Patent Application
20240028931
Embodiments pertain to computer architecture. Some embodiments relate to artificial intelligence. Some embodiments relate to a directed acyclic graph of actions for reinforcement learning.
Some problems may include multiple decisions, with multiple different options for each decision. This may result in a large possible recommendation space, which may be difficult for reinforcement learning agents to explore to adequately train the reinforcement learning agents. Techniques for leveraging reinforcement learning in spaces with multiple decisions may be desirable.
The following description and the drawings sufficiently illustrate specific embodiments to enable those skilled in the art to practice them. Other embodiments may incorporate structural, logical, electrical, process, and other changes. Portions and features of some embodiments may be included in, or substituted for, those of other embodiments. Embodiments set forth in the claims encompass all available equivalents of those claims.
Aspects of the present technology may be implemented as part of a computer system. The computer system may be one physical machine, or may be distributed among multiple physical machines, such as by role or function, or by process thread in the case of a cloud computing distributed model. In various embodiments, aspects of the technology may be configured to run in virtual machines that in turn are executed on one or more physical machines. It will be understood by persons of skill in the art that features of the technology may be realized by a variety of different suitable machine implementations.
The system includes various engines, each of which is constructed, programmed, configured, or otherwise adapted, to carry out a function or set of functions. The term engine as used herein means a tangible device, component, or arrangement of components implemented using hardware, such as by an application specific integrated circuit (ASIC) or field-programmable gate array (FPGA), for example, or as a combination of hardware and software, such as by a processor-based computing platform and a set of program instructions that transform the computing platform into a special-purpose device to implement the particular functionality. An engine may also be implemented as a combination of the two, with certain functions facilitated by hardware alone, and other functions facilitated by a combination of hardware and software.
In an example, the software may reside in executable or non-executable form on a tangible machine-readable storage medium. Software residing in non-executable form may be compiled, translated, or otherwise converted to an executable form prior to, or during, runtime. In an example, the software, when executed by the underlying hardware of the engine, causes the hardware to perform the specified operations. Accordingly, an engine is physically constructed, or specifically configured (e.g., hardwired), or temporarily configured (e.g., programmed) to operate in a specified manner or to perform part or all of any operations described herein in connection with that engine.
Considering examples in which engines are temporarily configured, each of the engines may be instantiated at different moments in time. For example, where the engines comprise a general-purpose hardware processor core configured using software, the general-purpose hardware processor core may be configured as respective different engines at different times. Software may accordingly configure a hardware processor core, for example, to constitute a particular engine at one instance of time and to constitute a different engine at a different instance of time.
In certain implementations, at least a portion, and in some cases, all, of an engine may be executed on the processor(s) of one or more computers that execute an operating system, system programs, and application programs, while also implementing the engine using multitasking, multithreading, distributed (e.g., cluster, peer-peer, cloud, etc.) processing where appropriate, or other such techniques. Accordingly, each engine may be realized in a variety of suitable configurations, and should generally not be limited to any particular implementation exemplified herein, unless such limitations are expressly called out.
In addition, an engine may itself be composed of more than one sub-engines, each of which may be regarded as an engine in its own right. Moreover, in the embodiments described herein, each of the various engines corresponds to a defined functionality; however, it should be understood that in other contemplated embodiments, each functionality may be distributed to more than one engine. Likewise, in other contemplated embodiments, multiple defined functionalities may be implemented by a single engine that performs those multiple functions, possibly alongside other functions, or distributed differently among a set of engines than specifically illustrated in the examples herein.
As used herein, the term “model” encompasses its plain and ordinary meaning. A model may include, among other things, one or more engines which receive an input and compute an output based on the input. The output may be a classification. For example, an image file may be classified as depicting a cat or not depicting a cat. Alternatively, the image file may be assigned a numeric score indicating a likelihood whether the image file depicts the cat, and image files with a score exceeding a threshold (e.g., 0.9 or 0.95) may be determined to depict the cat.
This document may reference a specific number of things (e.g., “six mobile devices”). Unless explicitly set forth otherwise, the numbers provided are examples only and may be replaced with any positive integer, integer or real number, as would make sense for a given situation. For example, “six mobile devices” may, in alternative embodiments, include any positive integer number of mobile devices. Unless otherwise mentioned, an object referred to in singular form (e.g., “a computer” or “the computer”) may include one or multiple objects (e.g., “the computer” may refer to one or multiple computers).
Machine learning is a field of study that gives computers the ability to learn without being explicitly programmed. Machine learning explores the study and construction of algorithms, also referred to herein as tools, which may learn from existing data and make predictions about new data. Such machine-learning tools operate by building a model from example training data 112 in order to make data-driven predictions or decisions expressed as outputs or assessments 120. Although example embodiments are presented with respect to a few machine-learning tools, the principles presented herein may be applied to other machine-learning tools.
In some example embodiments, different machine-learning tools may be used. For example, Logistic Regression (LR), Naive-Bayes, Random Forest (RF), neural networks (NN), matrix factorization, and Support Vector Machines (SVM) tools may be used for classifying or scoring job postings.
Two common types of problems in machine learning are classification problems and regression problems. Classification problems, also referred to as categorization problems, aim at classifying items into one of several category values (for example, is this object an apple or an orange). Regression algorithms aim at quantifying some items (for example, by providing a value that is a real number). The machine-learning algorithms utilize the training data 112 to find correlations among identified features 102 that affect the outcome.
