mapping the first member of the data pair to a first modality space; and mapping the second member of the data pair to a second modality space. 14. The method of any of claims 12-13, wherein the same data domain is compounds, and the first and second data modalities are selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 15. The method of any of claims 12-13, wherein the same data domain is diseases, and the first and second data modalities are selected from a group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 16. The method of any of claims 1-10, wherein first and second members of the data pair are of respective first and second domains. 17. The method of claim 16, wherein mapping the input data to the at least one space includes: mapping the first member of the data pair to a first domain space; and mapping the second member of the data pair to a second domain space. 18. The method of claim 17, wherein mapping the input data to the at least one space further includes: mapping the first member of the data pair to a first modality space of the first domain; and mapping the second member of the data pair to a first modality space of the second domain. 19. The method of any of claims 16-18, wherein the first and second domains are compounds and diseases, respectively. 20. The method of any of claims 16-18, wherein the first and second domains are financial market conditions and financial market events, respectively. 21. The method of any of claims 16-18, wherein the first and second domains are computer network conditions and computer network security events, respectively. 22. The method of any of claims 16-18, wherein the first and second domains are vehicle traffic conditions and vehicle accident events, respectively. 23. The method of any of claims 16-18, wherein the first and second domains are real images and/or videos and fake images and/or videos, respectively. 24. The method of any of claims 16-18, wherein the first and second domains are music and/or movie preferences and music and/or movies, respectively. 25. The method of any of claims 16-18, wherein the first and second domains are social media trends and political events, respectively. 26. The method of any of claims 16-18, wherein the first and second domains are environmental phenomena and/or human activity and natural disaster events, respectively. 27. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 1-26. 28. A method comprising: mapping, by at least one processor, input data to at least one hyper-geometric space; and determining, by the at least one processor, an association among the input data based on the mapping. 29. The method of claim 28, wherein the at least one hyper-geometric space includes a surface of a hypersphere. 30. The method of claim 28, wherein the at least one hyper-geometric space includes a surface of a polytope. 31. The method of claim 28, wherein the at least one hyper-geometric space includes a surface of a hyper-cube. 32. The method of any of claims 28-31, further comprising calculating, by the at least one processor, an energy metric relating to a distance separating a pair of the input data in the at least one hyper-geometric space. 33. The method of claim 32, wherein calculating the energy metric includes calculating an exponential term. 34. The method of claim 32, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 35. The method of any of claims 33-34, wherein the energy metric further relates to a hyper-parameter of the at least one hyper-geometric space. 36. The method of any of claims 28-35, wherein mapping the input data to the at least one hyper-geometric space includes grouping vector representations of the input data into a plurality of clusters in the at least one space. 37. The method of any of claims 29-36, wherein determining the association among the input data includes determining, based on the energy metric, a likelihood that members of a data pair are associated with one another. 38. The method of any of claims 28-37, further comprising filtering out, by the at least one processor, portions of the input data prior to mapping the input data to the at least one hyper-geometric space. 39. The method of claim 38, wherein filtering out the portions of the input data includes removing portions of the input data having less than a threshold level of correlation. 40. The method of any of claims 37-39, wherein the members of the data pair are of a same data domain. 41. The method of claim 40, wherein a first member of the data pair is of a first data modality within the same data domain, and a second member of the data pair is of a second data modality within the same data domain. 42. The method of claim 41, wherein mapping the input data to the at least one hyper- geometric space includes: mapping the first member of the data pair to a first modality hyper-geometric space; and mapping the second member of the data pair to a second modality hyper-geometric space. 43. The method of any of claims 41-42, wherein the same data domain is compounds, and the first and second data modalities are selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 44. The method of any of claims 41-42, wherein the same data domain is diseases, and the first and second data modalities are selected from a group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 45. The method of any of claims 37-39, wherein first and second members of the data pair are of respective first and second domains. 46. The method of claim 45, wherein mapping the input data to the at least one hyper- geometric space includes: mapping the first member of the data pair to a first domain hyper-geometric space; and mapping the second member of the data pair to a second domain hyper-geometric space. 47. The method of any of claims 45-46, wherein mapping the input data to the at least one hyper-geometric space further includes: mapping the first member of the data pair to a first modality hyper-geometric space of the first domain; and mapping the second member of the data pair to a first modality hyper-geometric space of the second domain. 48. The method of any of claims 45-47, wherein the first and second domains are compounds and diseases, respectively. 49. The method of any of claims 45-47, wherein the first and second domains are financial market conditions and financial market events, respectively. 50. The method of any of claims 45-47, wherein the first and second domains are computer network conditions and computer network security events, respectively. 51. The method of any of claims 45-47, wherein the first and second domains are vehicle traffic conditions and vehicle accident events, respectively. 52. The method of any of claims 45-47, wherein the first and second domains are real images and/or videos and fake images and/or videos, respectively. 