Professional Documents
Culture Documents
Abstract
The Jeshka community is building an open source ecosystem that max-
imizes the net worth of activities performed by humans and machines. The
ecosystem uses a directed acyclic graph network to efficiently interconnect
different networks, their machine learning services and users. The main
objective is to split activities/tasks into micro and nano tasks, so they can
be reused across a wider scope. A secondary goal is to enable learning
during completion.
In this paper we propose the Jeshka Internet of Tasks (JIoT), an in-
novative and distributed ecosystem designed to provide efficient collabo-
ration in near real-time.
This work implements, extends and makes the following contributions:
1. Efficient Distributed Activity Splitting (EDAS) - A machine learning
approach to identify the structure of activities and tasks to efficiently
split them into components that can be reused.
2. Efficient Distributed Activity Merging (EDAM) - A machine learn-
ing approach to merge the components created with EDAS to allow
parallel processing while securing privacy.
3. Proof-of-Reputation (PoR) - A consensus protocol that guarantees
quick response times and validity.
4. IOTA Identity of Things (IDoT) - Provides unique identities for all
participating resources.
5. IOTA Masked Authenticated Messaging (MAM) - Used to create
private data streams between processes/resources.
6. IOTA Oracles - Used to exchange processed results.
7. IOTA Flash Network - Payment Channels enable very rapid high
throughput of transactions.
8. IOTA Private Transactions - A layer to enable transactional privacy.
Note: Jeshka is a work in progress. Active research and testing is under
way and new versions of this paper will be updated at https: // jeshka. io
.
1
Contents
1 Real World Usage 3
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 JESH Token . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 IOTA Token . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Use Case Example . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Proof-of-Reputation 4
2.1 Reputation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Additional proof and loss of reputation . . . . . . . . . . . . . . . 4
2.3 Scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Transactions 5
3.1 Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Efficient Distributed Activity Splitting (EDAS) . . . . . . . . . . 5
3.3 Efficient Distributed Activity Merging (EDAM) . . . . . . . . . . 5
2
1 Real World Usage
1.1 Background
We are used to designing and using task/activity specific solutions. JIoT opens
up a revolutionary way to create and complete tasks, so that work is not only
done directly for the specific request, but across a network of tasks. Instead of
the current way, where alice performs an activity for bob, the activity performed
can be split up into many smaller activities that fulfill parts of multiple broader
tasks in parallel. Thereby maximizing the net worth of the activities performed,
because their result can be used by multiple parties at the same time.
3
2 Proof-of-Reputation
2.1 Reputation
The Proof-of-Reputation consensus protocol uses a highest reputation first ap-
proach. Reputation is not a single unit of measurement. Transactions are
grouped by content and different groups of transactions have different repu-
tation units. The protocol puts all processors on a spectrum of reputation.
High reputation processors are assigned with the processes first, high reputa-
tion processors then assign lower reputation processors to provide finality and
verifiability. The initiator has to vote on a result of another random network,
before the transaction is considered successful. Processors form mixed groups,
low reputation processors can gain reputation and help ensure the liveness of
the network. When a low reputation processors becomes a high reputation
processor, it inherently forms a new network with low reputation processors.
Reputation is increased by the number of validations from higher level reputa-
tion processors within the network. Because JIoT doesnt use sequential blocks,
this can take place in parallel. Processors can only earn reputation from high
reputation processors in the groups they join. Reputation is awarded based on
speed and validity of the transaction, as well as the quality of the distribution
of the transaction to lower level processors.
2.3 Scalability
With a growing number of transactions, there is a growing number of processors
to speed up the entire network. The amount of work done to verify a transaction
is not exponentially growing, but constantly optimized.
4
3 Transactions
The input flows towards lower reputation processors in the network to form
smaller and smaller partial inputs (chunks) and once processed sufficiently flows
to higher reputation processors to form larger and larger outputs (chunks).