The machine-learning algorithms utilize features 102 for analyzing the data to generate assessments 120. A feature 102 is an individual measurable property of a phenomenon being observed. The concept of a feature is related to that of an explanatory variable used in statistical techniques such as linear regression. Choosing informative, discriminating, and independent features is important for effective operation of the MLP in pattern recognition, classification, and regression. Features may be of different types, such as numeric features, strings, and graphs.
In one example embodiment, the features 102 may be of different types and may include one or more of words of the message 103, message concepts 104, communication history 105, past user behavior 106, subject of the message 107, other message attributes 108, sender 109, and user data 110.
The machine-learning algorithms utilize the training data 112 to find correlations among the identified features 102 that affect the outcome or assessment 120. In some example embodiments, the training data 112 includes labeled data, which is known data for one or more identified features 102 and one or more outcomes, such as detecting communication patterns, detecting the meaning of the message, generating a summary of the message, detecting action items in the message, detecting urgency in the message, detecting a relationship of the user to the sender, calculating score attributes, calculating message scores, etc.
With the training data 112 and the identified features 102, the machine-learning tool is trained at operation 114. The machine-learning tool appraises the value of the features 102 as they correlate to the training data 112. The result of the training is the trained machine-learning program 116.
When the machine-learning program 116 is used to perform an assessment, new data 118 is provided as an input to the trained machine-learning program 116, and the machine-learning program 116 generates the assessment 120 as output. For example, when a message is checked for an action item, the machine-learning program utilizes the message content and message metadata to determine if there is a request for an action in the message.
Machine learning techniques train models to accurately make predictions on data fed into the models (e.g., what was said by a user in a given utterance; whether a noun is a person, place, or thing; what the weather will be like tomorrow). During a learning phase, the models are developed against a training dataset of inputs to optimize the models to correctly predict the output for a given input. Generally, the learning phase may be supervised, semi-supervised, or unsupervised; indicating a decreasing level to which the “correct” outputs are provided in correspondence to the training inputs. In a supervised learning phase, all of the outputs are provided to the model and the model is directed to develop a general rule or algorithm that maps the input to the output. In contrast, in an unsupervised learning phase, the desired output is not provided for the inputs so that the model may develop its own rules to discover relationships within the training dataset. In a semi-supervised learning phase, an incompletely labeled training set is provided, with some of the outputs known and some unknown for the training dataset.
Models may be run against a training dataset for several epochs (e.g., iterations), in which the training dataset is repeatedly fed into the model to refine its results. For example, in a supervised learning phase, a model is developed to predict the output for a given set of inputs, and is evaluated over several epochs to more reliably provide the output that is specified as corresponding to the given input for the greatest number of inputs for the training dataset. In another example, for an unsupervised learning phase, a model is developed to cluster the dataset into n groups, and is evaluated over several epochs as to how consistently it places a given input into a given group and how reliably it produces the n desired clusters across each epoch.
Once an epoch is run, the models are evaluated and the values of their variables are adjusted to attempt to better refine the model in an iterative fashion. In various aspects, the evaluations are biased against false negatives, biased against false positives, or evenly biased with respect to the overall accuracy of the model. The values may be adjusted in several ways depending on the machine learning technique used. For example, in a genetic or evolutionary algorithm, the values for the models that are most successful in predicting the desired outputs are used to develop values for models to use during the subsequent epoch, which may include random variation/mutation to provide additional data points. One of ordinary skill in the art will be familiar with several other machine learning algorithms that may be applied with the present disclosure, including linear regression, random forests, decision tree learning, neural networks, deep neural networks, etc.
Each model develops a rule or algorithm over several epochs by varying the values of one or more variables affecting the inputs to more closely map to a desired result, but as the training dataset may be varied, and is preferably very large, perfect accuracy and precision may not be achievable. A number of epochs that make up a learning phase, therefore, may be set as a given number of trials or a fixed time/computing budget, or may be terminated before that number/budget is reached when the accuracy of a given model is high enough or low enough or an accuracy plateau has been reached. For example, if the training phase is designed to run n epochs and produce a model with at least 95% accuracy, and such a model is produced before the nth epoch, the learning phase may end early and use the produced model satisfying the end-goal accuracy threshold. Similarly, if a given model is inaccurate enough to satisfy a random chance threshold (e.g., the model is only 55% accurate in determining true/false outputs for given inputs), the learning phase for that model may be terminated early, although other models in the learning phase may continue training. Similarly, when a given model continues to provide similar accuracy or vacillate in its results across multiple epochs—having reached a performance plateau—the learning phase for the given model may terminate before the epoch number/computing budget is reached.
Once the learning phase is complete, the models are finalized. In some example embodiments, models that are finalized are evaluated against testing criteria. In a first example, a testing dataset that includes known outputs for its inputs is fed into the finalized models to determine an accuracy of the model in handling data that it has not been trained on. In a second example, a false positive rate or false negative rate may be used to evaluate the models after finalization. In a third example, a delineation between data clusterings is used to select a model that produces the clearest bounds for its clusters of data.
As illustrated at the bottom of
In some example embodiments, the neural network 204 (e.g., deep learning, deep convolutional, or recurrent neural network) comprises a series of neurons 208, such as Long Short Term Memory (LSTM) nodes, arranged into a network. A neuron 208 is an architectural element used in data processing and artificial intelligence, particularly machine learning, which includes memory that may determine when to “remember” and when to “forget” values held in that memory based on the weights of inputs provided to the given neuron 208. Each of the neurons 208 used herein are configured to accept a predefined number of inputs from other neurons 208 in the neural network 204 to provide relational and sub-relational outputs for the content of the frames being analyzed. Individual neurons 208 may be chained together and/or organized into tree structures in various configurations of neural networks to provide interactions and relationship learning modeling for how each of the frames in an utterance are related to one another.