53. The method of any of claims 45-47, wherein the first and second domains are music and/or movie preferences and music and/or movies, respectively. 54. The method of any of claims 45-47, wherein the first and second domains are social media trends and political events, respectively. 55. The method of any of claims 45-47, wherein the first and second domains are environmental phenomena and/or human activity and natural disaster events, respectively. 56. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 28-55. 57. A method of training an encoder, comprising: mapping, by at least one processor, labeled input data to at least one space; calculating, by the at least one processor, an energy metric relating to the mapping; and adjusting, by the at least one processor, weights and/or biases of the encoder based on the energy metric. 58. The method of claim 57, further comprising: calculating, by the at least one processor, a loss function using the energy metric, wherein adjusting the weights and/or biases of the encoder includes adjusting the weights and/or biases based on the loss function. 59. The method of any of claims 57-58, wherein calculating the energy metric includes using a distance, in the at least one space, separating a data pair of the input data. 60. The method of claim 59, wherein calculating the energy metric includes calculating an exponential term that includes the distance. 61. The method of claim 59, wherein calculating the energy metric includes calculating a logarithmic term that includes the distance. 62. The method of claim 59, wherein calculating the energy metric includes calculating a sigmoidal term that includes the distance. 63. The method of claim 59, wherein calculating the energy metric includes calculating a continuous piecewise linear term that includes the distance. 64. The method of any of claims 59-66, wherein: an impact of the energy metric on the weights and/or biases of the encoder increases as the distance increases; and the data pair share a common characteristic. 65. The method of claim 64, wherein: the energy metric further relates to a second distance, in the at least one space, separating a second data pair of the input data; the impact of the energy metric on the weights and/or biases of the encoder increases as the distance decreases; and the second data pair do not have the common characteristic. 66. The method of any of claims 64-65, further comprising labeling, by the at least one processor, input data to generate the labeled input data, wherein labeling the input data includes emphasizing the common characteristic. 67. The method of claim 66, wherein emphasizing the common characteristic includes de- emphasizing a second common characteristic. 68. The method of any of claims 57-67, wherein the at least one space includes a hyper- geometric space. 69. The method of claim 68, wherein the hyper-geometric space includes a surface of a hypersphere. 70. The method of any of claims 68-69, wherein the energy metric includes a hyper- parameter of the hyper-geometric space. 71. The method of any of claims 66-70, further comprising filtering out, by the at least one processor, portions of the input data prior to labeling the input data. 72. The method of claim 71, wherein filtering out the portions of the input data includes removing portions of the input data having less than a threshold level of correlation. 73. The method of any of claims 59-72, wherein the members of the data pair are of a same data domain. 74. The method of claim 73, wherein a first member of the data pair is of a first data modality within the same data domain, and a second member of the data pair is of a second data modality within the same data domain. 75. The method of claim 74, wherein mapping the labeled input data to the at least one space includes: mapping the first member of the data pair to a first modality space; and mapping the second member of the data pair to a second modality space. 76. The method of any of claims 73-75, wherein the same data domain is compounds, and the first and second data modalities are selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 77. The method of any of claims 73-75, wherein the same data domain is diseases, and the first and second data modalities are selected from a group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 78. The method of any of claims 59-72, wherein first and second members of the data pair are of respective first and second domains. 79. The method of claim 78, wherein mapping the labeled input data to the at least one space includes: mapping the first member of the data pair to a first domain space; and mapping the second member of the data pair to a second domain space. 80. The method of any of claims 78-79, wherein mapping the labeled input data to the at least one space further includes: mapping the first member of the data pair to a first modality space of the first domain; and mapping the second member of the data pair to a first modality space of the second domain. 81. The method of any of claims 78-80, wherein the first and second domains are compounds and diseases, respectively. 82. The method of any of claims 78-80, wherein the first and second domains are financial market conditions and financial market events, respectively. 83. The method of any of claims 78-80, wherein the first and second domains are computer network conditions and computer network security events, respectively. 84. The method of any of claims 78-80, wherein the first and second domains are vehicle traffic conditions and vehicle accident events, respectively. 85. The method of any of claims 78-80, wherein the first and second domains are real images and/or videos and fake images and/or videos, respectively. 86. The method of any of claims 78-80, wherein the first and second domains are music and/or movie preferences and music and/or movies, respectively. 87. The method of any of claims 78-80, wherein the first and second domains are social media trends and political events, respectively. 88. The method of any of claims 78-80, wherein the first and second domains are environmental phenomena and/or human activity and natural disaster events, respectively. 89. The method of any of claims 59-88, further comprising: initializing the encoder, by the at least one processor, at least in part by: mapping, to the at least one space, the labeled input data; calculating an initial loss function prior to calculating the energy metric; and adjusting the weights and/or biases based on the initial loss function. 90. The method of claim 89, wherein calculating the initial loss function includes using an initial distance separating the data pair. 91. The method of any of claims 89-90, further comprising uniformly distributing, by the at least one processor, the labeled input data in the at least one space after adjusting the weights and/or biases based on the initial loss function. 92. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 57-91. 93. A method comprising comparing, by at least one processor, disease data with compound data to identify a treatment compound. 94. The method of claim 93, wherein identifying the treatment compound includes predicting, by the at least one processor based on an energy metric, that the treatment compound is effective at treating a disease related to the disease data. 95. The method of claim 94, further comprising calculating, by the at least one processor, the energy metric using a distance separating the compound data from the disease data in a common space. 96. The method of claim 95, wherein calculating the energy metric includes calculating an exponential term. 97. The method of claim 95, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 98. The method of any of claims 95-97, further comprising encoding, by at least one trained encoder executed by the at least one processor, the compound data and the disease data to generate vector representations of the compound data and the disease data in the common space. 99. The method of claim 98, further including filtering out, by the at least one processor, portions of the compound data and the disease data prior to encoding. 100. The method of any of claims 98-99, further comprising grouping, by the at least one trained encoder in the common space, groups of the vector representations according to similarities and differences in biological characteristics of the compound data and the disease data. 101. The method of any of claims 95-100, wherein the common space includes a hyper- geometric space. 102. The method of claim 101, wherein the energy metric includes a hyper-parameter of the hyper-geometric space. 103. The method of any of claims 101-102, wherein the hyper-geometric space includes a surface of a hypersphere. 104. The method of any of claims 98-103, wherein the compound data includes at least one member selected from a compound data group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 105. The method of claim 104, further comprising: encoding, by a first trained encoder of the at least one encoder to a first space, a first member of the compound data group; encoding, by a second trained encoder of the at least one encoder to a second space, a second member of the compound data group; and encoding, by at least one third trained encoder of the at least one encoder to the common space, encoded versions of the first and second members of the compound data group. 106. The method of any of claims 98-103, wherein the disease data includes at least one member selected from a disease data group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 107. The method of claim 106, further comprising: encoding, by a first trained encoder of the at least one trained encoder to a first space, a first member of the disease data group; encoding, by a second trained encoder of the at least one trained encoder to a second space, a second member of the disease data group; and encoding, by at least one third trained encoder of the at least one trained encoder to the common space, encoded versions of the first and second members of the disease data group. 108. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 93-107. 109. A method comprising matching, by at least one processor, disease data with compound data to predict a treatment association. 110. The method of claim 109, further including predicting, by the at least one processor based on an energy metric, that a compound is effective at treating a disease, wherein the treatment association includes a portion of the compound data related to the compound and a portion of the disease data related to the disease. 111. The method of claim 110, further comprising calculating, by the at least one processor, the energy metric using a distance separating the compound data from the disease data in a common space. 112. The method of claim 111, wherein calculating the energy metric includes calculating an exponential term. 113. The method of claim 111, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 114. The method of any of claims 111-113, further comprising encoding, by at least one trained encoder executed by the at least one processor, the compound data and the disease data to generate vector representations of the compound data and the disease data in the common space. 115. The method of claim 114, further including filtering out, by the at least one processor, portions of the compound data and the disease data prior to encoding. 116. The method of claim 114-115, further comprising grouping, by the at least one trained encoder in the common space, the vector representations according to similarities and differences in biological characteristics of the compound data and the disease data. 117. The method of any of claims 111-116, wherein the common space includes a hyper- geometric space. 118. The method of claim 117, wherein the hyper-geometric space includes a surface of a hypersphere. 119. The method of any of claims 117-118, wherein the energy metric includes a hyper- parameter of the hyper-geometric space. 120. The method of any of claims 114-119, wherein the compound data includes at least one member selected from a compound data group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 121. The method of claim 120, further comprising: encoding, by a first trained encoder of the at least one encoder to a first space, a first member of the compound data group; encoding, by a second trained encoder of the at least one encoder to a second space, a second member of the compound data group; and encoding, by at least one third trained encoder of the at least one encoder to the common space, encoded versions of the first and second members of the compound data group. 122. The method of any of claims 114-119, wherein the disease data includes at least one member selected from a disease data group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 123. The method of claim 122, further comprising: encoding, by a first trained encoder of the at least one trained encoder to a first space, a first member of the disease data group; encoding, by a second trained encoder of the at least one trained encoder to a second space, a second member of the disease data group; and encoding, by at least one third trained encoder of the at least one trained encoder to the common space, encoded versions of the first and second members of the disease data group. 124. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 109-123. 125. A method comprising identifying, by at least one processor, a relationship between a disease of a patient and a compound at least in part by matching disease data associated with the disease to compound data associated with the compound. 