Commit 951ff42

committed

code update

1 parent 1147eb6 commit 951ff42Copy full SHA for 951ff42

File tree

+12

-6

lines changed

+12

-6

lines changed

Lines changed: 6 additions & 3 deletions

Original file line number	Diff line number	Diff line change
`@@ -649,6 +649,7 @@`
`649`	`649`	`"\n",`
`650`	`650`	`"$$v_t = \\gamma v_{t-1} + \\eta \\nabla_\\theta J( \\theta) -- (8) $$\n",`
`651`	`651`	`"\n",`
	`652`	`+ "Update theta by: \n",`
`652`	`653`	`"$${\\theta = \\theta - v_t } -- (9) $$"`
`653`	`654`	`]`
`654`	`655`	`},`
`@@ -763,7 +764,7 @@`
`763`	`764`	`"\n",`
`764`	`765`	`"$$v_t = \\gamma v_{t-1} + \\eta \\nabla_\\theta J( \\theta - \\gamma v_{t-1} ) -- (10) $$\n",`
`765`	`766`	`"\n",`
`766`		`- "\n",`
	`767`	`+ "Update theta by:\n",`
`767`	`768`	`"$${\\theta = \\theta - v_t } -- (11) $$"`
`768`	`769`	`]`
`769`	`770`	`},`
`@@ -980,14 +981,16 @@`
`980`	`981`	`"\n",`
`981`	`982`	`"The update equation of AdaDelta is expressed as:\n",`
`982`	`983`	`"\n",`
	`984`	`+ "Running average of gradients:\n",`
`983`	`985`	`"$$E\\left[g^{2}\\right]_{t}=\\gamma E\\left[g^{2}\\right]_{t-1}+(1-\\gamma) g_{t}^{2} -- (13) $$\n",`
`984`	`986`	`"\n",`
	`987`	`+ "Compute delta:\n",`
`985`	`988`	`"$$ \\nabla \\theta_{t}=-\\frac{R M S[\\Delta \\theta]_{t-1}}{R M S\\left[g_{t}\\right]} \\cdot g_{t}^{i} -- (14) $$\n",`
`986`	`989`	`"\n",`
`987`		`- "\n",`
	`990`	`+ "Running averge of parameter update:\n",`
`988`	`991`	`"$$ E\\left[\\Delta \\theta^{2}\\right]_{t}=\\gamma E\\left[\\Delta \\theta^{2}\\right]_{t-1}+(1-\\gamma) \\Delta \\theta_{t}^{2} -- (15) $$\n",`
`989`	`992`	`"\n",`
`990`		`- "\n",`
	`993`	`+ "Update theta:\n",`
`991`	`994`	`"$$ \\theta_{t}^{i}=\\theta_{t-1}^{i}+\\nabla \\theta_{t} -- (16) $$"`
`992`	`995`	`]`
`993`	`996`	`},`

Lines changed: 6 additions & 3 deletions

Original file line number	Diff line number	Diff line change
`@@ -649,6 +649,7 @@`
`649`	`649`	`"\n",`
`650`	`650`	`"$$v_t = \\gamma v_{t-1} + \\eta \\nabla_\\theta J( \\theta) -- (8) $$\n",`
`651`	`651`	`"\n",`
	`652`	`+ "Update theta by: \n",`
`652`	`653`	`"$${\\theta = \\theta - v_t } -- (9) $$"`
`653`	`654`	`]`
`654`	`655`	`},`
`@@ -763,7 +764,7 @@`
`763`	`764`	`"\n",`
`764`	`765`	`"$$v_t = \\gamma v_{t-1} + \\eta \\nabla_\\theta J( \\theta - \\gamma v_{t-1} ) -- (10) $$\n",`
`765`	`766`	`"\n",`
`766`		`- "\n",`
	`767`	`+ "Update theta by:\n",`
`767`	`768`	`"$${\\theta = \\theta - v_t } -- (11) $$"`
`768`	`769`	`]`
`769`	`770`	`},`
`@@ -980,14 +981,16 @@`
`980`	`981`	`"\n",`
`981`	`982`	`"The update equation of AdaDelta is expressed as:\n",`
`982`	`983`	`"\n",`
	`984`	`+ "Running average of gradients:\n",`
`983`	`985`	`"$$E\\left[g^{2}\\right]_{t}=\\gamma E\\left[g^{2}\\right]_{t-1}+(1-\\gamma) g_{t}^{2} -- (13) $$\n",`
`984`	`986`	`"\n",`
	`987`	`+ "Compute delta:\n",`
`985`	`988`	`"$$ \\nabla \\theta_{t}=-\\frac{R M S[\\Delta \\theta]_{t-1}}{R M S\\left[g_{t}\\right]} \\cdot g_{t}^{i} -- (14) $$\n",`
`986`	`989`	`"\n",`
`987`		`- "\n",`
	`990`	`+ "Running averge of parameter update:\n",`
`988`	`991`	`"$$ E\\left[\\Delta \\theta^{2}\\right]_{t}=\\gamma E\\left[\\Delta \\theta^{2}\\right]_{t-1}+(1-\\gamma) \\Delta \\theta_{t}^{2} -- (15) $$\n",`
`989`	`992`	`"\n",`
`990`		`- "\n",`
	`993`	`+ "Update theta:\n",`
`991`	`994`	`"$$ \\theta_{t}^{i}=\\theta_{t-1}^{i}+\\nabla \\theta_{t} -- (16) $$"`
`992`	`995`	`]`
`993`	`996`	`},`

Comments

(0)