For example, an LSTM node serving as a neuron includes several gates to handle input vectors (e.g., phonemes from an utterance), a memory cell, and an output vector (e.g., contextual representation). The input gate and output gate control the information flowing into and out of the memory cell, respectively, whereas forget gates optionally remove information from the memory cell based on the inputs from linked cells earlier in the neural network. Weights and bias vectors for the various gates are adjusted over the course of a training phase, and once the training phase is complete, those weights and biases are finalized for normal operation. One of skill in the art will appreciate that neurons and neural networks may be constructed programmatically (e.g., via software instructions) or via specialized hardware linking each neuron to form the neural network.
Neural networks utilize features for analyzing the data to generate assessments (e.g., recognize units of speech). A feature is an individual measurable property of a phenomenon being observed. The concept of feature is related to that of an explanatory variable used in statistical techniques such as linear regression. Further, deep features represent the output of nodes in hidden layers of the deep neural network.
A neural network, sometimes referred to as an artificial neural network, is a computing system/apparatus based on consideration of biological neural networks of animal brains. Such systems/apparatus progressively improve performance, which is referred to as learning, to perform tasks, typically without task-specific programming. For example, in image recognition, a neural network may be taught to identify images that contain an object by analyzing example images that have been tagged with a name for the object and, having learnt the object and name, may use the analytic results to identify the object in untagged images. A neural network is based on a collection of connected units called neurons, where each connection, called a synapse, between neurons can transmit a unidirectional signal with an activating strength that varies with the strength of the connection. The receiving neuron can activate and propagate a signal to downstream neurons connected to it, typically based on whether the combined incoming signals, which are from potentially many transmitting neurons, are of sufficient strength, where strength is a parameter.
A deep neural network (DNN) is a stacked neural network, which is composed of multiple layers. The layers are composed of nodes, which are locations where computation occurs, loosely patterned on a neuron in the human brain, which fires when it encounters sufficient stimuli. A node combines input from the data with a set of coefficients, or weights, that either amplify or dampen that input, which assigns significance to inputs for the task the algorithm is trying to learn. These input-weight products are summed, and the sum is passed through what is called a node's activation function, to determine whether and to what extent that signal progresses further through the network to affect the ultimate outcome. A DNN uses a cascade of many layers of non-linear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. Higher-level features are derived from lower-level features to form a hierarchical representation. The layers following the input layer may be convolution layers that produce feature maps that are filtering results of the inputs and are used by the next convolution layer.
In training of a DNN architecture, a regression, which is structured as a set of statistical processes for estimating the relationships among variables, can include a minimization of a cost function. The cost function may be implemented as a function to return a number representing how well the neural network performed in mapping training examples to correct output. In training, if the cost function value is not within a pre-determined range, based on the known training images, backpropagation is used, where backpropagation is a common method of training artificial neural networks that are used with an optimization method such as a stochastic gradient descent (SGD) method.
Use of backpropagation can include propagation and weight update. When an input is presented to the neural network, it is propagated forward through the neural network, layer by layer, until it reaches the output layer. The output of the neural network is then compared to the desired output, using the cost function, and an error value is calculated for each of the nodes in the output layer. The error values are propagated backwards, starting from the output, until each node has an associated error value which roughly represents its contribution to the original output. Backpropagation can use these error values to calculate the gradient of the cost function with respect to the weights in the neural network. The calculated gradient is fed to the selected optimization method to update the weights to attempt to minimize the cost function.
The training set 302 includes a plurality of images 306 for each class 304 (e.g., image 306), and each image is associated with one of the categories to be recognized (e.g., a class). The machine learning program is trained 308 with the training data to generate a classifier 310 operable to recognize images. In some example embodiments, the machine learning program is a DNN.
When an input image 312 is to be recognized, the classifier 310 analyzes the input image 312 to identify the class (e.g., class 314) corresponding to the input image 312.
With the development of deep convolutional neural networks, the focus in face recognition has been to learn a good face feature space, in which faces of the same person are close to each other, and faces of different persons are far away from each other. For example, the verification task with the LFW (Labeled Faces in the Wild) dataset has been often used for face verification.
Many face identification tasks (e.g., MegaFace and LFW) are based on a similarity comparison between the images in the gallery set and the query set, which is essentially a K-nearest-neighborhood (KNN) method to estimate the person's identity. In the ideal case, there is a good face feature extractor (inter-class distance is always larger than the intra-class distance), and the KNN method is adequate to estimate the person's identity.
Feature extraction is a process to reduce the amount of resources required to describe a large set of data. When performing analysis of complex data, one of the major problems stems from the number of variables involved. Analysis with a large number of variables generally requires a large amount of memory and computational power, and it may cause a classification algorithm to overfit to training samples and generalize poorly to new samples. Feature extraction is a general term describing methods of constructing combinations of variables to get around these large data-set problems while still describing the data with sufficient accuracy for the desired purpose.
In some example embodiments, feature extraction starts from an initial set of measured data and builds derived values (features) intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps. Further, feature extraction is related to dimensionality reduction, such as reducing large vectors (sometimes with very sparse data) to smaller vectors capturing the same, or similar, amount of information.