126. The method of claim 125, wherein identifying the relationship includes predicting, based on an energy metric, that the compound is effective at treating the disease. 127. The method of claim 126, further comprising calculating, by the at least one processor, the energy metric using a distance separating the compound data from the disease data in a common space. 128. The method of claim 127, wherein calculating the energy metric includes calculating an exponential term. 129. The method of claim 127, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 130. The method of any of claims 127-129, further comprising encoding, by at least one trained encoder executed by the at least one processor, the compound data and the disease data to generate vector representations of the compound data and the disease data in the common space. 131. The method of claim 130, further comprising filtering out, by the at least one processor, portions of the compound data and the disease data prior to encoding. 132. The method of any of claims 130-131, further comprising grouping, by the at least one trained encoder in the common space, the vector representations according to similarities and differences in gene expression characteristics of the compound data and the disease data. 133. The method of any of claims 127-132, wherein the common space includes a hyper- geometric space. 134. The method of claim 133, wherein the hyper-geometric space includes a surface of a hypersphere. 135. The method of any of claims 133-134, wherein the energy metric includes a hyper- parameter of the hyper-geometric space. 136. The method of any of claims 130-135, wherein the compound data includes at least one member selected from a compound data group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 137. The method of claim 136, further comprising: encoding, by a first trained encoder of the at least one encoder to a first space, a first member of the compound data group; encoding, by a second trained encoder of the at least one encoder to a second space, a second member of the compound data group; and encoding, by at least one third trained encoder of the at least one encoder to the common space, encoded versions of the first and second members of the compound data group. 138. The method of any of claims 130-135, wherein the disease data includes at least one member selected from a disease data group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 139. The method of claim 138, further comprising: encoding, by a first trained encoder of the at least one trained encoder to a first space, a first member of the disease data group; encoding, by a second trained encoder of the at least one trained encoder to a second space, a second member of the disease data group; and encoding, by at least one third trained encoder of the at least one trained encoder to the common space, encoded versions of the first and second members of the disease data group. 140. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 125-139. 141. A method comprising predicting, by at least one processor, a compound-disease association based on a grouping of an encoded compound profile with an encoded disease profile in a common space. 142. The method of claim 141, wherein predicting the compound-disease association includes predicting, based on an energy metric related to the grouping, that a compound related to the encoded compound profile is effective at treating a disease related to the encoded disease profile. 143. The method of claim 142, further comprising calculating, by the at least one processor, the energy metric using a distance separating the encoded compound profile from the encoded disease profile in the common space. 144. The method of claim 143, wherein calculating the energy metric includes using distances separating other encoded compound profiles from other encoded disease profiles in the common space. 145. The method of any of claims 143-144, wherein calculating the energy metric includes calculating an exponential term. 146. The method of any of claims 143-144, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 147. The method of any of claims 143-146, further comprising encoding, by at least one trained encoder executed by the at least one processor, compound data and disease data to generate the encoded compound profile and the encoded disease profile in the common space. 148. The method of claim 147, further comprising filtering out, by the at least one processor, portions of the compound data and the disease data prior to encoding. 149. The method of any of claims 147-148, further comprising: grouping, by the at least one trained encoder in the common space, encoded compound profiles and encoded disease profiles according to similarities and differences in biological characteristics of the encoded compound profiles and the encoded disease profiles, wherein the grouping includes pairing the encoded compound profile with the encoded disease profile. 150. The method of any of claims 143-149, wherein the common space includes a hyper- geometric space. 151. The method of claim 150, wherein the hyper-geometric space includes a surface of a hypersphere. 152. The method of any of claims 150-151, wherein the energy metric includes a hyper- parameter of the hyper-geometric space. 153. The method of any of claims 147-152, wherein the compound data includes at least one member selected from a compound data group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 154. The method of claim 153, further comprising: encoding, by a first trained encoder of the at least one encoder to a first space, a first member of the compound data group; encoding, by a second trained encoder of the at least one encoder to a second space, a second member of the compound data group; and encoding, by at least one third trained encoder of the at least one encoder to the common space, encoded versions of the first and second members of the compound data group. 155. The method of any of claims 147-152, wherein the disease data includes at least one member selected from a disease data group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 156. The method of claim 155, further comprising: encoding, by a first trained encoder of the at least one trained encoder to a first space, a first member of the disease data group; encoding, by a second trained encoder of the at least one trained encoder to a second space, a second member of the disease data group; and encoding, by at least one third trained encoder of the at least one trained encoder to the common space, encoded versions of the first and second members of the disease data group. 157. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 141-156. 