bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 2쪽 중 2쪽
.0
,
−
f
i
(
x
)
,
0
,
...0
]
Note, when the error function
Q
j
\ Q_{j} \:
Q
j
is the twice-differentiable cross-entropy, then
H
(
Q
j
)
\ H(Q_{j}) \:
H
(
Q
j
)
is its Fisher- information matrix, and
r
i
∈
R
\ r_{i} \in R \:
r
i
∈
R
is suitably measured as each peer’s informational significance to the network as a whole. However, information theoretic weights require the full Hessian of the error. In practice it is more reasonable to use a heuristic to propagate a contribution score from the error function through to the inputs Yu et al. [2017]. For instance, weights from the gating layer (Section 6) provide a useful differentiable proxy.
10 Collusion
We consider the scenario where a subset of the peers in the network have formed a ’cabal’: A set of colluding peers attempting to maximize their inflation without accurately scoring their neighbors. The fight between the honest graph A with stake
S
A
\ S_{A}
S
A
and the disjoint cabal B with stake
S
B
S_{B}
S
B
can be determined by the proportion of network stake held by each. The honest graph must attain more inflation to maintain its dominance and protect the network
I
A
≫
I
B
\ I_{A} \gg I_{B}
I
A
≫
I
B
We assume that the proportion of stake in the honest graph is more than that found in the dishonest graph
S
A
≫
S
B
\ S_{A} \gg S_{B} \:
S
A
≫
S
B
and that the chain has reached consensus
L
<
0
\ \mathcal{L} < 0 \:
L
<
0
Since all peers in B are disjoint from A, our loss term
−
R
B
⋅
(
C
B
−
0.5
)
>
0
\ -R_{B} \cdot (C_{B} - 0.5) > 0 \:
−
R
B
⋅
(
C
B
−
0.5
)
>
0
is positive. Because
L
<
0
\ \mathcal{L} < 0 \:
L
<
0
it must be the case that
R
A
⋅
(
C
A
−
0.5
)
<
0
\ R_{A} \cdot (C_{A} - 0.5) < 0 \:
R
A
⋅
(
C
A
−
0.5
)
<
0
is negative and there are peers in the honest sub-graph A who are connected to the majority.
As the chain progresses, newly minted stake is being emitted at our inflation rate τ in proportion to I = R · T. Importantly, the gradient of the incentive function with respect to the stake is positive and super-linear at our inflection point between the honest and dishonest graph. Notably,
δ
I
δ
S
=
5
2
\ \frac{\delta I}{\delta S} = \frac{5}{2}
δ
S
δ
I
=
2
5
, this ensures that the amount of stake held by each sub-graph reflects a non-linear change in their inflation at the next iteration.
Initially, since
S
A
>
0.5
a
n
d
S
B
<
0.5
\ S_{A} > 0.5 \: and \: S_{B} < 0.5
S
A
>
0.5
an
d
S
B
<
0.5
the proportion of stake emitted in sub-graph A exceeds that in sub-graph B, and sub-graph A’s incentive grows super-linearly compared to B. The result is that the ratio of stake
S
B
S
A
+
S
B
\ \frac{S_{B}}{S_{A} + S_{B}} \:
S
A
+
S
B
S
B
decreases – the cabal must continually add stake to its sub-graph to maintain itself through time.
We consider this proportion between the competing graphs under continuous inflation. Converting to python code ...
11 Conclusion
We have proposed an intelligence market that runs on a P2P network outside of a trusted environment. Crucially, the benchmark measures performance as representational knowledge production using other intelligence systems to determine its value. The fact that this can be done in a collaborative and high-resolution manner suggests that the benchmark could provide a better reward mechanism for the field in general.
To achieve this aim, the paper began with the definition of a P2P network composed of abstractly defined intelligence models. We showed how this framework allowed us to produce a ranking for each peer based on the cost to prune it from the network. Peers negotiated this score using a set of weights on a digital ledger. However, the system was incomplete without mechanisms that prevented participants from forming dishonest sub-graphs.
To resolve this, we proposed an incentive scheme based on peer connectivity which exponentially rewarded peers for being trusted by a large portion of the network. This ensured that over time dishonest sub-graphs decay to irrelevance.
Following this, we showed (`
1
`) how peers reduced the network bandwidth by learning connectivity using a differential layer and (`
2
`) how they could extract fully network-disconnected machine learning models to run in production. The result is an intelligence market that rewards participants for producing knowledge and making it available to new learners in the system.
Follow this link for the original version프로젝트로 돌아가기