Determining a subset of the initial features is called feature selection. The selected features are expected to contain the relevant information from the input data, so that the desired task can be performed by using this reduced representation instead of the complete initial data. DNN utilizes a stack of layers, where each layer performs a function. For example, the layer could be a convolution, a non-linear transform, the calculation of an average, etc. Eventually this DNN produces outputs by classifier 414. In
As shown in
In some example embodiments, the structure of each layer is predefined. For example, a convolution layer may contain small convolution kernels and their respective convolution parameters, and a summation layer may calculate the sum, or the weighted sum, of two pixels of the input image. Training assists in defining the weight coefficients for the summation.
One way to improve the performance of DNNs is to identify newer structures for the feature-extraction layers, and another way is by improving the way the parameters are identified at the different layers for accomplishing a desired task. The challenge is that for a typical neural network, there may be millions of parameters to be optimized. Trying to optimize all these parameters from scratch may take hours, days, or even weeks, depending on the amount of computing resources available and the amount of data in the training set.
Examples, as described herein, may include, or may operate on, logic or a number of components, modules, or mechanisms. Modules and components are tangible entities (e.g., hardware) capable of performing specified operations and may be configured or arranged in a certain manner. In an example, circuits may be arranged (e.g., internally or with respect to external entities such as other circuits) in a specified manner as a module. In an example, the whole or part of one or more computer systems/apparatus (e.g., a standalone, client or server computer system) or one or more hardware processors may be configured by firmware or software (e.g., instructions, an application portion, or an application) as a module that operates to perform specified operations. In an example, the software may reside on a machine readable medium. In an example, the software, when executed by the underlying hardware of the module, causes the hardware to perform the specified operations.
Accordingly, the term “module” (and “component”) is understood to encompass a tangible entity, be that an entity that is physically constructed, specifically configured (e.g., hardwired), or temporarily (e.g., transitorily) configured (e.g., programmed) to operate in a specified manner or to perform part or all of any operation described herein. Considering examples in which modules are temporarily configured, each of the modules need not be instantiated at any one moment in time. For example, where the modules comprise a general-purpose hardware processor configured using software, the general-purpose hardware processor may be configured as respective different modules at different times. Software may accordingly configure a hardware processor, for example, to constitute a particular module at one instance of time and to constitute a different module at a different instance of time.
The computing machine 500 may include a hardware processor 502 (e.g., a central processing unit (CPU), a GPU, a hardware processor core, or any combination thereof), a main memory 504 and a static memory 506, some or all of which may communicate with each other via an interlink (e.g., bus) 508. Although not shown, the main memory 504 may contain any or all of removable storage and non-removable storage, volatile memory or non-volatile memory. The computing machine 500 may further include a video display unit 510 (or other display unit), an alphanumeric input device 512 (e.g., a keyboard), and a user interface (UI) navigation device 514 (e.g., a mouse). In an example, the display unit 510, input device 512 and UI navigation device 514 may be a touch screen display. The computing machine 500 may additionally include a storage device (e.g., drive unit) 516, a signal generation device 518 (e.g., a speaker), a network interface device 520, and one or more sensors 521, such as a global positioning system (GPS) sensor, compass, accelerometer, or other sensor. The computing machine 500 may include an output controller 528, such as a serial (e.g., universal serial bus (USB), parallel, or other wired or wireless (e.g., infrared (IR), near field communication (NFC), etc.) connection to communicate or control one or more peripheral devices (e.g., a printer, card reader, etc.).
The drive unit 516 (e.g., a storage device) may include a machine readable medium 522 on which is stored one or more sets of data structures or instructions 524 (e.g., software) embodying or utilized by any one or more of the techniques or functions described herein. The instructions 524 may also reside, completely or at least partially, within the main memory 504, within static memory 506, or within the hardware processor 502 during execution thereof by the computing machine 500. In an example, one or any combination of the hardware processor 502, the main memory 504, the static memory 506, or the storage device 516 may constitute machine readable media.
While the machine readable medium 522 is illustrated as a single medium, the term “machine readable medium” may include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers) configured to store the one or more instructions 524.
The term “machine readable medium” may include any medium that is capable of storing, encoding, or carrying instructions for execution by the computing machine 500 and that cause the computing machine 500 to perform any one or more of the techniques of the present disclosure, or that is capable of storing, encoding or carrying data structures used by or associated with such instructions. Non-limiting machine readable medium examples may include solid-state memories, and optical and magnetic media. Specific examples of machine readable media may include: non-volatile memory, such as semiconductor memory devices (e.g., Electrically Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM)) and flash memory devices; magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; Random Access Memory (RAM); and CD-ROM and DVD-ROM disks. In some examples, machine readable media may include non-transitory machine readable media. In some examples, machine readable media may include machine readable media that is not a transitory propagating signal.
The instructions 524 may further be transmitted or received over a communications network 526 using a transmission medium via the network interface device 520 utilizing any one of a number of transfer protocols (e.g., frame relay, internet protocol (IP), transmission control protocol (TCP), user datagram protocol (UDP), hypertext transfer protocol (HTTP), etc.). Example communication networks may include a local area network (LAN), a wide area network (WAN), a packet data network (e.g., the Internet), mobile telephone networks (e.g., cellular networks), Plain Old Telephone (POTS) networks, and wireless data networks (e.g., Institute of Electrical and Electronics Engineers (IEEE) 802.11 family of standards known as Wi-Fi®, IEEE 802.16 family of standards known as WiMax®), IEEE 802.15.4 family of standards, a Long Term Evolution (LTE) family of standards, a Universal Mobile Telecommunications System (UMTS) family of standards, peer-to-peer (P2P) networks, among others. In an example, the network interface device 520 may include one or more physical jacks (e.g., Ethernet, coaxial, or phone jacks) or one or more antennas to connect to the communications network 526.
