Glp-1/gip dual agonists

Abstract

The present disclosure provides methods and systems for hyperparameter tuning and benchmarking.

1 . A computer-implemented method, said method comprising:
 (a) until a stopping criterion is met:
 (i) using a digital computer to generate a hyperparameter configuration from a hyperparameter search space, and storing said hyperparameter configuration in memory; 
 (ii) using said hyperparameter configuration to perform benchmarking on a computational procedure, said benchmarking comprising:
 (a) using a computational platform to:
 (i) perform said computational procedure for each computational task of a problem class, wherein said each computational task of said problem class is configured using at least said hyperparameter configuration, and 
 (ii) store a result generated upon performance of said computational procedure in memory, and 
 
 (b) calculating a value of said at least one metric to evaluate said result corresponding to a performance of said computational procedure, and storing said value in memory; 
 
   (b) using said result and said value to select at least one hyperparameter configuration from said hyperparameter configuration of (a); and   (c) electronically outputting said at least one hyperparameter configuration selected in (b) and said result corresponding to said performance of said at least one hyperparameter configuration selected in (c).   
     
     
         2 . The method of  claim 1 , further comprising, prior to (a), using an interface of said digital computer to obtain an indication of: (i) said problem class comprising at least one computational task, (ii) said hyperparameter search space, (iii) said computational procedure, and (iv) at least one metric for evaluating a performance of said computational procedure. 
     
     
         3 . The method of  claim 1 , wherein said using said digital computer to generate said hyperparameter configuration from said hyperparameter search space is performed using said result and said value of said at least one metric. 
     
     
         4 . The method of  claim 1 , wherein (a) further comprises obtaining a hyperparameter optimization algorithm, wherein said using said digital computer to generate said hyperparameter configuration is performed using said hyperparameter optimization algorithm. 
     
     
         5 . The method of  claim 1 , wherein (a) further comprises obtaining at least one hyperparameter constraint, wherein said using said digital computer to generate said hyperparameter configuration comprises evaluation of said at least one hyperparameter constraint. 
     
     
         6 . The method of  claim 1 , wherein (c) further comprises generating an interactive visualization of said result and said value. 
     
     
         7 . The method of  claim 1 , wherein said computational task is an optimization task. 
     
     
         8 . The method of  claim 1 , wherein said at least one hyperparameter configuration selected in (c) comprises a tuned hyperparameter configuration. 
     
     
         9 . The method of  claim 1 , wherein said benchmarking occurs in an absence of tuning said hyperparameter configuration. 
     
     
         10 . A system configured to benchmark at least one computational procedure using at least one problem class, each problem class comprising at least one computational task, said system comprising:
 (a) a digital computer comprising an interface and a non-transitory computer readable medium operatively coupled to a processor, said non-transitory computer readable medium comprising instructions, wherein said processor is configured to execute said instructions to at least:
 (i) generate a hyperparameter configuration from a hyperparameter search space, and store said hyperparameter configuration in memory; 
   (b) a memory operatively coupled to said digital computer, wherein said memory stores at least a computational result and value of said at least one metric; and   (c) a computational platform operatively coupled to said digital computer, said computational platform comprising at least one processor and a readout control system, wherein said computational platform is configured at least to:
 (ii) receive from said digital computer an indication of said problem class comprising said computational task, said computational procedure, and said hyperparameter configuration, and 
 (iii) perform said computational procedure using a received hyperparameter configuration to generate a result, and 
 (iv) via said readout control, store said result of said computational procedure in memory. 
   
     
     
         11 . The system of  claim 10 , wherein (a) further comprises obtaining an indication of: (i) said problem class comprising a computational task, (ii) said hyperparameter search space, (iii) said computational procedure, and (iv) at least one metric for evaluating a performance of said computational procedure. 
     
     
         12 . The system of  claim 10 , wherein said digital computer comprises multiple processors that are configured to perform said at least one computational procedure in parallel. 
     
     
         13 . The system of  claim 10 , wherein said computational platform is further configured to receive at least one metric for evaluating a performance of said computational procedure and to calculate a value of said at least one metric and store said value in a memory. 
     
     
         14 . The system of  claim 10 , wherein said digital computer is further configured to calculate a value of said at least one metric and store said value in said memory. 
     
     
         15 . The system of  claim 10 , wherein said computational platform comprises at least one member selected from the group consisting of a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU), a tensor streaming processor (TSP), a quantum computer, a quantum annealer, an integrated photonic coherent Ising machine, or an optical quantum computer. 
     
     
         16 . The system of  claim 10 , further comprising a plurality of virtual machines, and optionally wherein said plurality of virtual machines are hosted on said computational platform. 
     
     
         17 . The system of  claim 10 , wherein said memory is operatively coupled to said digital computer over a network. 
     
     
         18 . The system of  claim 10 , wherein one or more processors of said at least one processor is located on a cloud. 
     
     
         19 . The system of  claim 10 , wherein said indication of said computational procedure comprises an indication of a computer-implemented method and a type of hardware. 
     
     
         20 . The system of  claim 10 , wherein an indication of said computational procedure comprises an indication of a meta-solver comprising two or more computational procedures. 
     
     
         21 . The system of  claim 10 , wherein said memory is part of a database. 
     
     
         22 . The system of  claim 10 , wherein said system is configured to set an automated pipeline. 
     
     
         23 . The system of  claim 22 , wherein said automated pipeline comprises hyperparameter tuning or benchmarking. 
     
     
         24 . A method for setting an automated pipeline comprising hyperparameter tuning and benchmarking, said method comprising:
 (a) selecting one or more computational procedures;   (b) selecting one or more hardware types;   (c) determining whether to tune a hyperparameter for said selected one or more computational procedures, and, if so determined:
 (i) selecting a hyperparameter search space corresponding to said selected one or more computational procedures and said one or more hardware types; 
 (ii) determining how to select a computational task sample; 
   (d) determining whether to benchmark said selected one or more computational procedures; and   (e) determining whether to visualize said tuning and said benchmarking results.   
     
     
         25 . The method of  claim 24 , wherein said one or more hardware types comprises at least two machines, and wherein said automated pipeline comprises:
 (a) receiving an indication of at least one problem class comprising a plurality of computational tasks, at least one computational procedure, and at least one metric for evaluating a performance of said at least one computational procedure;   (b) for each problem class of said at least one problem class:
 (i) selecting a sample of a computational task of said plurality of computational tasks comprising at least one computational task from said problem class, 
 (ii) for each computational procedure of said at least one computational procedure:
 (a) generating one or more hyperparameter configurations, 
 (b) using said computational task sample to benchmark said computational procedure, said benchmark comprising performing said computational procedure for each of said selected computational tasks using said at least two machines and evaluating results generated by said performing by calculating values of said at least one metric and storing said values in said database, 
 (c) providing said results comprising hyperparameter configurations and corresponding computational results, 
 (d) repeating (a)-(c) until a stopping criterion is met, and 
 (e) selecting at least one hyperparameter configuration using results of said benchmarking, 
 
 (iii) using remaining computational tasks and said hyperparameter configuration to benchmark each of said at least one computational procedure; and 
   (c) generating an interactive visualization of benchmarking results from (iii) comprising said hyperparameter configurations, corresponding computational results, and values of said at least one metric.