158. A method, comprising: conditioning, by at least one processor, input data to address noise present in the input data; and predicting, by the at least one processor, a compound-disease association based on the input data. 159. The method of claim 158, wherein conditioning the input data includes filtering out portions of the input data having a level of correlation with one another that is below a threshold level. 160. The method of claim 159, wherein: the compound-disease association includes a compound and a disease; and predicting the compound-disease association includes predicting, based on an energy metric, that the compound is effective at treating the disease. 161. The method of claim 160, further comprising calculating, by the at least one processor, the energy metric using a distance, in a common space, separating a compound profile related to the compound from a disease profile related to the disease. 162. The method of claim 161, wherein calculating the energy metric further includes using distances separating other compound profiles from other disease profiles in the common space. 163. The method of any of claims 161-162, wherein calculating the energy metric includes calculating an exponential term. 164. The method of any of claims 161-162, wherein calculating the energy metric includes calculating at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 165. The method of any of claims 161-164, further comprising encoding, by at least one trained encoder executed by the at least one processor, compound data and disease data to generate vector representations of the compound profile and the disease profile in the common space. 166. The method of claim 165, wherein filtering out the portions of the input data occurs prior to encoding. 167. The method of any of claims 165-166, further comprising grouping, by the at least one trained encoder in the common space, the vector representations according to similarities and differences in biological characteristics of the compound profiles and the disease profiles. 168. The method of any of claims 161-167, wherein the common space includes a hyper- geometric space. 169. The method of claim 168, wherein the hyper-geometric space includes a surface of a hypersphere. 170. The method of any of claims 168-169, wherein the energy metric includes a hyper- parameter of the hyper-geometric space. 171. The method of any of claims 165-170, wherein the compound data includes at least one member selected from a compound data group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 172. The method of claim 171, further comprising: encoding, by a first trained encoder of the at least one encoder to a first space, a first member of the compound data group; encoding, by a second trained encoder of the at least one encoder to a second space, a second member of the compound data group; and encoding, by at least one third trained encoder of the at least one encoder to the common space, encoded versions of the first and second members of the compound data group. 173. The method of any of claims 165-170, wherein the disease data includes at least one member selected from a disease data group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 174. The method of claim 173, further comprising: encoding, by a first trained encoder of the at least one trained encoder to a first space, a first member of the disease data group; encoding, by a second trained encoder of the at least one trained encoder to a second space, a second member of the disease data group; and encoding, by at least one third trained encoder of the at least one trained encoder to the common space, encoded versions of the first and second members of the disease data group. 175. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 158-174. 176. A method comprising comparing, by at least one processor, compound data with disease data, using multiple modalities of the compound data and the disease data, to determine a treatment association among the compound and disease data. 177. The method of claim 176, wherein, for the compound data, the multiple modalities include at least member selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and/or compound side effect data. 178. The method of claim 176, wherein, for the disease data, the multiple modalities include at least one member selected from a group consisting of: disease gene expression data; disease symptom data; disease biological pathway data; and/or disease proteomic data; 179. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 176-178. 180. A method comprising comparing, by at least one processor, biological data within a domain, using multiple modalities of the biological data, to determine an association among the biological data. 181. The method of claim 180, wherein the domain is compound data and the multiple modalities include at least one member selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and/or compound side effect data. 182. The method of claim 180, wherein the domain is disease data and the multiple modalities include at least one member selected from a group consisting of: disease gene expression data; disease symptom data; disease biological pathway data; and/or disease proteomic data; 183. A non-transitory computer-readable medium having encoded thereon instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of claims 180-182. 184. A method of predicting a combination therapy treatment, comprising: mapping, by at least one processor, compound and disease data to a common space; associating, by the at least one processor, a grouping of compound data and disease data in the common space; and associating, by the at least one processor, a sub-grouping of the compound data in a compound space. 185. The method of claim 184, wherein associating the grouping of compound and disease data is based on a common energy metric. 186. The method of claim 185, further comprising calculating, by the at least one processor, the common space energy metric using common space distances separating the grouping of compound data and disease data. 187. The method of claim 186, wherein calculating the common space energy metric includes calculating at least one common space exponential function that includes at least one of the common space distances. 188. The method of any of claims 184-187, wherein associating the sub-grouping of the compound data is based on a compound space energy metric. 189. The method of claim 188, further comprising calculating, by the at least one processor, the compound space energy metric using compound space distances separating the sub-grouping of compound data. 190. The method of claim 189, wherein calculating the compound space energy metric includes calculating at least one compound space exponential function that includes at least one of the compound space distances. 