Reinforcement learning is an area of machine learning concerned with how intelligent agents ought to take actions in an environment in order to maximize the notion of cumulative reward. Reinforcement learning is one of three basic machine learning paradigms, alongside supervised learning and unsupervised learning.
Reinforcement learning differs from supervised learning in not needing labelled input/output pairs be presented, and in not needing sub-optimal actions to be explicitly corrected. Instead, the focus is on finding a balance between exploration (of uncharted territory) and exploitation (of current knowledge). Partially supervised reinforcement learning algorithms can combine the advantages of supervised and reinforcement learning algorithms.
The environment may be stated in the form of a Markov decision process (MDP) because many reinforcement learning algorithms for this context use dynamic programming techniques. One difference between the classical dynamic programming methods and reinforcement learning algorithms is that the latter do not assume knowledge of an exact mathematical model of the MDP and they target large MDPs where exact methods become infeasible. Due to its generality, reinforcement learning is studied in many disciplines, such as game theory, control theory, operations research, information theory, simulation-based optimization, multi-agent systems, swarm intelligence, and statistics.
The reinforcement learning 600 may be useful in contexts outside of playing chess. For example, in electronic marketing, the environment 602 may include information that a business stores (e.g., in its data repositories) about customers or prospects. The interpreter 604 may interpret data in the environment (e.g., whether a customer or prospect selected a link, responded to a message, or made a purchase). The agent 608 may compute the action 610, for example, to send a message to a customer or a prospect using a specified channel (e.g., email or short messaging service (SMS)) at a specified date and time, that impacts the environment 602.
The environment 602 may include data stored in a data repository and/or a memory of a computing machine (e.g., the main memory 504, the static memory 506, and/or the drive unit 516 of the computing machine 500). The interpreter 604 may be implemented using software, hardware, or a combination of software and hardware. The agent 608 may be implemented using software, hardware, or a combination of software and hardware. The environment 602, the interpreter 604, and/or the agent 608 may be implemented at one or more computing machines (e.g., the computing machine 500) using one or more engines, as described herein.
Some implementations relate to, among other things, using reinforcement learning to personalize recommendations. In the online marketing context, a computing machine (e.g., one or more of the computing machines 100) ingests data on customers (e.g., transactions, engagement with application or webpage, product usage, interaction with customer support, engagement with marketing communications (e.g., whether the customer opens or selects links in an email message, uses a coupon received via an instant messaging service, or swipes a push notification), or the like). A reinforcement learning agent (e.g., the agent 608) from a set of reinforcement learning agents uses that data, together with possible actions, to determine a recommendation. For example, a recommendation may be, “Send customer 3867813098 offer template 342 by email next Tuesday at time 8:00 am EST.” The computing machine may receive data every threshold time period (e.g., daily) and may use that data to train the agents to continually improve recommendations based on observed results. The training may be done with respect to a target metric that the agent aims to optimize. The target metric may include at least one or a mathematical combination of: net present value, customer lifetime value, profit from offer, conversion, or the like.
In some cases, the space of possible recommendations (that may be generated by a reinforcement learning agent) is very large. For example, in providing marketing messages to a given customer, the space may include multiple different offers (e.g., five different offer templates), multiple different possible times (e.g., every hour between midnight and 11 pm), multiple different possible days (e.g., seven days in a week), and multiple different possible channels (e.g., email or SMS). This results in a total of 5*24*7*2=1680 possible options for each customer. The space of 1680 possible actions for exploration is very large, making training a reinforcement learning agent to select action(s) effectively leverage a very large number of samples.
Some implementations attempt to solve this problem by using multiple reinforcement learning agents to compute an action, with each reinforcement learning agent computing a value which is a component of the computed action. The value may correspond to the action 610 by the agent 608 shown in
As shown, the dimension 702A has no input edges. Thus, the value for the dimension 702A is computed based on an input dataset (e.g., data about the customer, such as the customer's past purchases and data provided by the customer to a business associated with the DAG 700, such as the customer's home city and state, date of birth, default payment type (e.g., checking account, credit card, or cryptocurrency), and the like and not based on values computed for the dimensions 702B, 702C, and 702D. The dimension 702B has an input edge from the dimension 702A. Thus, the value for the dimension 702B is computed based on the input dataset and the computed value for the dimension 702A, but not based on the values computed for the dimensions 702C and 702D. The dimension 702C has input edges from the dimensions 702A and 702B. Thus, the value for the dimension 702C is computed based on the input dataset and the computed values for the dimensions 702A and 702B, but not based on the value computed for the dimension 702D. The dimension 702D has input edges from the dimensions 702A, 702B, and 702C. Thus, the value for the dimension 702D is computed based on the input dataset and the computed values for the dimensions 702A, 702B, and 702C.
The DAG 700 may be constructed based on a DAG configuration specified by a user (e.g., using a computing machine associated with an employee of a business). As a result, the user may be able to control an order in which the values for the dimensions are computed and which dimensions depend on which other dimensions. The DAG configuration may be stored in a configuration file or another data structure for storing configuration data.
It should be noted that a DAG may be linearized—the nodes of the DAG may be arranged in an order (1, 2, 3, . . . ) such that arrows only go from a lower node to a higher one. For example, the DAG 700 corresponds to the order: dimension 702A, dimension 702B, dimension 702C, dimension 702D. The disclosed technique takes the DAG, determines the order, and then computes the decisions of each agent. In a linearization of a DAG, the first node has no nodes upstream from itself, the second node has at most one node upstream from itself, the third node has at most two nodes upstream from itself, and the nth node has at most n-1 nodes upstream from itself, where n is a number between 1 and the total number of nodes in the DAG.
As used herein, the terms “top level dimension” and “non-top level dimension” encompass their plain and ordinary meaning. A top level dimension includes a dimension that has no nodes upstream from the dimension—in other words, no edges going into the dimension. A non-top level dimension includes a dimension that has at least one node upstream from the dimension—in other words, has at least one edge going into the dimension. In
Another DAG may include Dimension A, Dimension B, and Dimension C. An edge may exist from Dimension A to Dimension B, and there may be no other edges. In this DAG, Dimension A and Dimension C are top level dimensions because these dimensions have no nodes upstream from themselves and no edges pointing into themselves. Dimension B is a non-top level dimension because Dimension B has a node (Dimension A) upstream from Dimension B and has an edge (from Dimension A to Dimension B) going into Dimension B. It should be noted that this DAG may be linearized in multiple ways, as long as Dimension A precedes Dimension B in the linearization. Three possible linearizations are: (i) Dimension A, Dimension B, Dimension C; (ii) Dimension A, Dimension C, Dimension B; and (iii) Dimension C, Dimension A, Dimension B.
As shown in
As further illustrated in
In one example use case, the disclosed technology is being used in a vacation recommendation technology for a user of a travel website or application. The dimension 702A represents a recommended destination city (e.g., New York City, San Francisco, Miami, or the like), the dimension 702B represents a recommended set of flights (e.g., ABC Airlines flight 123 on April 1 and ABC Airlines Flight 456 on April 8), the dimension 702C represents a recommended hotel (e.g., Hotel DEF), and the dimension 702D represents a recommended rental car (e.g., midsize car at Rental Agency GHI) or alternative ground transportation (e.g., public transportation, taxi or ride sharing software).
The user accesses the travel website and requests a vacation recommendation. The user may specify a departure date (e.g., April 1), a return date (e.g., April 8), and his/her home airport (e.g., Detroit). The travel website may have stored data about the user including trips the user has taken in the past, past home airports near which the user has lived, the user's default payment method type, whether the user has a valid passport, and the like. The stored data about the user and the data input by the user may be included in the dataset 806.
The agent 808A accesses the dataset 806 and computes a recommendation for a destination city for the user as the value 810A. For example, the value 810A may be “Las Vegas.” The agent 808B accesses the dataset 806 and the value 810A to recommend a set of round trip flights from the user's home airport (Detroit) to the destination city (Las Vegas) as the value 810B. For example, the agent 808B may select direct round trip flights from Detroit to Las Vegas on Delta Airlines. The agent 808C then accesses the values 810A (Las Vegas) and 810B (recommended flights on Delta Airlines) and the dataset 806 including stored data about the user to recommend a hotel in Las Vegas for the user as the value 810C. For example, the value 810C may correspond to the Bellagio Hotel. The agent 808D then accesses the values 810A (Las Vegas), 810B (recommended flights on Delta Airlines), and 810C (Bellagio Hotel), to generate a rental car or alternative ground transportation recommendation as the value 810D. For example, the recommendation in the value 810D may be to use taxi transportation within Las Vegas.
The values 810A (Las Vegas), 810B (recommended flights on Delta Airlines), 810C (Bellagio Hotel), and 810D (taxi transportation) are used to generate an action 812. The action 812 may be providing the recommendation to the user. The user may then book the proposed itinerary in the action 812 and/or make changes to the proposed itinerary 812. For example, the user might accept the proposed destination city (Las Vegas), and the proposed flights, but might select a different hotel (e.g., Tropicana Hotel instead of Bellagio Hotel) and different ground transportation (e.g., economy car from Hertz Car Rental instead of taxi transportation). The user's selections and/or complete rejection of the itinerary are stored as the result 814. The result 814 may be stored in the dataset 816 (e.g., among the trips the user has taken or among the other data about the user) and/or used to further train the agents 808A, 808B, 808C, and/or 808D via online learning.
It should be noted that, alternative structures of the DAG that are different from the DAG 700 are possible. For example, in the travel recommendation case, the destination city may be computed first. The flight may be computed based on the destination city. The hotel may be computed based on the destination city and not based on the flight. The rental car or alternative ground transportation may be computed based on the destination city and the flight, and not based on the hotel.
At block 902, a computing machine (e.g., the computing machine 500) accesses an input dataset (e.g., the dataset 806) for computing an action (e.g., the action 812) including multiple decision items and a configuration for a DAG (e.g., the DAG configuration 804, which may be associated with the DAG 700). The DAG includes nodes representing dimensions (e.g., the dimensions 702A, 702B, 702C, and 702D). Each dimension is associated with a decision item of the action. The configuration of the DAG may be accessed via one or more configuration files or other data structures. The configuration for the DAG may be generated, by a user, via a graphical user interface.