191. A method of identifying a biomarker, comprising: mapping, by at least one processor to at least one space, first biological data extracted from a patient; associating, by the at least one processor based on a grouping in the at least one space, the first biological data with second biological data related to patients having responded to a treatment; and predicting, by the at least one processor, that the patient will respond to the treatment. 192. The method of claim 191, further comprising calculating, by at least one processor, an energy metric, wherein predicting that the patient will respond to the treatment is based on the energy metric. 193. The method of claim 192, wherein calculating the energy metric includes using a distance, in the at least one space, separating the first biological data from the second biological data. 194. The method of any of claims 191-193, wherein the first and second biological data include disease data. 195. The method of any of claims 191-193, wherein the first biological data includes disease data and the second biological data includes compound data. 196. A system for predicting an association among input data, comprising: at least one trained encoder configured to: map the input data to at least one space; and calculate an energy metric relating to the map; and at least one decoder configured to output a prediction, generated using the energy metric, associating members of a data pair of the input data. 197. The system of claim 196, further comprising at least one processor configured to execute the at least one trained encoder. 198. The system of claim 197, wherein the at least one processor is configured to generate the energy metric using a distance, in the at least one space, separating the members of the data pair. 199. The system of any of claims 196-198, wherein the energy metric includes an exponential term. 200. The system of any of claims 196-198, wherein the energy metric includes at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 201. The system of any of claims 197-200, wherein the at least one trained encoder is configured to group vector representations of the input data into a plurality of clusters in the at least one space. 202. The system of any of claims 197-201, wherein the at least one space includes a hyper- geometric space. 203. The system of claim 202, wherein the hyper-geometric space includes a surface of a hypersphere. 204. The system of any of claims 202-203, wherein the at least one processor is further configured to calculate the energy metric using a hyper-parameter of the hyper-geometric space. 205. The system of any of claims 197-204, wherein the at least one processor is configured to predict that the members of the data pair are associated with one another at least in part by determining, using the energy metric, a likelihood that the members of the data pair are associated with one another. 206. The system of any of claims 197-205, wherein the at least one processor is further configured to filter out portions of the input data prior to mapping the input data to the at least one space. 207. The system of claim 206, wherein the at least one processor is further configured to remove portions of the input data having less than a threshold level of correlation. 208. The system of any of claims 197-207, wherein the members of the data pair are of a same data domain. 209. The system of claim 208, wherein a first member of the data pair is of a first data modality within the same data domain, and a second member of the data pair is of a second data modality within the same data domain. 210. The system of claim 209, wherein the at least one trained encoder includes: a first modality encoder configured to map the first member of the data pair to a first modality space; and a second modality encoder configured to map the second member of the data pair to a second modality space. 211. The system of any of claims 209-210, wherein the same data domain is compounds, and the first and second data modalities are selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 212. The system of any of claims 209-210, wherein the same data domain is diseases, and the first and second data modalities are selected from a group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 213. The system of any of claims 197-207, wherein first and second members of the data pair are of respective first and second domains. 214. The system of claim 213, wherein the at least one trained encoder includes: at least one first domain encoder configured to map the first member of the data pair to a first domain space; and at least one second domain encoder configured to map the second member of the data pair to a second domain space. 215. The system of claim 214, wherein: the at least one first domain encoder includes a first modality encoder configured to map the first member of the data pair to a first modality space of the first domain; and the at least one second domain encoder includes a second modality encoder configured to map the second member of the data pair to a second modality space of the second domain. 216. The system of any of claims 213-215, wherein the first and second domains are compounds and diseases, respectively. 217. The system of any of claims 213-215, wherein the first and second domains are financial market conditions and financial market events, respectively. 218. The system of any of claims 213-215, wherein the first and second domains are computer network conditions and computer network security events, respectively. 219. The system of any of claims 213-215, wherein the first and second domains are vehicle traffic conditions and vehicle accident events, respectively. 220. The system of any of claims 213-215, wherein the first and second domains are real images and/or videos and fake images and/or videos, respectively. 221. The system of any of claims 213-215, wherein the first and second domains are music and/or movie preferences and music and/or movies, respectively. 222. The system of any of claims 213-215, wherein the first and second domains are social media trends and political events, respectively. 223. The system of any of claims 213-215, wherein the first and second domains are environmental phenomena and/or human activity and natural disaster events, respectively. 224. The system of any of claims 197-223, further comprising a user interface component coupled to the at least one processor, wherein the user interface component is configured to receive at least a first portion of the input data from a user. 225. The system of claim 224, wherein the user interface component includes at least one member selected from a group consisting of: a mouse; a keyboard; a touchscreen; and a microphone. 