At block 904, the computing machine computes a value (e.g., the value 810A) for a first decision item associated with a top level dimension (e.g., the dimension 702A) of the DAG. The value for the first decision item is computed based on the input dataset and using a reinforcement learning agent (e.g., the agent 808A) for the top level dimension. The DAG may include unidirectional edges between multiple nodes. As used herein, dimension X is upstream from dimension Y if a unidirectional edge or a path including multiple unidirectional edges exists in the DAG from dimension X to dimension Y. For example, as shown in
At block 906, the computing machine computing a value (e.g., the value 810B) for a second decision item associated with a non-top level dimension (e.g., the dimension 702B) from the DAG. The value for the second decision item is computed based on the input dataset and a specified subset of outputs of reinforcement learning engines for dimensions upstream from the non-top level dimension in the DAG and using a reinforcement learning agent (e.g., the agent 808B) for the non-top level dimension. For example, as shown in
According to some implementations, there may be a single top level dimension, which is upstream from one or more non-top level dimensions in the DAG. In other implementations, there may be multiple top level dimensions. In some implementations, a unidirectional edge or set of edges goes from a top level dimension to a non-top level dimension. (It should be noted that there may be multiple other dimensions, in addition to the top level dimension and the non-top level dimension. Some of the multiple other dimensions may be upstream or downstream from the non-top level dimension. Some of the dimensions may be downstream from the top level dimension. However, there are no dimensions upstream from the top level dimension because, as specified above, the top level dimension does not receive input from other dimensions.) The non-top level dimension may use the output of the top level dimension if there is an edge (or set of edges) going from the top level dimension to the non-top level dimension. In some implementations, the non-top level dimension does not use the output of the top level dimension if there is no edge (or set of edges) going from the top level dimension to the non-top level dimension.
One example of a DAG is an empty DAG with multiple nodes but no edges between the nodes. This DAG may be linearized with any order of the nodes, as all the nodes receive the original input set and none of the nodes receive input from other nodes. In a linearization of a DAG, for the Nth dimension of the DAG, for any N less than or equal to the number of dimensions in the DAG, dimensions upstream of the Nth dimension appear prior to the N in the ordered list.
At block 908, the computing machine provides the computed action (e.g., the action 812). The computed action includes at least the value for the first decision item and the value for the second decision item. In some implementations, the computing machine causes performance or simulation of the computed action. The computing machine determines a result of the computed action (e.g., by running a simulation or observing a result in the external environment, such as whether a customer decides to book a recommended flight, hotel, and/or ground transportation option). The computing machine iteratively trains (e.g., using online learning techniques) the reinforcement learning agent for the top level dimension or the reinforcement learning agent for the non-top level dimension based on the determined result. In one example, the action is transmitting an offer message to a messaging address selected from a set of addresses associated with a user, wherein the multiple decision items include a channel, a day, a time, and an offer message identifier. However, it should be noted that the disclosed technology may be useful in multiple other contexts.
Some embodiments are described as numbered examples (Example 1, 2, 3, etc.). These are provided as examples only and do not limit the technology disclosed herein.
Example 1 is a method comprising: accessing, at a computing machine, an input dataset for computing an action including multiple decision items and a configuration for a directed acyclic graph comprising nodes representing dimensions, each dimension being associated with a decision item of the action; computing a value for a first decision item associated with a top level dimension of the directed acyclic graph, wherein the value for the first decision item is computed based on the input dataset and using a reinforcement learning agent for the top level dimension; computing a value for a second decision item associated with a non-top level dimension from the directed acyclic graph, wherein the value for the second decision item is computed based on the input dataset and a specified subset of outputs of reinforcement learning engines for dimensions upstream from the non-top level dimension in the directed acyclic graph and using a reinforcement learning agent for the non-top level dimension; and providing, via the computing machine, the computed action including at least the value for the first decision item and the value for the second decision item.
In Example 2, the subject matter of Example 1 includes, wherein the directed acyclic graph comprises unidirectional edges between the multiple nodes, wherein an upstream dimension being upstream from the non-top level dimension comprises an edge existing from a node of the upstream dimension to a node of the non-top level dimension.
In Example 3, the subject matter of Examples 1-2 includes, generating an ordered list of the dimensions of the DAG based on a structure of the DAG, wherein the top level dimension occupies a first position in the ordered list, wherein the dimensions upstream from the non-top level dimension appear prior to the non-top level dimension in the ordered list.
In Example 4, the subject matter of Examples 1-3 includes, wherein the configuration for the direct acyclic graph is accessed via one or more configuration files.
In Example 5, the subject matter of Examples 1-4 includes, wherein the configuration for the direct acyclic graph is generated via a graphical user interface.
In Example 6, the subject matter of Examples 1-5 includes, wherein the top level dimension lacks upstream dimensions in the direct acyclic graph.
In Example 7, the subject matter of Examples 1-6 includes, wherein the top level dimension is upstream from the non-top level dimension in the direct acyclic graph, wherein the specified subset of outputs of reinforcement learning agents associated with the dimensions upstream from the non-top level dimension comprises an output of the reinforcement learning agent for the top level dimension.
In Example 8, the subject matter of Examples 1-7 includes, wherein the top level dimension is upstream from the non-top level dimension in the direct acyclic graph, wherein the specified subset of outputs of reinforcement learning engines associated with the dimensions upstream from the non-top level dimension lacks an output of the reinforcement learning agent for the top level dimension.
In Example 9, the subject matter of Examples 1-8 includes, causing, via the computing machine, performance of the computed action; determining a result of the computed action; and iteratively training the reinforcement learning agent for the top level dimension or the reinforcement learning agent for the non-top level dimension based on the determined result.
In Example 10, the subject matter of Examples 1-9 includes, computing a value for at least one decision item associated with a Nth dimension of the directed acyclic graph, wherein N is a positive integer corresponding to a position of the Nth dimension in a linearization of the directed acyclic graph comprising at least N dimensions, wherein the value of the decision item is computed based on the input dataset and a specified subset of outputs of reinforcement learning engines for dimensions upstream from the Nth dimension in the directed acyclic graph and using a reinforcement learning agent for the Nth dimension.
In Example 11, the subject matter of Examples 1-10 includes, wherein the action comprises transmitting an offer message to a user, wherein the multiple decision items comprise a channel, a day, a time, and an offer message identifier.