226. The system of any of claims 197-225, further comprising a network interface component coupled to the at least one processor, wherein the network interface component is configured to receive at least a second portion of the input data over a communication network. 227. A system comprising: at least one trained encoder configured to map input data to at least one hyper- geometric space; and at least one decoder configured to output an association of members of a data set of the input data based on the map. 228. The system of claim 227, further comprising at least one processor configured to execute the at least one trained encoder. 229. The system of any of claims 227-228, wherein the at least one hyper-geometric space includes a surface of a hypersphere. 230. The system of any of claims 227-228, wherein the at least one hyper-geometric space includes a surface of a polytope. 231. The system of any of claims 227-228, wherein the at least one hyper-geometric space includes a surface of a hyper-cube. 232. The system any of claims 228-231, wherein the at least one processor is further configured to calculate an energy metric relating to a distance separating the data set in the at least one hyper-geometric space. 233. The system of claim 232, wherein the energy metric includes an exponential term. 234. The system of claim 232, wherein the energy metric includes at least one term selected from a group consisting of: an exponential term; a logarithmic term; a sigmoidal term; and a continuous piecewise linear term. 235. The system of any of claims 232-234, wherein the energy metric further relates to a hyper-parameter of the at least one hyper-geometric space. 236. The system of any of claims 228-235, wherein the at least one trained encoder is further configured to group vector representations of the input data into a plurality of clusters in the at least one space. 237. The system of any of claims 228-236, wherein the at least one processor is further configured to determine, based on the energy metric, a likelihood that members of a data set are associated with one another. 238. The system of any of claims 228-236, wherein the at least one processor is further configured to filter out portions of the input data prior to mapping the input data to the at least one hyper-geometric space. 239. The system of claim 238, wherein the at least one processor is further configured to remove portions of the input data having less than a threshold level of correlation. 240. The system of any of claims 237-239, wherein the members of the data set are of a same data domain. 241. The system of claim 240, wherein a first member of the data set is of a first data modality within the same data domain, and a second member of the data set is of a second data modality within the same data domain. 242. The system of claim 241, wherein the at least one trained encoder includes: a first modality encoder configured to map the first member of the data set to a first modality hyper-geometric space; and a second modality encoder configured to map the second member of the data set to a second modality hyper-geometric space. 243. The system of any of claims 241-242, wherein the same data domain is compounds, and the first and second data modalities are selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 244. The system of any of claims 241-242, wherein the same data domain is diseases, and the first and second data modalities are selected from a group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 245. The system of any of claims 237-239, wherein first and second members of the data set are of respective first and second domains. 246. The system of claim 245, wherein the at least one trained encoder includes: at least one first domain encoder configured to map the first member of the data set to a first domain hyper-geometric space; and at least one second domain encoder configured to map the second member of the data set to a second domain hyper-geometric space. 247. The system of any of claims 245-246, wherein: the at least one first domain encoder includes a first modality encoder configured to map the first member of the data set to a first modality hyper-geometric space of the first domain; and the at least one second domain encoder includes a second modality encoder configured to map the second member of the data set to a second modality hyper-geometric space of the second domain. 248. The system of any of claims 245-247, wherein the first and second domains are compounds and diseases, respectively. 249. The system of any of claims 245-247, wherein the first and second domains are financial market conditions and financial market events, respectively. 250. The system of any of claims 245-247, wherein the first and second domains are computer network conditions and computer network security events, respectively. 251. The system of any of claims 245-247, wherein the first and second domains are vehicle traffic conditions and vehicle accident events, respectively. 252. The system of any of claims 245-247, wherein the first and second domains are real images and/or videos and fake images and/or videos, respectively. 253. The system of any of claims 245-247, wherein the first and second domains are music and/or movie preferences and music and/or movies, respectively. 254. The system of any of claims 245-247, wherein the first and second domains are social media trends and political events, respectively. 255. The system of any of claims 245-247, wherein the first and second domains are environmental phenomena and/or human activity and natural disaster events, respectively. 256. The system of any of claims 228-255, further comprising a user interface component coupled to the at least one processor, wherein the user interface component is configured to receive at least a first portion of the input data from a user. 257. The system of claim 256, wherein the user interface component includes at least one member selected from a group consisting of: a mouse; a keyboard; a touchscreen; and a microphone. 258. The system of any of claims 228-257, further comprising a network interface component coupled to the at least one processor, wherein the network interface component is configured to receive at least a second portion of the input data over a communication network. 259. A system, comprising: at least one encoder configured to: map labeled input data to at least one space; and adjust weights and/or biases of the encoder based on an energy metric relating to the map. 260. The system of claim 259, further comprising at least one processor configured to execute the at least one encoder. 261. The system of claim 260, wherein the at least one processor is further configured to: calculate a loss function using the energy metric; and adjust the weights and/or biases of the encoder based on the loss function. 262. The system of any of claims 260-261, wherein the at least one processor is further configured to calculate the energy metric using a distance, in the at least one space, separating a data pair of the input data. 263. The system of claim 262, wherein the energy metric includes an exponential term that includes the distance. 264. The system of claim 262, wherein the energy metric includes a logarithmic term that includes the distance. 265. The system of claim 262, wherein the energy metric includes a sigmoidal term that includes the distance. 266. The system method of claim 262, wherein the energy metric includes a continuous piecewise linear term that includes the distance. 267. The system of any of claims 262-266, wherein: an impact of the energy metric on the weights and/or biases of the encoder increases as the distance increases; and the data pair share a common characteristic. 268. The system of claim 267, wherein: the energy metric further relates to a second distance, in the at least one space, separating a second data pair of the input data; the impact of the energy metric on the weights and/or biases of the encoder increases as the distance decreases; and the second data pair do not have the common characteristic. 269. The system of any of claims 267-268, wherein the at least one processor is further configured to label the input data to emphasize the common characteristic. 270. The system of claim 269, wherein the at least one processor is further configured to de-emphasize a second common characteristic. 271. The system of any of claims 259-270, wherein the at least one space includes a hyper- geometric space. 272. The system of claim 271, wherein the hyper-geometric space includes a surface of a hypersphere. 273. The system of any of claims 270-271, wherein the energy metric includes a hyper- parameter of the hyper-geometric space. 274. The system of any of claims 269-273, wherein the at least one processor is further configured to filter out portions of the input data prior to labeling the input data. 275. The system of claim 273, wherein the at least one processor is further configured to remove portions of the input data having less than a threshold level of correlation. 276. The system of any of claims 262-275, wherein the members of the data pair are of a same data domain. 277. The system of claim 276, wherein a first member of the data pair is of a first data modality within the same data domain, and a second member of the data pair is of a second data modality within the same data domain. 278. The system of claim 277, wherein the at least one encoder comprises: a first modality encoder configured to map the first member of the data pair to a first modality space; and a second modality encoder configured to map the second member of the data pair to a second modality space. 279. The system of any of claims 276-278, wherein the same data domain is compounds, and the first and second data modalities are selected from a group consisting of: compound gene expression data; compound chemical structure data; compound target data; and compound side-effect data. 280. The system of any of claims 276-278, wherein the same data domain is diseases, and the first and second data modalities are selected from a group consisting of: disease gene expression data; disease symptom data; and disease biological pathway data. 281. The system of any of claims 262-275, wherein first and second members of the data pair are of respective first and second domains. 282. The system of claim 281, wherein the at least one encoder includes: at least one first domain encoder configured to map the first member of the data pair to a first domain space; and at least one second domain encoder configured to map the second member of the data pair to a second domain space. 283. The system of any of claims 281-282, wherein: the at least one first domain encoder includes a first modality encoder configured to map the first member of the data pair to a first modality space of the first domain; and the at least one second domain encoder includes a second modality encoder configured to map the second member of the data pair to a second modality space of the second domain. 284. The system of any of claims 281-283, wherein the first and second domains are compounds and diseases, respectively. 285. The system of any of claims 281-283, wherein the first and second domains are financial market conditions and financial market events, respectively. 286. The system of any of claims 281-283, wherein the first and second domains are computer network conditions and computer network security events, respectively. 287. The system of any of claims 281-283, wherein the first and second domains are vehicle traffic conditions and vehicle accident events, respectively. 288. The system of any of claims 281-283, wherein the first and second domains are real images and/or videos and fake images and/or videos, respectively. 289. The system of any of claims 281-283, wherein the first and second domains are music and/or movie preferences and music and/or movies, respectively. 290. The system of any of claims 281-283, wherein the first and second domains are social media trends and political events, respectively. 291. The system of any of claims 281-283, wherein the first and second domains are environmental phenomena and/or human activity and natural disaster events, respectively. 292. The system of any of claims 262-291, wherein the at least one processor is further configured to initialize the encoder at least in part by: mapping, to the at least one space, the labeled input data; calculating an initial loss function prior to calculating the energy metric; and adjusting the weights and/or biases based on the initial loss function. 293. The system of claim 292, wherein the at least one processor is configured to calculate the initial loss function using an initial distance separating the data pair. 294. The system of any of claims 292-293, wherein the at least one processor is further configured to uniformly distribute the labeled input data in the at least one space after adjusting the weights and/or biases based on the initial loss function. 295. The system of any of claims 260-294, further comprising a user interface component coupled to the at least one processor, wherein the user interface component is configured to receive at least a first portion of the input data from a user. 296. The system of claim 295, wherein the user interface component includes at least one member selected from a group consisting of: a mouse; a keyboard; a touchscreen; and a microphone. 297. The system of any of claims 260-296, further comprising a network interface component coupled to the at least one processor, wherein the network interface component is configured to receive at least a second portion of the input data over a communication network. |