Example 12 is a non-transitory machine-readable medium storing instructions which, when executed by one or more computing machines, cause the one or more computing machines to perform operations comprising: accessing an input dataset for computing an action including multiple decision items and a configuration for a directed acyclic graph comprising nodes representing dimensions, each dimension being associated with a decision item of the action; computing a value for a first decision item associated with a top level dimension of the directed acyclic graph, wherein the value for the first decision item is computed based on the input dataset and using a reinforcement learning agent for the top level dimension; computing a value for a second decision item associated with a non-top level dimension from the directed acyclic graph, wherein the value for the second decision item is computed based on the input dataset and a specified subset of outputs of reinforcement learning engines for dimensions upstream from the non-top level dimension in the directed acyclic graph and using a reinforcement learning agent for the non-top level dimension; and providing the computed action including at least the value for the first decision item and the value for the second decision item.
In Example 13, the subject matter of Example 12 includes, wherein the directed acyclic graph comprises unidirectional edges between the multiple nodes, wherein an upstream dimension being upstream from the non-top level dimension comprises an edge existing from a node of the upstream dimension to a node of the non-top level dimension.
In Example 14, the subject matter of Examples 12-13 includes, wherein the configuration for the direct acyclic graph is accessed via one or more configuration files.
In Example 15, the subject matter of Examples 12-14 includes, the operations further comprising: generating an ordered list of the dimensions of the DAG based on a structure of the DAG, wherein the top level dimension occupies a first position in the ordered list, wherein the dimensions upstream from the non-top level dimension appear prior to the non-top level dimension in the ordered list.
In Example 16, the subject matter of Examples 12-15 includes, wherein the configuration for the direct acyclic graph is generated via a graphical user interface.
In Example 17, the subject matter of Examples 12-16 includes, wherein the top level dimension lacks upstream dimensions in the direct acyclic graph.
Example 18 is a system comprising: processing circuitry; and a memory storing instructions which, when executed by the processing circuitry, cause the processing circuitry to perform operations comprising: accessing an input dataset for computing an action including multiple decision items and a configuration for a directed acyclic graph comprising nodes representing dimensions, each dimension being associated with a decision item of the action; computing a value for a first decision item associated with a top level dimension of the directed acyclic graph, wherein the value for the first decision item is computed based on the input dataset and using a reinforcement learning agent for the top level dimension; computing a value for a second decision item associated with a non-top level dimension from the directed acyclic graph, wherein the value for the second decision item is computed based on the input dataset and a specified subset of outputs of reinforcement learning engines for dimensions upstream from the non-top level dimension in the directed acyclic graph and using a reinforcement learning agent for the non-top level dimension; and providing the computed action including at least the value for the first decision item and the value for the second decision item.
In Example 19, the subject matter of Example 18 includes, wherein the directed acyclic graph comprises unidirectional edges between the multiple nodes, wherein an upstream dimension being upstream from the non-top level dimension comprises an edge existing from a node of the upstream dimension to a node of the non-top level dimension.
In Example 20, the subject matter of Examples 18-19 includes, wherein the configuration for the direct acyclic graph is accessed via one or more configuration files.
Example 21 is at least one machine-readable medium including instructions that, when executed by processing circuitry, cause the processing circuitry to perform operations to implement of any of Examples 1-20.
Example 22 is an apparatus comprising means to implement of any of Examples 1-20.
Example 23 is a system to implement of any of Examples 1-20.
Example 24 is a method to implement of any of Examples 1-20.
Although an embodiment has been described with reference to specific example embodiments, it will be evident that various modifications and changes may be made to these embodiments without departing from the broader spirit and scope of the present disclosure. Accordingly, the specification and drawings are to be regarded in an illustrative rather than a restrictive sense. The accompanying drawings that form a part hereof show, by way of illustration, and not of limitation, specific embodiments in which the subject matter may be practiced. The embodiments illustrated are described in sufficient detail to enable those skilled in the art to practice the teachings disclosed herein. Other embodiments may be utilized and derived therefrom, such that structural and logical substitutions and changes may be made without departing from the scope of this disclosure. This Detailed Description, therefore, is not to be taken in a limiting sense, and the scope of various embodiments is defined only by the appended claims, along with the full range of equivalents to which such claims are entitled.
Although specific embodiments have been illustrated and described herein, it should be appreciated that any arrangement calculated to achieve the same purpose may be substituted for the specific embodiments shown. This disclosure is intended to cover any and all adaptations or variations of various embodiments. Combinations of the above embodiments, and other embodiments not specifically described herein, will be apparent to those of skill in the art upon reviewing the above description.
In this document, the terms “a” or “an” are used, as is common in patent documents, to include one or more than one, independent of any other instances or usages of “at least one” or “one or more.” In this document, the term “or” is used to refer to a nonexclusive or, such that “A or B” includes “A but not B,” “B but not A,” and “A and B,” unless otherwise indicated. In this document, the terms “including” and “in which” are used as the plain-English equivalents of the respective terms “comprising” and “wherein.” Also, in the following claims, the terms “including” and “comprising” are open-ended, that is, a system, user equipment (UE), article, composition, formulation, or process that includes elements in addition to those listed after such a term in a claim are still deemed to fall within the scope of that claim. Moreover, in the following claims, the terms “first,” “second,” and “third,” etc. are used merely as labels, and are not intended to impose numerical requirements on their objects.
The Abstract of the Disclosure is provided to comply with 37 C.F.R. § 1.72(b), requiring an abstract that will allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, it can be seen that various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Thus the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment.
